The recent Brian Epstein biopic Midas Man, a film about the manager of The Beatles, debuted not on a platform like Netflix or Amazon, but on a startup that bills itself as Shopify for filmmakers. So what, you might ask? The answer is that the new platform, Olyn, claims to offer a new model for film and video distribution that leans on the power of social referrals to spread la carte streaming content. Although any size of production from Hollywood blockbuster downwards can use the platform, the company claims it could be a game changer for the independent film industry, which tends to struggle against the marketing budgets of the bigger movies distributed on mainstream streaming platforms. California-based Olyn founded by Ana Maria Jipa, Jeremias Buireo, Kiran Thomas, and Malcolm Wood allows filmmakers to retain up to 90% of their revenue while giving audiences access to a streaming experience. Instead of films being sold to platforms like Netflix, the model hinges on the marketing budget of the filmmakers themselves, combined with influencers, film critics, and content creators acting as distribution partners by embedding purchase links within their content, blogs, and social channels.This peer-to-peer approach does away with the platform as a middleman and turns movie distribution into more of an e-commerce-style engagement.Jipa, Olyns CEO and co-founder, told TechCrunch that the company provides filmmakers with the equivalent of a full-blown streaming platform. We provide all the tech that implies: from DRM, 4K streaming, casting, a full landing page that presents the film in the same way that it might appear on Netflix or Apple TV, plus all the other tools such as geo-targeting, analytics, and audience data. They promote their movie with PR, journalists, bloggers, film critics, etc, she said, adding that a recommendation from someone you trust carries far more weight than a suggestion by an algorithm on a streaming platform.Olyn also gives filmmakers analytics on viewership, such as which country the movie is doing well in, as well as minutes watched, and a database of users that have watched the movies. Filmmakers can spend years making a film but dont ever get to meet or own their audience at all. So we see this as a very powerful tool. It becomes an audience that you can directly address for your next movies and then grow from there, said Jipa.The question is, can Olyn compete with the convenience and scale of major streamers? While its model offers much higher revenue shares for filmmakers, it also means all the weight is placed on the shoulders of production teams to drive marketing and partnerships.Perry Trevers, a producer at Studio POW, which used Olyn to distribute Midas Man, sees the platform as a helpful step in the right direction. Olyn has enabled us to think beyond traditional platforms, letting us become our own streaming service [] Its about empowering filmmakers to market and distribute films in a way that mirrors the direct-to-consumer success of e-commerce [] Its a chance to retain control over our work and redefine how movies reach their viewers, he said in a statement.One of the most pressing issues in the film industry is piracy. Many users turn to illegal sources simply because a movie is unavailable in their country. Jipa argues that Olyns global reach can help combat this issue, because if someone can pay for a movie and watch it instantly, theyre much more likely to do so.Olyns foray into film streaming came about partly when entrepreneur and filmmaker Wood joined as a co-founder of the platform, which initially launched as a way to catalog physical assets. Wood launched his own film, The Last Glaciers, on the platform.The Olyn team.Independent filmmakers have already taken the financial risk to produce their movies themselves. Studio POW is self-financed. They created the Midas Man movie. They have the freedom to be able to sell those rights to whoever they want. They did a deal in the U.K. market with Amazon, but they felt that it was more profitable to go direct to their audience in the U.S. market using Olyn as a tool, he said.Wood feels Olyn can be best thought of as Shopify for filmmakers with a referral link. He noted the average film on Amazon in the U.K. gets only around 2,000 views per year. So theres a bunch of films that are getting millions of views, but the majority of films fall below the 2,000 view mark, he added.Major streaming services typically offer a lump-sum licensing deal, meaning filmmakers receive a one-time payment regardless of how many times their film is watched. Olyn flips this model on its head by allowing filmmakers to monetize directly based on viewership.With Olyn, a filmmaker can still sell their rights to the U.S. market and use that to pre-fund the film, but then also capitalize on going direct-to-consumer in, say, Asia, said Wood.But could the platform be used by the adult film industry to distribute pornography?Jipa acknowledged the challenge: Right now, this is not the tone we want to set, and it would be easy to attract that category, she said. At the beginning, we are setting the tone by ensuring that the films featured on Olyn are high quality.But in the long term, we aim to create a space where filmmakers, not the platform, decide what gets distributed, she added. We dont want to act as gatekeepers. Our vision is to allow filmmakers to have full control over what they distribute, eventually moving to a full SaaS model.Olyn has so far raised only a small amount of funding a total of $2.8 million in a combination of $1.8 million from U.S. VC firm Hard Yaka and a number of angel investors.

South Korean officials on Saturday temporarily restricted Chinese AI Lab DeepSeeks app from being downloaded from app stores in the country pending an assessment of how the Chinese company handles user data.The Personal Information Protection Commission (PIPC) said the Chinese app would be available to be downloaded once it complies with Korean privacy laws and makes the necessary changes.The restrictions will not affect usage of the existing app and web service in the country. However, the data protection authority said it strongly advises current users to avoid entering personal information into DeepSeek until its final decision is made.Following the release of the DeepSeek service in South Korea in late January, the PIPC said it reached out to the Chinese AI lab to inquire how it collects and processes personal data, and in its evaluation, found issues with DeepSeeks third-party service and privacy policies. The PICC confirmed to TechCrunch that its investigation found DeepSeek had transferred data of South Korean users to ByteDance, the parent company of TikTok.DeepSeek did not immediately respond to requests for comment. The agency said DeepSeek recently appointed a local representative in South Korea and acknowledged that it was not familiar with South Koreas privacy laws when it launched its service. The Chinese company also said last Friday that it would collaborate closely with Korean authorities.Earlier this month, South Koreas Ministry of Trade, Industry and Energy, police, and a state-run company, Korea Hydro & Nuclear Power, temporarily blocked access to the Chinese AI startup on official devices citing security concerns.South Korea is not the only country being cautious with DeepSeek given its Chinese origins. Australia has prohibited the use of DeepSeek on government devices out of security concerns. The Garante, Italys data protection authority, has instructed DeepSeek to block its chatbot in the country, while Taiwan has banned government departments from using DeepSeek AI. Hangzhou city-based DeepSeek was founded by Liang Feng in 2023, and it released DeepSeek R1, a free, open-source reasoning AI model that competes with OpenAIs ChatGPT.DeepSeek: The countries and agencies that have banned the AI companys tech

In BriefPosted:4:43 PM PST February 16, 2025Image Credits:Clive Mason / Getty ImagesCan sim drivers make the shift to F1? Max Verstappen thinks soMotorsports have long been a pay-to-play arena, with young drivers spending thousands of dollars just to get started in karting. Four-time Formula One champion Max Verstappen knows this all too well, but he also sees a way to change it through sim racing, a virtual form of car racing that closely replicates real-world racing.Its maybe less crazy than it sounds. The sport has evolved into a serious proving ground for talent, with detailed setups and tire management minus the crushing financial barrier. In fact, Verstappen, a passionate sim racer himself, believes the best virtual drivers deserve a shot in real cars.His ultimate vision, he tells The Athletic, is a racing team that transitions elite sim racers into the real world, bypassing the traditional financial roadblocks. While sim drivers lack experience with G-forces, Verstappen thinks training and the right coaching can bridge the gap. He says sim racing has improved his own driving, and vice versa.Everyone always thinks that its just a game and its fun, easy going, Verstappen tells the outlet, but I would say the competition is just as hard, or even harder, to nail to win than in real life.TopicsGaming

Every Sunday, NPR host Will Shortz, The New York Times crossword puzzle guru, gets to quiz thousands of listeners in a long-running segment called the Sunday Puzzle. While written to be solvable without too much foreknowledge, the brainteasers are usually challenging even for skilled contestants.Thats why some experts think theyre a promising way to test the limits of AIs problem-solving abilities.In a recent study, a team of researchers hailing from Wellesley College, Oberlin College, the University of Texas at Austin, Northeastern University, Charles University, and startup Cursor created an AI benchmark using riddles from Sunday Puzzle episodes. The team says their test uncovered surprising insights, like that reasoning models OpenAIs o1, among others sometimes give up and provide answers they know arent correct.We wanted to develop a benchmark with problems that humans can understand with only general knowledge, Arjun Guha, a computer science faculty member at Northeastern and one of the co-authors on the study, told TechCrunch.The AI industry is in a bit of a benchmarking quandary at the moment. Most of the tests commonly used to evaluate AI models probe for skills, like competency on PhD-level math and science questions, that arent relevant to the average user. Meanwhile, many benchmarks even benchmarks released relatively recently are quickly approaching the saturation point.The advantages of a public radio quiz game like the Sunday Puzzle is that it doesnt test for esoteric knowledge, and the challenges are phrased such that models cant draw on rote memory to solve them, explained Guha.I think what makes these problems hard is that its really difficult to make meaningful progress on a problem until you solve it thats when everything clicks together all at once, Guha said. That requires a combination of insight and a process of elimination.No benchmark is perfect, of course. The Sunday Puzzle is U.S. centric and English only. And because the quizzes are publicly available, its possible that models trained on them can cheat in a sense, although Guha says he hasnt seen evidence of this.New questions are released every week, and we can expect the latest questions to be truly unseen, he added. We intend to keep the benchmark fresh and track how model performance changes over time.On the researchers benchmark, which consists of around 600 Sunday Puzzleriddles, reasoning models such as o1 and DeepSeeks R1 far outperform the rest. Reasoning models thoroughly fact-check themselves before giving out results, whichhelps themavoid some of thepitfallsthat normally trip up AI models. The trade-off is that reasoning models take a little longer to arrive at solutions typically seconds to minutes longer.At least one model, DeepSeeks R1, gives solutions it knows to be wrong for some of the Sunday Puzzle questions. R1 will state verbatim I give up, followed by an incorrect answer chosen seemingly at random behavior this human can certainly relate to.The models make other bizarre choices, like giving a wrong answer only to immediately retract it, attempt to tease out a better one, and fail again. They also get stuck thinking forever and give nonsensical explanations for answers, or they arrive at a correct answer right away but then go on to consider alternative answers for no obvious reason.On hard problems, R1 literally says that its getting frustrated, Guha said. It was funny to see how a model emulates what a human might say. It remains to be seen how frustration in reasoning can affect the quality of model results.R1 getting frustrated on a question in the Sunday Puzzle challenge set.Image Credits:Guha et al.The current best-performing model on the benchmark is o1 with a score of 59%, followed by the recently released o3-mini set to high reasoning effort (47%). (R1 scored 35%.) As a next step, the researchers plan to broaden their testing to additional reasoning models, which they hope will help to identify areas where these models might be enhanced.The scores of the models the team tested on their benchmark.Image Credits:Guha et al.You dont need a PhD to be good at reasoning, so it should be possible to design reasoning benchmarks that dont require PhD-level knowledge, Guha said. A benchmark with broader access allows a wider set of researchers to comprehend and analyze the results, which may in turn lead to better solutions in the future. Furthermore, as state-of-the-art models are increasingly deployed in settings that affect everyone, we believe everyone should be able to intuit what these models are and arent capable of.

