Straw Stack Space in Houhu Art Park / WCY Regional StudioSave this picture!© Jinquan KongCommunity Center, Cultural Center•Chang Sha Shi, China Architects: WCY Regional Studio Area Area of this architecture project Area:  748 m² Year Completion year of this architecture project Year:  2023 Photographs Photographs:Jinquan Kong, Haohao Xu Lead Architects: Chunyu Wei, Sheng Ouyang More SpecsLess Specs Save this picture! Text description provided by the architects. Straw Stack Space is located in Houhu Art Park in Changsha, Hunan Province, which is a special 'village in the middle of the city' in Changsha that grew up spontaneously, located at the foot of Yuelu Mountain, and in the centre of the university town. Since the 1950s, the Houhu Art Park has become a unique cultural landmark in Changsha through its history of fisheries, art training and art exhibitions. And, Straw Stack Space is a small community cultural complex in Houhu Lake, integrates exhibitions, art salons, music performances and cultural and creative exchanges.Save this picture!Using pure 'Euclidean Geometry' ---square and circle as the basic graphic language, Straw Stack Space tries to revive the poetic atmosphere of Houhu's past as a field and countryside through the abstraction and metaphor of 'wheat (rice) haystacks', round barns, sunbathing yards and other memorable scenes in the rural landscape. The building occupies two residential sites, and we combined one house with six 'cylinders' in juxtaposition, while the other house has one 'cylinder' implanted in a pure 'square'. The cylinder array and the standard square form an isomorphic and inverted base relationship.Save this picture!Save this picture!Save this picture!The public cultural and artistic activities that take place in the Straw Stack Space with a sense of 'autonomy' or 'accident', such as contemporary art exhibitions, university graduation design exhibitions, small-scale music parties, modern dances, lawn music festivals, etc. At this point, the 'space' has the ability of 'functional transformation',and it has already departed from the original design preconception, and has generated other interpretations, but it is also an unexpected praise to us.Save this picture!The windows of the Straw Stack Space are mainly orientated to spatial experience, especially strengthening the use of natural top light: when the light is thrown down from the upper part of the building's cylindrical space, through the vertical directivity of the cylinders, it draws one's visual focus to the sky, establishing a relationship between the space, the body and the sky and the earth. Together with the geometrical order of the arrayed cylinders, it tries to let the body deeply perceive a certain ritual, quiet and poetic spatial atmosphere.Save this picture!Save this picture!Save this picture!After the completion of the project, whether it is the indigenous villagers, neighbouring citizens, teachers and students of the university, or foreign tourists, when they are relaxing in the Straw Stack Space, they will feel a special kind of 'spatial' information, which tries to evoke the memory of a familiar place or unfamiliar reverie, which belongs to the unique 'sense of place' of the Houhu Art Park or the Straw Stack Space, and it is also the power of the Archetype that we are looking for.Save this picture! Project gallerySee allShow less Project locationAddress:Chang Sha Shi, ChinaLocation to be used only as a reference. It could indicate city/country but not exact address.About this officeWCY Regional StudioOffice••• MaterialBrickMaterials and TagsPublished on May 15, 2025Cite: "Straw Stack Space in Houhu Art Park / WCY Regional Studio" 15 May 2025. ArchDaily. Accessed . <https://www.archdaily.com/1030097/straw-stack-space-in-houhu-art-park-wcy-regional-studio&gt ISSN 0719-8884Save世界上最受欢迎的建筑网站现已推出你的母语版本!想浏览ArchDaily中国吗?是否 You've started following your first account!Did you know?You'll now receive updates based on what you follow! Personalize your stream and start following your favorite authors, offices and users.Go to my stream

Tencent's president Martin Lau said the company can rely on a "pretty strong stockpile of chips that we acquired previously." Ying Tang/NurPhoto via Getty Images 2025-05-15T04:30:03Z Save Saved Read in app This story is available exclusively to Business Insider subscribers. Become an Insider and start reading now. Have an account? Tencent's president Martin Lau says company has a "pretty strong stockpile of chips." Last month, Nvidia said the Trump administration imposed new chip export restrictions to China. Lau said Tencent has enough chips to train its AI models "for a few more generations going forward." The Chinese tech giant Tencent said it has a "pretty strong stockpile of chips" to tide it through America's chip sale restrictions.The company's president, Martin Lau, was speaking to investors during an earnings call on Monday when he was asked how Tencent would deal with US chip restrictions.Lau said "it's a very dynamic situation" that Tencent is managing, and it's trying to "figure out the right solution" to make sure its AI strategy "can still be executed." Lau told investors that Tencent can rely on a "pretty strong stockpile of chips that we acquired previously." Lau said the chips will be deployed to projects that can "generate immediate returns for us," such as in Tencent's advertising business.When it comes to training Tencent's large language models, Lau said the company will not need a large number of chips to improve its performance. This is because companies are starting to move away from the scaling law "which required continuous expansion of the training cluster," Lau said."And now we can see even with a smaller cluster, you can actually achieve very good training results. And there's a lot of potential that we can get on the post-training side which do not necessarily meet very large clusters," Lau said."So that actually helps us to look at our existing inventory of high-end chips and say we should have enough high-end chips to continue our training of models for a few more generations going forward," Lau added.Last month, Nvidia told investors in a regulatory filing that the Trump administration will be imposing new export licensing restrictions for chips being sold to China and other countries.The chip giant said it would incur a charge of up to $5.5 billion in inventory, purchase commitments, and reserves for its H20 chip in the first quarter, which ended on April 27. Nvidia designed the H20 chip based on the Biden administration's chip export restrictions.Analysts, however, have said the Trump administration's new restrictions will not slow down China's progress in AI."Banning the H20 would make no sense as its performance is already well below Chinese alternatives; a ban would simply hand the Chinese AI market completely over to Huawei," Bernstein analysts wrote in a note to investors on April 23.Tencent did not respond to a request for comment from Business Insider. Recommended video

options aplenty OpenAI adds GPT-4.1 to ChatGPT amid complaints over confusing model lineup Two new AI models join 7 others, leaving some paid users wondering which one is best. Benj Edwards – May 14, 2025 6:16 pm | 18 Credit: Getty Images Credit: Getty Images Story text Size Small Standard Large Width * Standard Wide Links Standard Orange * Subscribers only   Learn more On Wednesday, OpenAI announced that ChatGPT users now have access to GPT-4.1, an AI language model previously available only through the company's API since its launch one month ago. The update brings what OpenAI describes as improved coding and web development capabilities to paid ChatGPT subscribers, with wider enterprise rollout planned in the coming weeks. GPT-4.1 and 4.1 mini to ChatGPT adds to an already complex model selection that includes GPT-4o, various specialized GPT-4o versions, o1-pro, o3-mini, and o3-mini-high models. There are technically nine AI models available for ChatGPT Pro subscribers. Wharton professor Ethan Mollick recently publicly lampooned the awkward situation on social media. As of May 14, 2025, ChatGPT Pro users have access to eight main AI models, plus Deep Research. Credit: Benj Edwards Deciding which AI model to use can be daunting for AI novices. Reddit users and OpenAI forum members alike commonly voice confusion about the available options. "I do not understand the reason behind having multiple models available for use," wrote one Reddit user in March. "Why would anyone use anything but the best one?" Another Redditor said they were "a bit lost" with the many ChatGPT models available after switching back from using Anthropic Claude. Reportedly better at coding So, what is actually different about GPT-4.1? Notably, it features a very large 1 million token context window that allows processing roughly 3,000 pages of text in a single conversation. The API launch included three versions: GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano. So far, only the full and mini versions are available in ChatGPT. The full GPT-4.1 model reportedly prioritizes instruction following and coding tasks, which the company positions as an alternative to its o3 and o4-mini simulated reasoning models for basic programming needs. For the smaller of the two models in ChatGPT, the company claims that GPT-4.1 mini performs better in instruction following, coding, and "overall intelligence" compared to GPT-4o mini. OpenAI is replacing GPT-4o mini with GPT-4.1 mini across all ChatGPT tiers, including free accounts. Free users will automatically switch to GPT-4.1 mini after reaching usage limits for GPT-4o. ChatGPT subscribers using Plus, Pro, or Team plans can access GPT-4.1 through a "more models" dropdown menu in the platform's model picker. The release comes just two weeks after OpenAI made GPT-4 unavailable in ChatGPT on April 30. That earlier model, which launched in March 2023, once sparked widespread hype about AI capabilities. Compared to that hyperbolic launch, GPT-4.1's rollout has been a fairly understated affair—probably because it's tricky to convey the subtle differences between all