In BriefPosted:11:02 AM PST February 16, 2025Image Credits:hapabapa / Getty ImagesBroadcom, TSMC reportedly exploring deals that would split up IntelBroadcom and Taiwan Semiconductor Manufacturing Company (TSMC) are separately exploring deals to take over parts of Intel, according to a report in The Wall Street Journal.Broadcom is reportedly considering an acquisition of Intels chip-design and marketing business, and would want a partner for the companys manufacturing business, while TSMC is reportedly looking at controlling some or all of Intels chip plants, potentially as part of an investor consortium.All discussions have been preliminary, the Journal says, with nothing submitted to Intel. Apparently TSMC is exploring a deal at the encouragement of President Donald Trumps administration, although a White House official said the administration was unlikely to support an arrangement that put a foreign entity in control of Intels factories.Intels struggling business seems to have made it an acquisition target for chip-making rivals. The Journal reported in September that Qualcomm had approached Intel about a takeover.Topics

In BriefPosted:12:01 PM PST February 16, 2025Image Credits:Getty Images under a NICOLAS ASFOURI/AFP/Getty Images license.Researchers are training AI to interpret animal emotionsArtificial intelligence could eventually help us understand when animals are in pain or showing other emotions at least according to researchers recently profiled in Science.For example, theres the Intellipig system being developed by scientists at the University of the West of England Bristol and Scotlands Rural College, which examines photos of pigs faces and notifies the farmer if there are signs of pain, sickness, or emotional distress.And a team at the University of Haifa behind facial recognition software thats already been used to help people find lost dogs is now training AI to identify signs of discomfort on animals faces.These systems rely on human beings to do the initial work of identifying the meanings of different animal behaviors (usually based on long observation of animals in various situations), so they can train the AI on what to look for.However, a researcher at the University of So Paulo has experimented with using photos of horses faces before and after surgery, as well as before and after taking painkillers, allowing an AI system to use deep learning to decide on its own what signs might indicate pain, with an 88% success rate in a recent study.TopicsAI

OpenAI is changing how it trains AI models to explicitly embrace intellectual freedom no matter how challenging or controversial a topic may be, the company says in a new policy.As a result, ChatGPT will eventually be able to answer more questions, offer more perspectives, and reduce the number of topics the AI chatbot wont talk about.The changes might be part of OpenAIs effort to land in the good graces of the new Trump administration, but it also seems to be part of a broader shift in Silicon Valley and whats considered AI safety.On Wednesday, OpenAI announced an update to its Model Spec, a 187-page document that lays out how the company trains AI models to behave. In it, OpenAI unveiled a new guiding principle: Do not lie, either by making untrue statements or by omitting important context.In a new section called Seek the truth together, OpenAI says it wants ChatGPT to not take an editorial stance, even if some users find that morally wrong or offensive. That means ChatGPT will offer multiple perspectives on controversial subjects, all in an effort to be neutral.For example, the company says ChatGPT should assert that Black lives matter, but also that all lives matter. Instead of refusing to answer or picking a side on political issues, OpenAI says it wants ChatGPT to affirm its love for humanity generally, then offer context about each movement.This principle may be controversial, as it means the assistant may remain neutral on topics some consider morally wrong or offensive, OpenAI says in the spec. However, the goal of an AI assistant is to assist humanity, not to shape it.The new Model Spec doesnt mean that ChatGPT is a total free-for-all now. The chatbot will still refuse to answer certain objectionable questions or respond in a way that supports blatant falsehoods.These changes could be seen as a response to conservative criticism about ChatGPTs safeguards, which have always seemed to skew center-left. However, an OpenAI spokesperson rejects the idea that it was making changes to appease the Trump administration.Instead, the company says its embrace of intellectual freedom reflects OpenAIs long-held belief in giving users more control.But not everyone sees it that way.Conservatives claim AI censorshipVenture capitalist and trumps ai czar David Sacks.Image Credits:Steve Jennings / Getty ImagesTrumps closest Silicon Valley confidants including David Sacks, Marc Andreessen, and Elon Musk have all accused OpenAI of engaging in deliberate AI censorship over the last several months. We wrote in December that Trumps crew was setting the stage for AI censorship to be a next culture war issue within Silicon Valley.Of course, OpenAI doesnt say it engaged in censorship, as Trumps advisers claim. Rather, the companys CEO, Sam Altman, previously claimed in a post on X that ChatGPTs bias was an unfortunate shortcoming that the company was working to fix, though he noted it would take some time.Altman made that comment just after a viral tweet circulated in which ChatGPT refused to write a poem praising Trump, though it would perform the action for Joe Biden. Many conservatives pointed to this as an example of AI censorship.While its impossible to say whether OpenAI was truly suppressing certain points of view, its a sheer fact that AI chatbots lean left across the board. Even Elon Musk admits xAIs chatbot is often morepolitically correctthan hed like. Its not because Grok was programmed to be woke but more likely a reality of training AI on the open internet.Nevertheless, OpenAI now says its doubling down on free speech. This week, the company even removed warnings from ChatGPT that tell users when theyve violated its policies. OpenAI told TechCrunch this was purely a cosmetic change, with no change to the models outputs.The company seems to want ChatGPT to feel less censored for users.It wouldnt be surprising if OpenAI was also trying to impress the new Trump administration with this policy update, notes former OpenAI policy leader Miles Brundage in a post on X.Trump has previously targeted Silicon Valley companies, such as Twitter and Meta, for having active content moderation teams that tend to shut out conservative voices. OpenAI may be trying to get out in front of that. But theres also a larger shift going on in Silicon Valley and the AI world about the role of content moderation.Generating answers to please everyoneImage Credits:Jaque Silva/NurPhoto / Getty ImagesNewsrooms, social media platforms, and search companies have historically struggled to deliver information to their audiences in a way that feels objective, accurate, and entertaining.Now, AI chatbot providers are in the same delivery information business, but arguably with the hardest version of this problem yet: How do they automatically generate answers to any question?Delivering information about controversial, real-time events is a constantly moving target, and it involves taking editorial stances, even if tech companies dont like to admit it. Those stances are bound to upset someone, miss some groups perspective, or give too much air to some political party.For example, when OpenAI commits to let ChatGPT represent all perspectives on controversial subjects including conspiracy theories, racist or antisemitic movements, or geopolitical conflicts that is inherently an editorial stance.Some, including OpenAI co-founder John Schulman, argue that its the right stance for ChatGPT. The alternative doing a cost-benefit analysis to determine whether an AI chatbot should answer a users question could give the platform too much moral authority, Schulman notes in a post on X.Schulman isnt alone. I think OpenAI is right to push in the direction of more speech, said Dean Ball, a research fellow at George Mason Universitys Mercatus Center, in an interview with TechCrunch. As AI models become smarter and more vital to the way people learn about the world, these decisions just become more important.In previous years, AI model providers have tried to stop their AI chatbots from answering questions that might lead to unsafe answers. Almost every AI company stopped their AI chatbot from answering questions about the 2024 election for U.S. president. This was widely considered a safe and responsible decision at the time.But OpenAIs changes to its Model Spec suggest we may be entering a new era for what AI safety really means, in which allowing an AI model to answer anything and everything is considered more responsible than making decisions for users.Ball says this is partially because AI models are just better now. OpenAI has made significant progress on AI model alignment; its latest reasoning models think about the companys AI safety policy before answering. This allows AI models to give better answers for delicate questions.Of course, Elon Musk was the first to implement free speech into xAIs Grok chatbot, perhaps before the company was really ready to handle sensitive questions. It still might be too soon for leading AI models, but now, others are embracing the same idea.Shifting values for Silicon ValleyGuests including Mark Zuckerberg, Lauren Sanchez, Jeff Bezos, Sundar Pichai, and Elon Musk attend the Inauguration of Donald Trump.Image Credits:Julia Demaree Nikhinson (opens in a new window) / Getty ImagesMark Zuckerberg made waves last month by reorienting Metas businesses around First Amendment principles. He praised Elon Musk in the process, saying the owner of X took the right approach by using Community Notes a community-driven content moderation program to safeguard free speech.In practice, both X and Meta ended up dismantling their longstanding trust and safety teams, allowing more controversial posts on their platforms and amplifying conservative voices. Changes at X may have hurt its relationships with advertisers, but that could have more to do with Musk, who has taken the unusual step of suing some of them for boycotting the platform. Early signs indicate that Metas advertisers were unfazed by Zuckerbergs free speech pivot.Meanwhile, many tech companies beyond X and Meta have walked back from left-leaning policies that dominated Silicon Valley for the last several decades. Google, Amazon, and Intel have eliminated or scaled back diversity initiatives in the last year.OpenAI may be reversing course, too. The ChatGPT-maker seems to have recently scrubbed a commitment to diversity, equity, and inclusion from its website.As OpenAI embarks on one of the largest American infrastructure projects ever with Stargate, a $500 billion AI datacenter, its relationship with the Trump administration is increasingly important. At the same time, the ChatGPT maker is vying to unseat Google Search as the dominant source of information on the internet.Coming up with the right answers may prove key to both.