of the available OpenAI models. As if 4.1's launch wasn't confusing enough, the release also roughly coincides with OpenAI's July 2025 deadline for retiring the GPT-4.5 Preview from the API, a model one AI expert called a "lemon." Developers must migrate to other options, OpenAI says, although GPT-4.5 will remain available in ChatGPT for now. A confusing addition to OpenAI’s model lineup In February, OpenAI CEO Sam Altman acknowledged his company's confusing AI model naming practices on X, writing, "We realize how complicated our model and product offerings have gotten." He promised that a forthcoming "GPT-5" model would consolidate the o-series and GPT-series models into a unified branding structure. But the addition of GPT-4.1 to ChatGPT appears to contradict that simplification goal. So, if you use ChatGPT, which model should you use? If you're a developer using the models through the API, the consideration is more of a trade-off between capability, speed, and cost. But in ChatGPT, your choice might be limited more by personal taste in behavioral style and what you'd like to accomplish. Some of the "more capable" models have lower usage limits as well because they cost more for OpenAI to run. For now, OpenAI is keeping GPT-4o as the default ChatGPT model, likely due to its general versatility, balance between speed and capability, and personable style (conditioned using reinforcement learning and a specialized system prompt). The simulated reasoning models like 03 and 04-mini-high are slower to execute but can consider analytical-style problems more systematically and perform comprehensive web research that sometimes feels genuinely useful when it surfaces relevant (non-confabulated) web links. Compared to those, OpenAI is largely positioning GPT-4.1 as a speedier AI model for coding assistance. Just remember that all of the AI models are prone to confabulations, meaning that they tend to make up authoritative-sounding information when they encounter gaps in their trained "knowledge." So you'll need to double-check all of the outputs with other sources of information if you're hoping to use these AI models to assist with an important task. Benj Edwards Senior AI Reporter Benj Edwards Senior AI Reporter Benj Edwards is Ars Technica's Senior AI Reporter and founder of the site's dedicated AI beat in 2022. He's also a tech historian with almost two decades of experience. In his free time, he writes and records music, collects vintage computers, and enjoys nature. He lives in Raleigh, NC. 18 Comments

ZDNET's key takeaways The Ultrahuman Ring Air is the brand's first foray into the smart ring space, and it's available for $349 (though it's currently $50 off at QVC), no subscription required The ring is great for hardcore fitness enthusiasts and recreational exercisers looking to use their health data to optimize their wellness routines The app's user interface could be improved for easier access to daily logging functions. View now at QVC The Ultrahuman Ring Air is a competitive smart ring that tracks your sleep and fitness. Right now, you can take an additional $50 off the smart ring when you buy it at QVC.Also: Two popular smart rings just got caught copying OuraAs one of the hottest smart rings on the market, the Ultrahuman Ring Air offers features and data collection that goes a step beyond Oura for individuals looking to optimize their health. I tested one alongside the Oura Horizon ring for a month, wearing both around the clock and logging everything from my morning coffee to my late-night cocktails. Suffice it to say, the Ultrahuman ring will excite people dialed into their health and fitness.Also: Oura Ring 3 vs Oura Ring 4: Should you buy the discounted smart ring still on sale or the brand's newest?While Oura seemingly appeals to just about anyone looking to invest in a smart ring, Ultrahuman's offering, while more niche, is arguably better. If the Oura Ring is the girl next door, the Ultrahuman Ring Air is the Silicon Valley-based, Bitcoin-mining neighbor who drinks Soylent. But while it will set hardcore fitness enthusiasts on the right path, this ring will also help anyone, even recreational exercisers, optimize their body to a tee. details View at QVC Unlike the Oura Ring, which requires a $6 monthly subscription, the Ultrahuman Ring Air is subscription-free. The Ultrahuman comes in matte gray, matte black, silver, and gold. Out of all the smart rings I have tested, it is by far the thinnest, most discreet, and unassuming. Plus, the matte gray finish hasn't scratched or tarnished after a month of use, something I can't say about the other smart rings with shiny metal finishes.From the moment you wake up to the minute you fall asleep, Ultrahuman is sending you notifications on how to make the most of your available energy and body clock. With all these scientific insights into my body, I couldn't help but feel like I was hacking my health -- or playing God. Within an hour of waking, Ultrahuman pings me about my "residual adenosine levels." My what? I had to look it up as well. Adenosine is a