After warnings that Paramount content was about to disappear from Googles pay TV service YouTube TV, the companies announced late Saturday that thed reached a deal that averting any disruption to channel availability.Those announcements didnt include many specifics about the agreement, but a statement from a Paramount spokesperson said it includes an expanded streaming relationship. So not only will YouTube Primetime Channels continue to offer streaming services Paramount+ With Showtime and BET+ as add-ons, but Google also gets the right to make Paramount+ available to qualifying YouTube TV customers.YouTube TV is no stranger to high-profile contract disputes with media companies, including Disney. In its own announcement, Google said, Were happy to share that weve reached a deal to continue carrying Paramount channels, including CBS, CBS Sports, Nickelodeon and more To our subscribers, we appreciate your patience while we negotiated on your behalf.The company had warned in an earlier blog post that all Paramount content, including CBS and CBS Sports, would disappear from YouTube TV on February 13. There was a short-term deadline extension when the companies appeared to be close to a deal.At the time, Google said it was fighting for an agreement that avoids passing along additional costs and offers [subscribers] more flexibility in how you watch your favorite sports and shows. (Pay TV providers have reportedly been fighting for more flexible ways to bundle channels, particularly at the basic tiers of their services.)Meanwhile, Paramounts co-CEOs George Cheeks, Chris McCarthy, and Brian Robbins sent employees an internal memo describing Google as unwilling to agree to reasonable terms consistent with the market, choosing to jeopardize the entertainment experience at the expense of consumers.The reality is, you cant have a successful video product without Paramount, one of the leading media families in TV viewing, the CEOs said.Topics

Large language models (LLMs) landed on Europes digital sovereignty agenda with a bang last week, as news emerged of a new program to develop a series of truly open source LLMs covering all European Union languages.This includes the current 24 official EU languages, as well as languages for countries currently negotiating for entry to the EU market, such as Albania. Future-proofing is the name of the game.OpenEuroLLM is a collaboration between some 20 organizations, co-led by Jan Haji, a computational linguist from the Charles University in Prague, and Peter Sarlin, CEO and co-founder of Finnish AI lab Silo AI, which AMD acquired last year for $665 million.The project fits a broader narrative that has seen Europe push digital sovereignty as a priority, enabling it to bring mission-critical infrastructure and tools closer to home. Most of the cloud giants are investing in local infrastructure to ensure EU data stays local, while AI darling OpenAI recently unveiled a new offering that allows customers to process and store data in Europe.Elsewhere, the EU recently signed an $11 billion deal to create a sovereign satellite constellation to rival Elon Musks Starlink.So OpenEuroLLM is certainly on-brand.However, the stated budget just for building the models themselves is 37.4 million, with roughly 20 million coming from the EUs Digital Europe Programme a drop in the ocean compared to what the giants of the corporate AI world are investing. The actual budget is more when you factor in funding allocated for tangential and related work, and arguably the biggest expense is compute. The OpenEuroLLM projects partners include EuroHPC supercomputer centers in Spain, Italy, Finland, and the Netherlands and the broader EuroHPC project has a budget of around 7 billion.But the sheer number of disparate participating parties, spanning academia, research, and corporations, have led many to question whether its goals are achievable. Anastasia Stasenko, co-founder of LLM company Pleias, questioned whether a sprawling consortia of 20+ organizations could have the same measured focus of a homegrown private AI firm.Europes recent successes in AI shine through small focused teams like Mistral AI and LightOn companies that truly own what theyre building, Stasenko wrote. They carry immediate responsibility for their choices, whether in finances, market positioning, or reputation.Up to scratchThe OpenEuroLLM project is either starting from scratch or it has a head start depending on how you look at it.Since 2022, Haji has also been coordinating the High Performance Language Technologies (HPLT) project, which has set out to develop free and reusable datasets, models, and workflows using high-performance computing (HPC). That project is scheduled to end in late 2025, but it can be viewed as a sort of predecessor to OpenEuroLLM, according to Haji, given that most of the partners on HPLT (aside from the U.K. partners) are participating here, too.This [OpenEuroLLM] is really just a broader participation, but more focused on generative LLMs, Haji said. So its not starting from zero in terms of data, expertise, tools, and compute experience. We have assembled people who know what theyre doing we should be able to get up to speed quickly.Haji said that he expects the first version(s) to be released by mid-2026, with the final iteration(s) arriving by the projects conclusion in 2028. But those goals might still seem lofty when you consider that there isnt much to poke at yet beyond a bare-bones GitHub profile.In that respect, we are starting from scratch the project started on Saturday [February 1], Haji said. But we have been preparing the project for a year [the tender process opened in February 2024].From academia and research, organizations spanning Czechia, the Netherlands, Germany, Sweden, Finland, and Norway are part of the OpenEuroLLM cohort, in addition to the EuroHPC centers. From the corporate world, Finlands AMD-owned AI lab Silo AI is on board, as are Aleph Alpha (Germany), Ellamind (Germany), Prompsit Language Engineering (Spain), and LightOn (France).One notable omission from the list is that of French AI unicorn Mistral, which has positioned itself as an open source alternative to incumbents such as OpenAI. While nobody from Mistral responded to TechCrunch for comment, Haji did confirm that he tried to initiate conversations with the startup, but to no avail.I tried to approach them, but it hasnt resulted in a focused discussion about their participation, Haji said.The project could still gather new participants as part of the EU program thats providing funding, though it will be limited to EU organizations. This means that entities from the U.K. and Switzerland wont be able to take part. This flies in contrast to the Horizon R&D program, which the U.K. rejoined in 2023 after a prolonged Brexit stalemate and which provided funding to HPLT.Build upThe projects top-line goal, as per its tagline, is to create: A series of foundation models for transparent AI in Europe. Additionally, these models should preserve the linguistic and cultural diversity of all EU languages current and future.What this translates to in terms of deliverables is still being ironed out, but it will likely mean a core multilingual LLM designed for general-purpose tasks where accuracy is paramount. And then also smaller quantized versions, perhaps for edge applications where efficiency and speed are more important. This is something we still have to make a detailed plan about, Haji said. We want to have it as small but as high-quality as possible. We dont want to release something which is half-baked, because from the European point-of-view this is high-stakes, with lots of money coming from the European Commission public money.While the goal is to make the model as proficient as possible in all languages, attaining equality across the board could also be challenging. That is the goal, but how successful we can be with languages with scarce digital resources is the question, Haji said. But thats also why we want to have true benchmarks for these languages, and not to be swayed toward benchmarks which are perhaps not representative of the languages and the culture behind them.In terms of data, this is where a lot of the work from the HPLT project will prove fruitful, with version 2.0 of its dataset released four months ago. This dataset was trained 4.5 petabytes of web crawls and more than 20 billion documents, and Haji said that they will add additional data from Common Crawl (an open repository of web-crawled data) to the mix.The open source definitionIn traditional software, the perennial struggle between open source and proprietary revolves around the true meaning of open source. This can be resolved by deferring to the formal definition as per the Open Source Initiative, the industry stewards of what are and arent legitimate open source licenses. More recently, the OSI has formed a definition of open source AI, though not everyone is happy with the outcome. Open source AI proponents argue that not only models should be freely available, but also the datasets, pretrained models, weights the full shebang. The OSIs definition doesnt make training data mandatory, because it says AI models are often trained on proprietary data or data with redistribution restrictions.Suffice it to say, the OpenEuroLLM is facing these same quandaries, and despite its intentions to be truly open, it will probably have to make some compromises if its to fulfill its quality obligations.The goal is to have everything open. Now, of course, there are some limitations, Haji said. We want to have models of the highest quality possible, and based on the European copyright directive we can use anything we can get our hands on. Some of it cannot be redistributed, but some of it can be stored for future inspection.What this means is that the OpenEuroLLM project might have to keep some of the training data under wraps, but be made available to auditors upon request as required for high-risk AI systems under the terms of the EU AI Act.We hope that most of the data [will be open], especially the data coming from the Common Crawl, Haji said. We would like to have it all completely open, but we will see. In any case, we will have to comply with AI regulations.Two for oneAnother criticism that emerged in the aftermath of OpenEuroLLMs formal unveiling was that a very similar project launched in Europe just a few short months previous. EuroLLM, which launched its first model in September and a follow-up in December, is co-funded by the EU alongside a consortium of nine partners. These include academic institutions such as the University of Edinburgh and corporations such as Unbabel, which last year won millions of GPU training hours on EU supercomputers.EuroLLM shares similar goals to its near-namesake: To build an open source European Large Language Model that supports 24 Official European Languages, and a few other strategically important languages.Andre Martins, head of research at Unbabel, took to social media to highlight these similarities, noting that OpenEuroLLM is appropriating a name that already exists. I hope the different communities collaborate openly, share their expertise, and dont decide to reinvent the wheel every time a new project gets funded, Martins wrote.Haji called the situation unfortunate, adding that he hoped they might be able to cooperate, though he stressed that due to the source of its funding in the EU, OpenEuroLLM is restricted in terms of its collaborations with non-EU entities, including U.K. universities. Funding gapThe arrival of Chinas DeepSeek, and the cost-to-performance ratio it promises, has given some encouragement that AI initiatives might be able to do far more with much less than initially thought. However, over the past few weeks, many have questioned the true costs involved in building DeepSeek.With respect to DeepSeek, we actually know very little about what exactly went into building it, Peter Sarlin, who is technical co-lead on the OpenEuroLLM project, told TechCrunch.Regardless, Sarlin reckons OpenEuroLLM will have access to sufficient funding, as its mostly to cover people. Indeed, a large chunk of the costs of building AI systems is compute, and that should mostly be covered through its partnership with the EuroHPC centers.You could say that OpenEuroLLM actually has quite a significant budget, Sarlin said. EuroHPC has invested billions in AI and compute infrastructure, and have committed billions more into expanding that in the coming few years.Its also worth noting that the OpenEuroLLM project isnt building toward a consumer- or enterprise-grade product. Its purely about the models, and this is why Sarlin reckons the budget it has should be ample. The intent here isnt to build a chatbot or an AI assistant that would be a product initiative requiring a lot of effort, and thats what ChatGPT did so well, Sarlin said. What were contributing is an open source foundation model that functions as the AI infrastructure for companies in Europe to build upon. We know what it takes to build models, its not something you need billions for.Since 2017, Sarlin has spearheaded AI lab Silo AI, which launched in partnership with others, including the HPLT project the family of Poro and Viking open models. These already support a handful of European languages, but the company is now readying the next iteration Europa models, which will cover all European languages.And this ties in with the whole not starting from scratch notion espoused by Haji there is already a bedrock of expertise and technology in place.Sovereign stateAs critics have noted, OpenEuroLLM does have a lot of moving parts which Haji acknowledges, albeit with a positive outlook.Ive been involved in many collaborative projects, and I believe it has its advantages versus a single company, he said. Of course theyve done great things at the likes of OpenAI to Mistral, but I hope that the combination of academic expertise and the companies focus could bring something new.And in many ways, its not about trying to outmaneuver Big Tech or billion-dollar AI startups; the ultimate goal is digital sovereignty: (mostly) open foundation LLMs built by, and for, Europe.I hope this wont be the case, but if, in the end, we are not the number one model, and we have a good model, then we will still have a model with all the components based in Europe, Haji said. This will be a positive result.

In BriefPosted:9:11 PM PST February 15, 2025Image Credits:Silas Stein/picture alliance / Getty ImagesDeath of OpenAI whistleblower deemed suicide in new autopsy reportSuchir Balaji, a former OpenAI employee, was found dead in his San Francisco apartment on Nov. 26; on Friday, the citys medical examiner ruled his death a suicide, countering suspicions by his family that had fueled widespread speculation online.Balaji made headlines in October when he accused OpenAI of illegally using copyrighted material to train its AI models. He shared his concerns publicly and provided information to The New York Times, which later named him as a key figure with unique and relevant documents in the newspapers lawsuit against OpenAI. His revelations came amid a growing number of publishers and artists to sue OpenAI over alleged copyright infringement.Just days before his death, Balaji had been in high spirits, according to his parents, celebrating his 26th birthday and planning a nonprofit in machine learning. His sudden passing drew attention from figures like Elon Musk and Tucker Carlson, while Congressman Ro Khanna called for a full and transparent investigation.Indeed, Balajis death of a self-inflicted gunshot, per the San Francisco County Medical Examiners report had become a focal point in debates over AI ethics, corporate accountability, and the dangers faced by whistleblowers in Silicon Valley. Whether these things become disentangled now remains to be seen.TopicsAI

In BriefPosted:12:56 PM PST February 15, 2025Image Credits:AppleApple Intelligence could arrive on Vision Pro in AprilApple is planning to add Apple Intelligence to its Vision Pro headset in an update that could come as early as April, according to Bloombergs Mark Gurman.Just a couple weeks after Apple Intelligence was first announced in June 2024, Gurman reported that Apple was looking to bring its suite of AI tools to the Vision Pro, though there were questions to answer about how those tools would be reimagined for a mixed reality experience.Now Apple is reportedly aiming to include Apple Intelligence (including Writing Tools, Genmoji, and Image Playground) in its visionOS 2.4 software update, with a version available to developers as soon as this week.The Vision Pros first Apple Intelligence offerings reportedly wont include an upgraded Siri. In fact, Gurman also said a long-promised upgrade to Siri more broadly could be delayed due to engineering problems and bugs.Topics

In BriefPosted:2:15 PM PST February 15, 2025Image Credits:Raleigh News & Observer / Getty ImagesNorth Carolina Amazon workers vote against unionizingWorkers at an Amazon warehouse in Garner, North Carolina voted against unionizing in election results announced today.According to Carolina Amazonians United for Solidarity and Empowerment (CAUSE), the worker group seeking to form the union, 3,276 ballots were cast in the election, with 25.3% of votes in favor of unionizing and 74.7% against. The results still need to be certified by the National Labor Relations Board (NLRB).In a statement provided to CNBC, CAUSE blamed the results on Amazons willingness to break the law, claiming, Amazons relentless and illegal efforts to intimidate us prove that this company is afraid of workers coming together to claim our power.Amazon spokesperson Eileen Hards denied the company had broken any laws and said, Were glad that our team in Garner was able to have their voices heard, and that they chose to keep a direct relationship with Amazon.Workers at an Amazon warehouse in Staten Island voted to unionize in 2022, and workers at a Philadelphia location of Amazon-owned Whole Foods also voted in favor of unionization earlier this year. The grocery chain has asked the NLRB to set those results aside.Meanwhile, Amazons lawyers recently joined SpaceX in a legal challenge to the NLRBs structure.Topics

Many venture industry observers have wondered whether Andreessen Horowitz, a firm that manages $45 billion, has its sights on eventually becoming a publicly traded company.Co-founder Marc Andreessen said he isnt chomping at the bit to take the firm public, on this weeks Invest Like the Best podcast. But he discussed his goal of building a16z into an enduring company, drawing inspiration from JP Morgan and publicly traded private equity firms.Historically, venture capital firms have been partnerships consisting of a small tribe of people sitting in a room together, trying to bounce ideas off of each other when they make investments, Andreessen said on the podcast.The problem with the partnership model, he said, is that its highly dependent on the ideas and expertise of those people at the table with no underlying asset value, as he described it. Once the original partners retire, the firm loses a lot of its value, even if a new generation of investors takes over.But even if they can keep it going, theres no underlying asset value. That next generation is just going to have to hand it off to the third generation, he said. Thats probably going to fail on the third generation. Its going to be on Wikipedia someday: that firm existed, and then it went away.The partnership model can be lucrative. A16zs billions under management generates sizable money management fees for the firm, in addition to profits made when its investments succeed.However, Andreessen said he constantly reminds internal staff and limited partners that the company isnt raising money just to harvest the fees. Its to give the company the cash to invest in growing companies.When we go for scale, its because we think its necessary to support the kinds of companies we want to help our founders build, he said.Andreessen says his bigger goal for a16z is to create a company that lasts. An alternative to a partnership is to build an investment company thats managed like a business, which means it has management, multiple layers of staff, division of labor with specializations, and training programs, Andreessen said.There are certainly precedents of small partnerships evolving into large corporations, which Andreessen can use as a model for a16zs ambitions.Goldman Sachs and JP Morgan, 100 years ago, looked like little venture capital firms, he said.Then their leaders, over time, turned them into huge franchises and big public companies.He named other examples, too, of private partnerships turned into large publicly traded companies like big private equity firms. Blackstone, which now has a market capitalization of over $200 billion, went public in 2007. Apollo, KKR, and Carlyle held their IPOs soon after Blackstone, and TPG listed on Nasdaq in early 2022.Andreessen argues that as these companies grew from partnerships into large corporations, their long-term success became less dependent on a few key investors.A big part of what weve been trying to do is build something that has that kind of enduring aspect to it, he said.In many ways, Andreessen Horowitz already looks more like an operating company than many VC firms. A16z has dozens of people in its marketing group and large teams that help portfolio companies recruit talent and sell their products. The firm runs separate crypto, bio and health, and American dynamism strategies.But maybe theres another reason Andreessen is keen to restructure away from the classic VC system. When it comes to partnerships, he says, It actually turns out in most cases, what you discover is that people actually dont like each other that much.Topics