naturally occurring chemical associated with wakefulness. It tells me to delay consuming stimulants like caffeine until around two hours after I've been awake to flush adenosine out of my system and boost the impact of caffeine on my energy levels. It also reminds me to expose myself to bright light or exercise during this window if I want an earlier bedtime and wake-up time.Throughout the day, I get pings like this. Ultrahuman tells me when my stimulant permissible window, the window where I can consume caffeine without it disrupting my sleep, is coming to a close. Not only that: Ultrahuman also tells me how much caffeine is in certain stimulants, how long they will stay in my system, and whether they exceed the caffeine intake of my window.As the day goes on, Ultrahuman pings me to stretch my legs and when the ideal bedtime for a sufficient night of sleep would be. All this is calculated based on my wake-up times, sleep, and recovery data. Also: The best smart rings of 2025: Expert tested and reviewedThe Ultrahuman Ring Air measures heart rate, skin temperature, heart rate variability (HRV), and resting heart rate to glean data for sleep, recovery, and movement. It also measures VO2 Max, or how the body utilizes oxygen during workouts. Ultrahuman presents all this data in an easy-to-understand format through an app, with scores (and explanations) for the big three: sleep, movement, and recovery. For example, when I got a 95 sleep score, it told me I fell asleep faster and stayed asleep for most of my rest duration. In tandem with my high sleep score, I got a 90 recovery score, and the blurb told me that my scores were in great shape, so I "might want to push a little more towards your cognitive and physical fitness goals." On a day when my sleep score was 49 and my recovery score 64, the app instructed me to go on a long walk and try a "non-sleep deep rest session" during the day. Nina Raemont/ZDNETOther health data, like HRV, VO2 Max, and resting heart rate, are tracked and the day-by-day data is aggregated and shown through graphs with daily, weekly, and monthly trends -- depicted with green and red indicators.  All these data points come with explanations of what each indicator means about one's overall health, elucidating the complexity of them.Another health feature the Ultrahuman ring offers is food logging, which leverages ChatGPT-powered AI food insights. I log the food I eat throughout the day, and its Food Optimization AI provides what it calls cyborg insights -- no, really -- about how to avoid glucose spikes when I consume certain foods. When I logged a bag of potato chips, it told me to pair the chips with cheese or other protein, like a hard-boiled egg, to slow down the absorption of glucose. It also told me to drink water before and after eating the chips and take a brisk walk after eating them to lower my glycemic response. While I rarely took this advice, this is a helpful feature for fitness and people who want to optimize their diet and health data.Also: The Oura smart ring's brilliant new features outshine even its titanium finishOf course, you also can log workouts, something I did often while training for my half marathon. The ring doesn't automatically detect that you're exercising, however, if it goes unlogged. I hope Ultrahuman improves this in future updates. I liked that I could see the map of where I ran, my average heart rate, my max heart rate, the calories I burned, and my average pace. Plus, it showed my workout zones, too, something essential for tracking long runs.The last feature that makes this ring ideal for exercise junkies is the Discover tab, which offers video classes for everything from Pilates and weight training to HIIT and yoga. There are also podcasts available for meditation and soundscapes for falling asleep.  Nina Raemont/ZDNETYou can wear the ring for about five days before the battery dies. I tested its battery life compared to Oura's by charging both to 100% on a Tuesday afternoon and waiting to see how long each would last through regular wear and activity. By Sunday morning, the Oura ring was at 7%, and the Ultrahuman Ring Air was at 6%. I will note, however, that the Ultrahuman ring gets hot to the touch after charging.In the next update of the Ultrahuman Ring Air, I'd like to see the app's user interface improve. It could take some notes from Oura's app, which offers sleep, readiness, resilience, and activity on top of a home tab that aggregates this data on the bottom. Ultrahuman's bottom tabs include home, metabolism, zones, Discover, and a bare-bones profile section. I'd also like to see a more accessible logging feature wherein you don't have to scroll all the way down the app to use for food, exercise tracking, weight, and more.ZDNET's buying adviceWho should buy this ring? I can imagine that anybody who's dialed into fitness and health metrics will get full use of the Ultrahuman Ring Air. And given how often my sleep duration and sleep and recovery scores from the Ultrahuman mimicked that of Oura's, I wouldn't be afraid to call this a subscription-free Oura dupe with rivaling data and battery life.   Ultimately, the Ultrahuman Ring Air impressed me. It's exciting to see such a new product step up to be a competitive fitness smart ring, being just niche enough to carve out an audience of devout fitness freaks but with data presentation, health metrics, and AI suggestions that could cater to the average Joe just as well.When will this deal expire? This deal will expire on May 19. The deal runs for four days. We're sorry if you've missed out on this deal, but don't fret -- we're constantly finding new chances to save and sharing them with you at ZDNET.comFeatured reviews