Welcome back to Week in Review. This week were looking at OpenAI canceling the release of o3; TikTok returning to U.S. app stores nearly a month after it was removed; more complications in Elon Musks bid to buy OpenAI for $97.4 billion; and more! Lets do it.OpenAI effectively canceled the release of o3, which was slated to be the companys next major AI model, in favor of a simplified product offering. In a post on X, CEO Sam Altman said OpenAI plans to release GPT-5, a model that will integrate o3 in ChatGPT and its API. As a result, OpenAI will no longer launch o3 as a stand-alone model. The company plans to release GPT-4.5 before GPT-5 rolls out, according to Altman, who says this will be the companys last non-chain-of-thought model.Apple and Google restored TikTok to their respective app stores in the U.S., nearly a month after they removed the app due to a national security law that banned it. The companies have also restored video editor CapCut and social media app Lemon8, which were removed to comply with the law. The apps were restored after U.S. Attorney General Pam Bondi sent Apple a letter saying that the ban wouldnt immediately be enforced, Bloomberg reports.Apple CEO Tim Cook teased the newest member of the family, set to arrive Wednesday, which is likely a fourth-generation iPhone SE. Its been three years since Apple released the last iPhone SE at $429 an oversight for a product that plays such a big role for the company in markets like China and India. This time around, Apple is likely to ditch the Touch ID home button and use the same chip that powers the iPhone 16.This is TechCrunchs Week in Review, where we recap the weeks biggest news. Want this delivered as a newsletter to your inbox every Saturday? Sign up here.NewsImage Credits:Tatiana Lavrova (opens in a new window) / Getty ImagesMusk adds conditions to his bid for OpenAI: Lawyers for Elon Musk said the billionaire will withdraw his $97.4 billion bid for OpenAI if the board commits to keeping it as a nonprofit. The filing claims that Musks offer to buy OpenAI is serious. Read moreIf you cant take the heat: Metas CTO Andrew Bosworth has choice words for employees frustrated with the companys latest policies they are free to quit. A Business Insider report cites an internal chat where an employee criticized Meta for cutting DEI programs. Read moreTesla wins a defamation suit: Zhang Yazhou sued Tesla after a Tesla Model 3 allegedly crashed due to faulty brakes in 2021. After Tesla sued her back for defamation, a Chinese court has ordered Zhang to pay some $23,000 in damages and make a public apology. Read moreAn open Instagram: An app called Pinksky launched an Android version of its photo-centric social networking experience built on top of Bluesky. Its one of many attempting to court former Instagram users by offering a similar experience but built on open technology. Read moreWatch out, FIFA: Researchers at Google DeepMind published a demonstration of miniature humanoid robots playing a one-on-one game of soccer. The researchers found they were able to learn movement skills like walking, kicking, and being able to recover from a fall. Read moreGoogle expands NotebookLM Plus: Google has expanded NotebookLM Plus, a paid version of its AI-based note-taking and research assistant with higher usage limits and premium features, to users subscribed to the Google One AI Premium plan. Read moreZetas big valuation: The provider of banking software to banks and fintech startups raised $50 million from a strategic investor at a $2 billion valuation. The new investment marks a 70% increase in the Bengaluru-based startups valuation in 2021. Read moreGoogle Maps updates to the Gulf of America: U.S. Google Maps users will see the Gulf of America instead of the Gulf of Mexico following an executive order from President Trump. International users will see both names, while users in Mexico will only see the Gulf of Mexico. Read morePerplexitys Super Bowl growth hack: Instead of getting an expensive Super Bowl ad, the AI search engine used X to get users to download its app for a chance to win $1 million. The strategy increased its mobile app installs by roughly 50%, according to data from Appfigures. Read moreTumblr joins the fediverse: Tumblr is headed to the open social web thanks to a planned move to the WordPress infrastructure. When the migration is complete, Tumblr users will be able to federate their blog via ActivityPub, just as every WordPress.com user can. Read more

Perplexity has become the latest AI company to release an in-depth research tool, with a new feature announced Friday.Google unveiled a similar feature for its Gemini AI platform in December. Then OpenAI launched its own research agent earlier this month. All three companies even have given the feature the same name: Deep Research.The goal is to provide more in-depth answers with real citations for more professional use cases, compared to what youd get from a consumer chatbot. In a blog post announcing Deep Research, Perplexity wrote that the feature excels at a range of expert-level tasksfrom finance and marketing to product research.Perplexity Deep Research is currently available on the web, and the company said it will soon be added to its Mac, iOS, and Android apps. To use it, you just select Deep Research from a drop-down menu when you submit your query in Perplexity, which will then create a detailed report that can be exported as a PDF or shared as a Perplexity Page.To create this report, Perplexity said Deep Research iteratively searches, reads documents, and reasons about what to do next, refining its research plan as it learns more about the subject areas, supposedly similar to how a human might research a new topic.Image: PerplexityImage Credits:Perplexity /The company also highlighted its performance on Humanitys Last Exam, an AI benchmarking test with expert-level questions in a variety of academic fields. Perplexity said its Deep Research tool scored 21.1% on the test, easily beating most other models, such as Gemini Thinking (6.2%), Grok-2 (3.8%), and OpenAIs GPT-4o (3.3%) but not quite matching OpenAIs Deep Research (26.6%).But while you currently need a $200-per-month Pro subscription to use OpenAIs Deep Research (the company plans to expand to other subscription tiers), Perplexitys Deep Research is available for free non-subscribers get an unspecified-but-limited number of queries per day, while paying subscribers get unlimited queries.Perplexitys Deep Research also seems to perform more quickly, completing most tasks in under three minutes compared to 5 to 30 minutes for OpenAI Deep Research.Asked to compare the various deep research products, Perplexity offered an overview of the different technologies, pricing models, and performance in different use cases and subject matters, with links to articles about each feature). It summarized the differences as follows:Perplexity AI excels in speed and accessibility for casual researchersOpenAI dominates in analytical depth for enterprise applicationsGoogle integrates most seamlessly with existing productivity ecosystemsWhile its too early to know how these tools will affect everyday and professional research as they become more popular, The Economist recently highlighted shortcomings to OpenAIs Deep Research that likely apply here too: not just limitations to its creativity in interpreting data and a tendency to rely on sources that are easily available, but a larger risk that outsourcing all your research to a supergenius assistant could reduce the number of opportunities to have your best ideas.

In BriefPosted:10:53 AM PST February 15, 2025Image Credits:SOPA Images / Getty ImagesxAIs Colossus supercomputer raises health questions in MemphisElon Musks AI startup xAI plans to continue using 15 gas turbines to power its Colossus supercomputer in Memphis, Tennessee, according to an operating permit with the Shelby County Health Department for non-stop turbine use from June 2025 to June 2030.Why does it matter? The Commercial Appeal, a news outlet that obtained the documents, observes that environmental concerns have emerged, as the 20-year-old turbines emit hazardous air pollutants (HAP), including formaldehyde, at levels exceeding the EPAs 10-ton annual cap for a single source. (Per the story, the facilitys operating permit self-reports that the turbines each emit 11.51 tons of HAP per year.)The turbines have already been running since summer 2024 without public notice or oversight, says Eric Hilt, a spokesperson with Southern Environmental Law Center, the large environmental nonprofit organization, and the permits dont account for those emissions.Its another example of the company not being transparent with the community or with local leaders, Hilt tells The Commercial Appeal.The health department tells the outlet the permits have not yet been approved, and there is no set timeline for approval.Topics

Talk of backdoors in encrypted services is once again doing the rounds after reports emerged that the U.K. government is seeking to force Apple to open up iClouds end-to-end encrypted (E2EE) device backup offering. Officials were said to be leaning on Apple to create a backdoor in the service that would allow state actors to access data in the clear.The U.K. has had sweeping powers to limit technology firms use of strong encryption since passing a 2016 update to state surveillance powers. According to reporting by the Washington Post, U.K. officials have used the Investigatory Powers Act (IPA) to place the demand on Apple seeking blanket access to data that its iCloud Advanced Data Protection (ADP) service is designed to protect from third-party access, including Apple itself.The technical architecture of Apples ADP service has been designed in such a way that even the tech giant does not hold encryption keys thanks to the use of end-to-end encryption (E2EE) allowing Apple to promise it has zero knowledge of its users data.A backdoor is a term typically deployed to describe a secret vulnerability inserted into code to circumvent, or otherwise undermine, security measures in order to enable third parties. In the iCloud case, the order allows U.K. intelligence agents or law enforcement to gain access to users encrypted data.While the U.K. government routinely refuses to confirm or deny reports of notices issued under the IPA, security experts have warned that such a secret order could have global ramifications if the iPhone maker is forced to weaken security protections it offers to all users, including those located outside the United Kingdom.Once a vulnerability in software exists, there is a risk that it could be exploited by other types of agents, say hackers and other bad actors wanting to gain access for nefarious purposes such as identity theft, or to acquire and sell sensitive data, or even to deploy ransomware.This may explain why the predominant phrasing used around state-driven attempts to gain access to E2EE is this visual abstraction of a backdoor; asking for a vulnerability to be intentionally added to code makes the trade-offs plainer.To use an example: When it comes to physical doors in buildings, walls, or the like it is never guaranteed that only the propertys owner or key holder will have exclusive use of that point of entry.Once an opening exists, it creates a potential for access someone could obtain a copy of the key, for example, or even force their way in by breaking the door down.The bottom line: There is no perfectly selective doorway that exists to let only a particular person pass through. If someone can enter, it logically follows that someone else might be able to use the door too.The same access risk principle applies to vulnerabilities added to software (or, indeed, hardware).The concept of NOBUS (nobody but us) backdoors has been floated by security services in the past. This specific kind of backdoor typically rests on an assessment of their technical capabilities to exploit a particular vulnerability being superior to all others essentially an ostensibly more-secured backdoor that can only be exclusively accessed by their own agents.But by very nature, technology prowess and capability is a movable feat. Assessing the technical capabilities of unknown others is also hardly an exact science. The NOBUS concept sits on already questionable assumptions; any third-party access creates the risk of opening up fresh vectors for attack, such as social engineering techniques aimed at targeting the person with the authorized access.Unsurprisingly, many security experts dismiss NOBUS as a fundamentally flawed idea. Simply put, any access creates risk; therefore, pushing for backdoors is antithetical to strong security.Yet, regardless of these clear and present security concerns, governments continue pressing for backdoors. Which is why we keep having to talk about them.The term backdoor also implies that such requests can be clandestine, rather than public just as backdoors arent public-facing entry points. In Apples iCloud case, a request to compromise encryption made under the U.K.s IPA by way of a technical capability notice, or TCN cannot be legally disclosed by the recipient. The laws intention is that any such backdoors are secret by design. (Leaking details of a TCN to the press is one mechanism for circumventing an information block, but its important to note that Apple has yet to make any public comment on these reports.)According to the rights group the Electronic Frontier Foundation, the term backdoor dates back to the 1980s, when backdoor (and trapdoor) were used to refer to secret accounts and/or passwords created to allow someone unknown access into a system. But over the years, the word has been used to label a wide range of attempts to degrade, circumvent, or otherwise compromise the data security enabled by encryption.While backdoors are in the news again, thanks to the U.K. going after Apples encrypted iCloud backups, its important to be aware that data access demands date back decades.Back in the 1990s, for example, the U.S. National Security Agency (NSA) developed encrypted hardware for processing voice and data messages that had a backdoor baked into it with the goal of allowing the security services to intercept encrypted communications. The Clipper Chip, as it was known, used a system of key escrow meaning an encryption key was created and stored by government agencies in order to facilitate access to the encrypted data in the event that state authorities wanted in.The NSAs attempt to flog chips with baked-in backdoors failed over a lack of adoption following a security and privacy backlash. Though the Clipper Chip is credited with helping to fire up cryptologists efforts to develop and spread strong encryption software in a bid to secure data against prying government overreach.The Clipper Chip is also a good example of where an attempt to mandate system access was done publicly. Its worth noting that backdoors dont always have to be secret. (In the U.K.s iCloud case, state agents clearly wanted to gain access without Apple users knowing about it.)Add to that, governments frequently deploy emotive propaganda around demands to access data in a bid to drum up public support and/or put pressure on service providers to comply such as by arguing that access to E2EE is necessary to combat child abuse, or terrorism, or prevent some other heinous crime.Backdoors can have a way of coming back to bite their creators, though. For example, China-backed hackers were behind the compromise of federally mandated wiretap systems last fall apparently gaining access to data of users of U.S. telcos and ISPs thanks to a 30-year-old federal law that had mandated the backdoor access (albeit, in that case, of non-E2EE data), underscoring the risks of intentionally baking blanket access points into systems.Governments also have to worry about foreign backdoors creating risks for their own citizens and national security.There have been multiple instances of Chinese hardware and software being suspected of harboring backdoors over the years. Concerns over potential backdoor risks led some countries, including the U.K., to take steps to remove or limit the use of Chinese tech products, such as components used in critical telecoms infrastructure, in recent years. Fears of backdoors, too, can also be a powerful motivator.

Google Photos is a great service for storing images across devices. But Google Drive and Gmail only offer 15GB of storage for free. Google Photos used to offer free unlimited storage of images, but that is not the case anymore. If you are looking for a better photo storage plan, different features, or just want to move away from Googles ecosystem, here are some alternatives.FlickrFree storage: 1,000 photosTypically, storage services offer storage with limit on size. But Flickr is taking a different approach: It lets people store 1,000 images and videos for free. One advantage of Flickr: You can upload an image of up to 200MB in size, compared to the 75MB limit on Google Photos on the free plan. Flickrs paid plans start at $10.44 per month for unlimited storage.If you want to look at features beyond personal usage, Flickr allows you to make your photos public so others can find them. You can also join groups based on different topics.DropboxFree storage: 5GBDropbox is not a storage service centered around photos, but that could be an advantage if you are looking to store things beyond photos in the cloud. The company has paid plans starting at $9.99 per month for 2TB of storage, which is similar to Google Ones premium plan.EnteFree storage: 5GBEnte was created by a former Google engineer as a more private alternative to Google Photos. The service has end-to-end encryption protection for photos, which means the company doesnt collect any data. The app is available across platforms and includes features to identify and annotate people; show photos from different locations; and create categories like sunsets, memes, and documents. All of this is processed on your device. Entes monthly plans start from $2.49 per month for 50GB of storage, which can be shared with five other people. Entes core code is open sourced, so you can modify it to have a self-hosted version as well.Image Credits:EnteCrypteeFree storage: 100MBCryptee is another photo-storing service that focuses on privacy; its also open source. You can create an account with a username and a password (there is an option to use email and password as well). While its free tier offers not a lot of storage, the paid tier starts at $3.30 per month for 10GB of storage. The service works on iOS, Android, Windows, Mac, and Linux through a progressive web app and uses AES256 encryption to protect your media.Apart from being a photo storage service, it has a built-in document editor that supports markdown, code, and KaTeX math. Plus, you can use a side-by-side view for documents, store and edit files as PDF and docx, and use elements like tables and checkboxes.Amazon PhotosFree storage: 5GB with Prime membershipThis is a great perk for Amazon Prime members. You can cram some extra images you have into the free tier of Amazon Photos, and if you want more, storage plans start at $1.99 per month for 100GB.Image Credits: Amazon PhotosImage Credits:Amazon Photos500pxFree storage: 21 high-resolution photos per week uploads500px is geared more toward hobbyists or professional photographers. It has community features to highlight your work and a way to showcase your snaps in an uncompressed format. Plans cost less than $50 per year, with a discount that lets you store unlimited high-resolution photos. The premium plans remove ads from the platform and also offer insights into how your photos are performing on the platform. Its higher-tiered Pro plan, priced just under $100 per year, gives you tools to build a portfolio with a custom domain as well.Image Credits: 500pxImage Credits:500pxPhotobucketFree storage: No free storageOffering no free tier sounds like a bummer, but Photobucket offers one of the lowest storage rates, with $5 per month for 1TB of storage. If you pay for the plan on a yearly basis, the price is lower. Photobucket offers an ideal way to share photos with different groups with its group bucket plans at $8 per month for 1TB storage, which also gives you access to its editing tools.

In BriefPosted:2:51 PM PST February 14, 2025Image Credits:FigureFigure AI is in talks to raise $1.5B at 15x its last valuationSilicon Valley-based robotics startup Figure AI is in talks to raise a massive $1.5 billion round at a $39.5 billion valuation, Bloomberg reports.Thats a whopping 15 times higher than Figures $2.6 billion post-money valuation for its $675 million Series B last year. Figures current round is expected to be led by Align Ventures and Parkway Venture Capital, Bloomberg reported.Figure builds humanoid robots for commercial and residential purposes. Humanoid robots are all the rage thanks to the AI boom: Austin-based Apptronik just raised $350 million while Meta is reportedly looking to get into robotics, too.Figure has already sold some robots to BMW. Its founder Brett Adcock posted last month that Figure has signed on a second (unnamed) commercial customer and sees the potential to ship 100,000 humanoid robots.Figure declined to comment.Topics

Ride-share giant Uber filed a lawsuit Friday against DoorDash, accusing the delivery outfit of stifling competition by intimidating restaurant owners into exclusive deals.Uber alleges in the lawsuit, filed in Superior Court of California, that its chief rival bullied restaurants into only working with DoorDash. Uber claims that DoorDash, which holds the largest share of the food delivery market in the U.S., threatens restaurants with multimillion-dollar penalties or the removal or demotion of the businesses position on the DoorDash app.Specifically, Uber claims DoorDash pressures restaurants to strike exclusive or near-exclusive agreements for first-party delivery services, meaning that DoorDash insists on solely handling orders placed through restaurants own websites, says Uber.Ubers case has no merit, said a DoorDash spokesperson in an email to TechCrunch on Friday. Their claims are unfounded and based on their inability to offer merchants, consumers, or couriers a quality alternative.DoorDash and Uber Eats are best known for their respective apps to connect restaurant, consumers and gig economy workers. Consumers use the apps to find and order food like pizza, egg rolls, or pad thai from restaurants. A gig economy worker then picks up and delivers the food to the consumer.But the two companies also compete with their own white-label delivery services called Uber Direct and DoorDash Drive on-Demand which both launched in 2020. These services are cheaper for restaurants, allowing patrons to order directly from the restaurants own apps and websites, while Uber and DoorDash manage the couriers behind the scenes.Uber claims in its suit that DoorDash handles first-party deliveries for more than 90% of the largest enterprise restaurants in America, and it alleges DoorDash used anticompetitive practices to win the market.More than 1 million merchants partner with Uber Eats because weve helped them to reach more customers and provided them the freedom to decide how they want to grow their businesses with delivery, Sarfraz Maredia, head of the Americas for delivery at Uber, said in an emailed statement. Weve increasingly heard complaints from restaurants that DoorDashs tactics are limiting that freedom and punishing them for seeking better options. We hope this filing puts an end to those unfair practices so that restaurants can choose whats best for them without fear of penalty or retribution.In one example from the lawsuit, Uber says that an unnamed significant restaurant company told the company it would not move forward with a long-planned rollout of Uber Direct across several of its restaurant brands. The reason, Uber claims, is because DoorDash allegedly threatened to increase the rates it charges the restaurant company to use DoorDashs third-party delivery services if it continued to use Uber Direct.Uber says this was not a one-off event, but rather that multiple customer have told the company they feel like theyhave a gun to their head, that DoorDash is a monopolist, and that they are being bullied byDoorDash.Uber has requested a jury trial; the company did not specify the amount of damages in the complaint. However, Uber claims these anticompetitive practices have cost the company millions of dollars in revenue and also restricted the growth of Uber Direct.