May 14, 20253 min readMigraine Drug Ubrogepant Tackles Debilitating Early SymptomsResults from a phase 3 clinical trial suggest that taking ubrogepant at the first sign of an oncoming migraine can prevent preheadache fatigue and light sensitivityBy Fred Schwaller & Nature magazine Migraine symptoms that appear long before a headache can include fatigue, light sensitivity, neck pain and difficulty concentrating. nemke/Getty ImagesScientists have shown that a drug approved to treat migraine headaches can also alleviate debilitating non-headache symptoms, such as fatigue, brain fog and blinding light sensitivity, that occur as the migraine is starting.The drug — called ubrogepant — is already known to stop the onset of a full-blown migraine attack in some people if they take it when the headache begins. But a phase III clinical trial, described in Nature Medicine on 12 May, shows that it can also tackle the ‘prodrome’ symptoms that arrive hours or even days earlier.The results suggest that ubrogepant could “free patients from a disabling part of migraine,” says study co-author Peter Goadsby, a neuroscientist at King’s College London.On supporting science journalismIf you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.Early interventionThe process of a migraine starts long before the head pain, when brain circuits involving the hypothalamus — a region that regulates several vital bodily functions — become dysregulated. In the prodrome, or premonitory phase, people can experience various unpleasant symptoms, including fatigue, neck pain, an aversion to light (photophobia) or sound (phonophobia) and difficulty concentrating.“Not enough attention has been given to prodrome symptoms,” says Goadsby. The trial aimed to “fill this gap” by investigating whether ubrogepant has an effect on the initial stages of a migraine.The trial included 438 participants who could reliably identify incoming migraine attacks from their prodromal symptoms. During the 60 days of the trial, they took ubrogepant or a placebo whenever they felt any prodromal symptoms coming on, and reported whether doing this had any effect.The results suggest that for some participants, the drug increased their ability to concentrate one hour after treatment, reduced their photophobia two hours after treatment and reduced their fatigue and neck pain after three hours.But the “effect sizes were small” and never more than 15 percentage points compared with the placebo, points out Gregory Dussor, a neuroscientist and migraine specialist at the University of Texas at Dallas. For example, 27% of participants who took ubrogepant reported an absence of fatigue, compared with 17% who took the placebo. Dussor suggests this modest improvement is because the class of drugs to which ubrogepant belongs is a “life-changing therapeutic” for a minority of people — perhaps only one in five, according to some studies. It makes little difference to others, so its effect for that minority “gets diluted in the data”.“Nevertheless, this study answers the question of whether ubrogepant works when you dose it early in the migraine, when people feel prodromal symptoms coming on,” he says.Know your migraineUbrogepant is already known to stop migraine headaches, but Goadsby says the data show that it works substantially better if taken in the prodromal stage rather than once the attack starts. In the trial, it was crucial that participants could predict their preliminary symptoms. “People who know their migraine would benefit more from this medicine,” he says. The next challenge for the field, he adds, is to train people to understand their migraine symptoms so they can judge when to take ubrogepant.Dussor thinks that once the prodrome has begun, it is probably still too late for ubrogepant to block the migraine entirely. He says research should explore another question: “Is there some way to alert people much earlier to take the drug, before they feel symptoms coming on?”The study could also provide clues about the cause of prodromal symptoms. Ubrogepant works by blocking receptors that bind to calcitonin gene-related peptide (CGRP), a molecule that sensitizes the nerves in the head and skull during a migraine attack. “Knowing there’s some CGRP component earlier than the headache phase is important to know in our mechanistic understanding of migraine,” says Dussor.This article is reproduced with permission and was first published on May 12, 2025.