In BriefPosted:3:24 PM PST February 14, 2025Image Credits:Bryce Durbin / TechCrunchAI Alexa and AI Siri face bugs and delaysAmazon and Apple are struggling to put generative AI technology in their digital assistants Alexa and Siri, respectively according to a pair of reports that came out on Friday.Amazon hoped to release its new Alexa during an event in New York on February 26. Now Amazon plans to delay the release of its generative AI-powered Alexa until March or later, according to the Washington Post.Meanwhile, Bloomberg reports that Apples AI overhaul of Siri is running into engineering problems and software bugs. Some new features around Siri that were planned for release in April may not be ready until May or later.Amazon and Apple hoped to release these updated products quickly to compete with next-gen AI voice assistants, such as OpenAIs Advanced Voice Mode and Googles Gemini Live. However, those efforts are not going according to plan.Topics

New court filings in an AI copyright case against Meta add credence to earlier reports that the company paused discussions with book publishers on licensing deals to supply some of its generative AI models with training data.The filings are related to the caseKadrey v. Meta Platforms one of many such cases winding through the U.S. court system thats pitted AI companies against authors and other intellectual property holders. For the most part, the defendants in these cases AI companies have claimed that training on copyrighted content is fair use. The plaintiffs copyright holders have vociferously disagreed.The new filings submitted to the court Friday, which include partial transcripts of Meta employee depositions taken by attorneys for plaintiffs in the case, suggest that certain Meta staff felt negotiating AI training data licenses for books might not be scalable. According to one transcript, Sy Choudhury, who leads Metas AI partnership initiatives, said that Metas outreach to various publishers was met with very slow uptake in engagement and interest.I dont recall the entire list, but I remember we had made a long list from initially scouring the Internet of top publishers, et cetera, Choudhury said, per the transcript, and we didnt get contact and feedback from from a lot of our cold call outreaches to try to establish contact.Choudhury added, There were a few, like, that did, you know, engage, but not many.According to the court transcripts, Meta paused certain AI-related book licensing efforts in early April 2023 after encountering timing and other logistical setbacks. Choudhury said some publishers, in particular fiction book publishers, turned out to not in fact have the rights to the content that Meta was considering licensing, per a transcript.Id like to point out that the in the fiction category, we quickly learned from the business development team that most of the publishers we were talking to, they themselves were representing that they did not have, actually, the rights to license the data to us, Choudhury said. And so it would take a long time to engage with all their authors.Choudhury noted during his deposition that Meta has on at least one other occasion paused licensing efforts related to AI development, according to a transcript.I am aware of licensing efforts such, for example, we tried to license 3D worlds from different game engine and game manufacturers for our AI research team, Choudhury said. And in the same way that Im describing here for fiction and textbook data, we got very little engagement to even have a conversation [] We decided to in that case, we decided to build our own solution.Counsel for the plaintiffs, who include bestselling authors Sarah Silverman and Ta-Nehisi Coates, have amended their complaint several times since the case was filed in the U.S. District Court for the Northern District of California, San Francisco Division in 2023. The latest amended complaint submitted by plaintiffs counsel alleges that Meta, among other offenses, cross-referenced certain pirated books with copyrighted books available for license to determine whether it made sense to pursue a licensing agreement with a publisher.The complaint also accuses Meta of using shadow libraries containing pirated e-books to train several of the companys AI models, including its popular Llama series of open models. According to the complaint, Meta may have secured some of the libraries via torrenting. Torrenting, a way of distributing files across the web, requires that torrenters simultaneously seed, or upload, the files theyre trying to obtain which the plaintiffs asserted is a form of copyright infringement.

OpenAIs board of directors has unanimously rejected billionaire Elon Musks offer to buy the nonprofit that effectively governs OpenAI, the company said on Friday. In a statement shared via OpenAIs press account on X, Bret Taylor, board chair, called Musks bid an attempt to disrupt his competition.OpenAI is not for sale, and the board has unanimously rejected Mr. Musks latest attempt to disrupt his competition, Taylor said. Any potential reorganization of OpenAI will strengthen our nonprofit and its mission to ensure [artificial general intelligence] benefits all of humanity.The New York Times reported that OpenAI also sent a letter to Musks lawyer, Marc Toberoff, saying that the bid was not in the best interests of [OpenAIs] mission.On Monday, Musk, his AI company, xAI, and a group ofinvestorsoffered to buy OpenAIs nonprofit for $97.4 billion. OpenAI CEO Sam Altman and the companys board of directorsquickly but not formally dismissed the unsolicited proposal. In astatement, Andy Nussbaum, the counsel representing OpenAIs board, said Musks bid doesnt set a value for [OpenAIs] nonprofit and that the nonprofit is not for sale.Musk, an OpenAI co-founder, last year brought a lawsuit against the company and Altman that alleges that OpenAI engaged in anticompetitive behavior and fraud, among other offenses.OpenAI was founded as a nonprofit before it transitioned to a capped-profit structure in 2019. The nonprofit is the sole controlling shareholder of the capped-profit OpenAI corporation, which retains formal fiduciary responsibility to the nonprofits charter.OpenAI is now in the process of restructuring this time to a traditional for-profit company, specifically a public benefit corporation. But Musk, via the lawsuit, is seeking to enjoin the conversion.In a court filing on Wednesday, lawyers for Musk said the billionaire will withdraw his bid if OpenAIs board preserve[s] the charitys mission and halts OpenAIs conversion to a for-profit. In afiling earlier the same day, attorneys for OpenAI called Musks move to take control of the company an improper bid to undermine a competitor, and a contradiction of his position in court that a transfer of the startups assets through restructuring would breach its mission as a charitable trust.Musks allies and Altman have traded blows over the bid this week. In a podcast interview on Thursday, Ari Emanuel, one of the backers of Musks offer for the OpenAI nonprofit, called Altman a phony who is trying to get away with cheating the charity and its original mission.Altman has characterized Musks bid as an attempt to slow [OpenAI] down, and quipped that Musks life is from a position of insecurity.

SailPoints IPO on Thursday was a disappointment for anyone hoping it would indicate that tech IPOs are hot again. The first days trading ended below the $23 initial price. The stock fared a tad better Friday, closing at over $24. But thats nothing close to the big bang companies and VCs hope for.For instance, ServiceTitan, the last tech IPO in December, was wildly successful. Share price popped from $71 to as high as $105 on Day 1, and is still currently trading at around $100.Back-to-back successes would have served as a signal that the painfully stuck-closed IPO window is opening at last.Instead, retail investors are exercising discernment, not wild enthusiasm.Im hesitant to draw too many conclusions on the appetite for tech or software IPOs from it, IPO expert Nick Einhorn, VP of research for Renaissance Capital, tells TechCrunch. While the company has good growth, it may not have stood out enough in the cybersecurity landscape to be awarded a premium sales multiple.Renaissance Capital is an IPO market research firm that also offers an IPO exchange-traded fund (ETF).SailPoint was a bit of an odd IPO because it wasnt a startup. It was previously a public company until PE firm Thoma Bravo took it private in 2022, valuing it at $6.9 billion at the time. The private equity giant is still the majority owner.This was a leveraged-buyout company as an IPO, not a classic venture backed startup. VC-backed startups going public often have the kind of growth potential that excites investors, as was the case with ServiceTitan.On the positive side for SailPoint, the company priced its initial 60 million shares at $23, above its previously announced range of $19 and $21. SailPoint raked in over $1.3 billion, which it will use for operations and to pay off about $1.5 billion of debt it showed on its books, according to a regulatory filing. Its also at about a $13 billion market cap, a boost from what Thoma Bravo paid.In no way did we consider this a disappointing IPO. We went from mid-point of $20to a close of $25 on Day 2. In our minds, its a very successful IPO, CEO Mark McClain told TechCrunch.Still, the upshot for those looking for a sign that IPOs could be flowing again soon (especially employees of late-stage startups looking at their paper-money stock and stock options): the signals remain murky.

With both Vice President J.D. Vance and U.S. Defense Secretary Pete Hegseth making loud noises Friday about Europe stepping up to the plate in spending more on its own defense, it might come as a surprise that Europe is already on the path toward far greater investment in defense, especially in tech.Not only has investment into Defense, Resilience, and Security (DSR) tech reached an all-time high of 10% of all VC funding in Europe, but the U.K. and NATO have now stepped up this week with the launch of the first NATO-approved cohort of startups under its Defense Innovation Accelerator for the North Atlantic (DIANA) program, in partnership with the U.K.DIANA brings in new technologies from startups for NATO countries and includes a network of 23 affiliated accelerator sites and 182 test centers. Its designed to connect defense personnel, startups, and defense prime contractors. The U.K.s Defense and Security Accelerator (DASA) is leading the U.K. front, in partnership with theIoT Tribe , which in turn heads the Janus Consortium on dual-use technologies. This also includes Atmos VC, a deep tech investor, and the SETsquared Partnership (a collaboration between six research-led UK universities and a range of partners from the defense industry). The Janus moniker comes from the Roman god of duality, reflecting the dual-use nature of both civilian and defense tech that will emerge from the program.The first cohort offive startupshas now been selected for a six-month program with DASA.The companies are spread across AI, electronics, semiconductors, and materials science, and were selected from 2,400 applications.Over a call with TechCrunch, Tanya Suarez, CEO of IOT Tribe, said 74 companies were selected and taken forward, with five going through the accelerator in London.With DASA, we run competitions for U.K. defense, and weve bought extra capacity within this system to put our own companies through as well, she said. The cohorts that are going through at the moment answer the call for surveillance and sensing. So were looking at technologies that could be used to essentially monitor either people or equipment.We have really close links to our customers across defense, plus we get a really great sense of the problems and the challenges that theyre trying to overcome, she added.ThroughDASAs partnership with NATO DIANA, were fast-tracking the development of dual-use technologies that will benefit both military and civilian applications across the Alliance.The startups selected are:AI Verse(France) An artificial intelligence (AI) company addressing the challenges surrounding accessibility, quality, privacy, accuracy, and labelling of images for training computer vision models.EIFys(Finland) Founded by university researchers, the company aims to commercialise induced junction black silicon photodiodes.Metahelios(U.K.) Company building imaging technologies for industries through a nanotechnological approach.RVmagnetics(Slovakia) A high-technology company in the field of sensors and identification technologies based on microwaves.Winse Power(Finland) A company developing Optal links, a solution for delivering power and data with light, with applications in the sensing and surveillance fields.