Author(s): Felix Pappe Originally published on Towards AI. Image created by the author using gpt-image-1 Introduction In the past, extracting specific information from documents or images using traditional methods could have become quickly cumbersome and frustrating, especially when the final results stray far from what you intended. The reasons for this can be diverse, ranging from overly complex document layouts to improperly formatted files, or an avalanche of visual elements, which machines struggle to interpret. However, vision-enabled language (vLMs)models have come to the rescue. Over the past months and years, these models have gained ever-greater capabilities, from rough image descriptions to detailed text extraction. Notably, the extraction of complex textual information from images has seen astonishing progress. This allows for rapid knowledge extraction from diverse document types without brittle, rule-based systems that break as soon as the document structure changes — and without the time-, data-, and cost-intensive specialized training of custom models. However, there is one flaw: vLMs, like their text-only counterparts, tend to produce verbose output around the information you actually want. Phrases such as “Of course, here is the information you requested” or “This is the extracted information about XYZ” commonly surround the essential content. You could use regular expressions together with advanced prompt engineering to constrain the vLM to output only the requested information. However, crafting the perfect prompt and matching regex for a given task is difficult and requires much trial and error. In this blog post, I’d like to introduce a simpler approach: combining the rich capabilities of vLLMs with the strict validation offered by Pydantic classes to extract exactly the desired information for your document processing pipeline. Description of post tackled issue The example in this blog post describes a situation that every job applicant has likely experienced many times. I am sure of it. After you have carefully and thoroughly created your CV, thinking about every word and maybe even every letter, you upload the file to a job portal. But after successfully uploading the file, including all the requested information, you are asked once again to fill out the same details in standard HTML forms by copying and pasting the information from your CV into the correct fields.Some companies attempt to autofill these fields based on the information extracted from your CV, but the results are often far from accurate or complete.In the following code, I combine Pixtral, LangChain, and Pydantic to provide a simple solution. The code extracts the first name, last name, phone number, email, and birthday from the CV if they exist. This helps keep the example simple and focuses on the technical aspects.The code can be easily adapted for other use cases or extended to extract all required information from a CV.So let us dive into the code. Code walkthrough Importing required libraries In the first step, the required libraries are imported, including: os, pathlib, and typing for standard Python modules providing filesystem access and type annotations base64 dontenv.env file into os.environ pydanticLLM output ChatMistralAILLM interface PIL import osimport base64from pathlib import Pathfrom typing import Optionalfrom dotenv import load_dotenvfrom pydantic import BaseModel, Fieldfrom langchain_mistralai.chat_models import ChatMistralAIfrom langchain_core.messages import HumanMessagefrom PIL import Image Loading environment variables Subsequently, the environment variables are loaded using load_dotenv(), and the MISTRAL_API_KEY is retrieved. load_dotenv()MISTRAL_API_KEY = os.getenv("MISTRAL_API_KEY")if not MISTRAL_API_KEY: raise ValueError("MISTRAL_API_KEY not set in environment") Defining the output schema with pydantic Following that, the output schema is defined using Pydantic. Pydantic is a Python library for data parsing and validation based on Python type hints. At its core, Pydantic’s BaseModel offers various useful features, such as the declaration of data types (e.g. str, int, List[str], nested models, etc.) and automatic coercion of incoming data into the required types when possible (e.g., converting "102" into 102). Moreover, it validates whether the incoming data matches the predefined schema and raises an error if it does not. Thanks to these clearly defined schemas, the data can be quickly serialized into other formats such as JSON. Likewise, Pydantic also allows the creation of document fields with metadata that tools such as LLMs can inspect and utilize. The next code block defines the structure of the expected output using Pydantic. These are the data points that the model should extract from the CV image.class BasicCV(BaseModel): first_name: Optional[str] = Field(None, description="first name") last_name: Optional[str] = Field(None, description="last name") phone: Optional[str] = Field(None, description="Telephone number") email: Optional[str] = Field(None, description="Email address") birthday: Optional[str] = Field(None, description="Date of birth (e.g., YYYY-MM-DD)") Converting images to base64 Subsequently, the first function is defined for the script. The function encode_image_to_base64() does exactly what its name suggests. It loads an image and converts it into a base64 string, which is passed into the vLM later. Moreover, an upscaling factor has been integrated. Although no additional information is gained by simply increasing the height and width, in my experience, the results tend to improve, especially in situations where the original resolution of the image is low. def encode_image_to_base64(image_path: Path, upscale_factor: float = 1.0) -> str: with Image.open(image_path) as