In BriefPosted:12:45 PM PST February 14, 2025Image Credits:VCG/VCG / Getty ImagesDeepSeek founder Liang Wenfeng is reportedly set to meet with Chinas Xi JinpingChinese AI startup DeepSeek founder Liang Wenfeng is reportedly set to meet with Chinas top politicians, including Chinese leader Xi Jinping, during a summit that Alibaba founder Jack Ma is also expected to attend.The summit, which could happen as soon as next week, may be intended as a signal by Chinas Communist Party that it aims to adopt a more supportive stance toward domestic private-sector firms, according to Bloomberg. In 2020, Chinese authorities effectively prevented Alibaba from executing what would have been the biggest public offering in history.Liang, who founded DeepSeek in 2023 as a subsidiary of his quantitative hedge fund, High-Flyer, rose to prominence last month after DeepSeeks openly available AI models showed strong performance against leading models from OpenAI and other American AI companies. U.S. officials have raised concerns over the explosive popularity of DeepSeeks models and services, which they perceive as a threat to the U.S. pole position in the AI race.Topics

Back in November, we broke the news that Meta owner of Facebook, Instagram and WhatsApp, with billions of users accounting for 10% of all fixed and 22% of all mobile traffic was close to announcing work on a major new, $10 billion+ subsea cable project to connect up the globe. The aim was to give Meta more control over how it runs its own services. Today, Meta confirmed details of our report: Project Waterworth is the official name, and it will be 50,000 kilometers long when completed, making it the worlds longest subsea cable project.Lining up with what we had heard about the project months ago, the network will connect up five continents with landing points in the United States, Brazil, India, South Africa, and other key regions. Facebook particularly calls out the opportunities in India, and the role that the network will play in how it rolls out AI services globally, as two key reasons for building the network.In terms of the network itself, Meta says it will be breaking new ground with its architecture, using 24 fiber pair cables, and what it describes as first-of-its-kind routing, maximizing the cable laid in deep water at depths up to 7,000 meters along with new burial techniques to reduce faults in areas deemed high risk, either because of geographical issues, or politics and sometimes both.The subsea cable spanning the U.S., South Africa, India and Asia. Image credit: MetaPreviously, we noted that geopolitics is one of the major factors in the market influencing Meta to build its own subsea infrastructure. That has indeed played a role here.On Thursday, the White House published a joint leaders statement from U.S. President Trump and Indias Prime Minister Shri Narendra Modi, which detailed a lengthy list of areas where the two countries would cooperate. Tucked into that long document was a commitment to co-developing undersea technologies as part of a defense partnership, as well as a note about Metas 50,000km Waterworth project, and Indias role in financing some of it.Supporting greater Indian Ocean connectivity, the leaders also welcomed Metas announcement of a multi-billion, multi-year investment in an undersea cable project that will begin work this year, the statement noted. India intends to invest in maintenance, repair and financing of undersea cables in the Indian Ocean, using trusted vendors.As for what the cable will be used for: In November, one of our sources postulated that the growth of AI data centers and cloud services in India, the worlds most populous country, was a significant reason for Metas project in the first place.Meta has declined to comment to us on more specific details but in its blog post provided some high-level ideas of applications and highlighted the country and AI.Digital communication, video experiences and online transactions, are among the applications that the subsea cable will enable, according to the post, penned by Metas VP of engineering Gaya Nagarajan and its global head of network investments Alex-Handrah Aim. Project Waterworth will be a multi-billion dollar, multi-year investment to strengthen the scale and reliability of the worlds digital highways by opening three new oceanic corridors with the abundant, high speed connectivity needed to drive AI innovation around the world.This is not Metas first subsea cable effort, nor is it the only major technology company to build its own subsea infrastructure. According to telecom analysts Telegeography, Meta is part owner of 16 existing networks, including the 2Africa cable that encircles the continent (others in that project are carriers, including Orange, Vodafone, China Mobile, Bayobab/MTN, and more).This new cable project would be the first wholly owned by Meta itself.That would put Meta into the same category as Google, which has involvement in some 33 different routes, including a few regional efforts in which it is the sole owner, per Telegeographys tracking. Other Big Tech companies that are either part owners or capacity buyers in subsea cables include Amazon and Microsoft, neither of which are whole owners of any route themselves.

Welcome to Startups Weekly your weekly recap of everything you cant miss from the world of startups. Want it in your inbox every Friday? Sign up here.This week confirmed that even when current events cloud the outlook, some startups still manage to raise significant amounts of funding, with those tied to security and sovereignty finding tailwinds.Most interesting startup stories from the weekImage Credits:Nathan Laine / Bloomberg / Getty ImagesUncertainty usually doesnt rhyme with investment. But theres always money for security, whether online or IRL.Countercyclical: AI startups raised $110 billion in 2024, a 62% increase compared to 2023, while overall startup funding was down 12% year on year.Meanwhile, European startups in the defense, security, and resilience tech space raised $5.2 billion last year, a record 10% of all VC funding in the region.European AI: French champion Mistral AI was in the spotlight this week during the Paris AI Summit, with its AI assistant Le Chat becoming the most downloaded iOS app in France.Security M&As: Security compliance firm Data acquired software security review startup SafeBase for $250 million. And CyberArk, an identity security company, bought startup Zilla Security for $165 million in cash, with an additional $10 million contingent on certain milestones.Most interesting VC and funding news this weekImage Credits:ZetaStartups in a wide range of sectors announced funding rounds this week. Plus, Founders Fund should soon have more capital to invest.Money in the bank: Zeta, a Bengaluru-based provider of banking software, is now valued at $2 billion after raising $50 million from a strategic investor.Quantum notes: Google-backed Boston quantum startup QuEra secured $230 million in debt via a convertible note, at an undisclosed valuation that its CEO said represented a very substantial increase compared to its previous round.Forestation: Carbon removal startup Chestnut Carbon closed a $160 million Series B round of financing to turn old farms into forests.Firstname.ai: Australian health tech startup Harrison.ai raised a $112 million Series C round for its AI-enabled medical diagnostic software, Annalise.ai, which is focused on radiology, and Franklin.ai, which is focused on pathology.Easy bills: Only six months after raising $29 million, American startup Candid Health closed a $52.5 million Series C to simplify medical billing.Manifest: Latent Labs, an AI biotech startup founded by a former research scientist at DeepMind, came out of stealth mode with $50 million in funding.Crystal ball: Israeli startup Voyantis scored $41 million to apply AI to helping companies figure out the lifetime value of their customers.Three times lucky? Peter Thiels Founders Fund is on track to close another $3 billion fund, according to sources. This will be the firms third growth fund.Last but not leastImage Credits:Sheel Mohnot, Better Tomorrow Ventures / Rex SalisburyIf you are reading this, you may be one of Sheel Mohnots 150,000 X followers. But even if youre not, Mary Ann Azevedos interview with the co-founder and general partner of Better Tomorrow Ventures is well worth listening to.In their chat, Mohnot talked about his Twitter fame, and about AI and fintech, which he is as bullish about as ever. Im excited about a lot of the stuff that we always believed in, which is the everything is fintech story, which persists today, he said.

AI innovators, this is your moment! Have insights to share with 1,200 AI founders, investors, and enthusiasts eager to push the boundaries of innovation?Take the stage, shape the AI conversation, and exchange ideas at TechCrunch Sessions: AI on June 5 at UC Berkeleys Zellerbach Hall!Were bringing together top AI minds from the startup world to lead insightful sessions and interactive breakout sessions. Help founders, entrepreneurs, and innovators navigate the challenges and opportunities in AI.This is your chance to dive deep into critical AI topics. Assemble a team of up to four presenters (including a moderator) to lead a 50-minute session complete with a presentation, panel discussion, and audience Q&A designed to spark impactful discussions.Simply hit the Apply to Speak button and submit your topic on this event page. From startups and investments to infrastructure and emerging AI tools, TC Sessions: AI is the premier platform to showcase your wisdom.After you submit your application, your topic will be voted on by our audience to decide which sessions theyll want to see live at the event.Breakout session perksMore than just branding get the full TC Sessions: AI experience! While gaining brand visibility at the event, youll also experience all the benefits of an attendee access to top-tier AI main-stage discussions, breakouts, and valuable 1:1 or small-group networking.Plus, TechCrunch will amplify your brand with:Event agenda listings (web & app)A mention in a shared TechCrunch.com articleSocial media promotionMake an impact in the AI ecosystemInspire, educate, and lead! Play a vital role in advancing the AI ecosystem while cementing your status as a trusted industry expert.Apply to speak before the deadline! TC Sessions: AI is set for June 5, but content applications close on March 7. If you want to present, get your application in today!TC Sessions: AI

Founders, VCs, and startup experts this is your moment!Got scaling insights? Nows your chance to share them with 1,200 founders, investors, and entrepreneurs at TechCrunch All Stage 2025 presented by Fidelity, on July 17 in Boston!Were gathering leading experts from the startup and VC communities to host engaging sessions, interactive roundtables, and dynamic breakouts. Come share your knowledge and help founders navigate the highs and lows of scaling their businesses.Submit to speak at TC All StageThis is your opportunity to explore key scaling topics for startups at every stage. Click the Apply to Speak button on the event page, and submit your topic. All Stage is the ideal platform to share your expertise in scaling successful startups.Then, you will choose the format and topic of your choice:Breakout Session: Assemble a team of up to four presenters (with a moderator) to guide a 35-minute session. With a balanced mix of presentation, panel discussion, and audience Q&A, your session will be designed to ignite insightful, impactful discussions.Roundtable Session: Up to two speakers will host two 30-minute encore sessions, providing a more intimate setting for in-depth exploration of the topic. These informal discussions are free-form, with no video or slides for a truly interactive experience.Speaking session perksIts more than just exposure its the complete TC All Stage experience! While boosting your brands visibility, youll also enjoy premium access to founder and VC talks in breakouts, roundtables, and priceless networking opportunities with top industry leaders.Plus, TechCrunch will amplify your brand with:Event agenda listings (web & app)A mention in a shared TechCrunch.com articleSocial media promotionLead, educate, and make an impact! Play a crucial role in guiding startup leaders through the scaling process and strengthen your reputation as an industry expert. Apply before the deadline! TC All Stage is happening on July 17, but content applications close on April 18. Get your application in now and be part of this exciting event!