img: if upscale_factor != 1.0: new_size = (int(img.width * upscale_factor), int(img.height * upscale_factor)) img = img.resize(new_size, Image.LANCZOS) from io import BytesIO buffer = BytesIO() img.save(buffer, format="PNG") image_bytes = buffer.getvalue() return base64.b64encode(image_bytes).decode() Processing the CV with a vision language model Now, let’s move on to the main function of this script. The process_cv() function begins by initializing the Mistral interface using a previously generated API key. This model is then wrapped using the .with_structured_output(BasicCV) function, in which the Pydantic model defined above is passed as input. If you are using a different vLM, make sure that it supports structured output, as not all vLMs do. Afterwards, the input image is converted into a base64 (b64) string, which is then transformed into a Uniform Resource Identifier (URI) by attaching a metadata string in front of the b64 string. Next, a simple system prompt is defined, which leaves room for improvement in more complex extraction tasks but works perfectly for this scenario. Finally, the URI and system prompt are combined into a LangChain HumanMessage, which is passed to the structured vLM. The model then returns the requested information in the previously defined Pydantic format. def process_cv() -> BasicCV: image_path: Path, api_key: Optional[str] = None llm = ChatMistralAI( model="pixtral-12b-latest", mistral_api_key=api_key or MISTRAL_API_KEY, ) structured_llm = llm.with_structured_output(BasicCV) image_b64 = encode_image_to_base64(image_path) data_uri = f"data:image/png;base64,{image_b64}" system_text = ( "Extract only the following fields from this CV: first name, last name, " "telephone number, email address, and birthday. Return JSON matching the schema." ) message = HumanMessage( content=[ {"type": "text", "text": system_text}, {"type": "image_url", "image_url": data_uri}, ] ) result: BasicCV = structured_llm.invoke([message]) return result Running the script This function is executed by the main, where the path is defined and the final information is printed out. if __name__ == "__main__": image_file = Path("cv-test.png") cv_data = process_cv(image_file) print(f"First Name: {cv_data.first_name}") print(f"Last Name: {cv_data.last_name}") print(f"Phone: {cv_data.phone}") print(f"Email: {cv_data.email}") print(f"Birthday: {cv_data.birthday}") Conclusion This simple Python script provides only a first impression of how powerful and flexible vLMs have become. In combination with Pydantic and with the support of the powerful LangChain framework, vLMs can be turned into a meaningful solution for many document processing workflows, such as application processing or invoice handling. What experience have you had with vision Large Language Models? Do you have other fields in mind where such a workflow might be beneficial? Source Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor. Published via Towards AI

Music-sharing platform SoundCloud is saying it “has never used artist content to train AI models,” and that it’s “making a formal commitment that any use of AI on SoundCloud will be based on consent, transparency, and artist control.” The update comes several days after artists reported that changes made last year to its terms of use could mean it reserved the right to use their music and other content to train generative AI tools.“The language in the Terms of Use was too broad and wasn’t clear enough. It created confusion, and that’s on us,” writes SoundCloud CEO Eliah Seton.The terms that SoundCloud is currently using were updated in February last year with text including this passage:In the absence of a separate agreement that states otherwise, You explicitly agree that your Content may be used to inform, train, develop or serve as input to artificial intelligence or machine intelligence technologies or services as part of and for providing the services.”But Seton says that “in the coming weeks,” that line will be replaced with this:We will not use Your Content to train generative AI models that aim to replicate or synthesize your voice, music, or likeness without your explicit consent, which must be affirmatively provided through an opt-in mechanism.Seton reiterates that SoundCloud has never used member content to train AI, including large language models, for music creation or to mimic or replace members’ work. And, echoing what a SoundCloud spokesperson told The Verge in an email over the weekend, Seton says if the company does use generative AI, it “may make this opportunity available to our human artists with their explicit consent, via an opt-in mechanism.”Ed Newton-Rex, the tech ethicist who first discovered the change, isn’t satisfied with the changes. In an X post, he says the tweaked language could still allow for “models trained on your work that might not directly replicate your style but that still compete with you in the market.” According to Rex, “If they actually want to address concerns, the change required is simple. It should just read “We will not use Your Content to train generative AI models without your explicit consent.” SoundCloud did not immediately respond to The Verge’s request for comment.See More:

Music-sharing platform SoundCloud is saying it “has never used artist content to train AI models,” and that it’s “making a formal commitment that any use of AI on SoundCloud will be based on consent, transparency, and artist control.” The update comes several days after artists reported that changes made last year to its terms of use could mean it reserved the right to use their music and other content to train generative AI tools.“The language in the Terms of Use was too broad and wasn’t clear enough. It created confusion, and that’s on us,” writes SoundCloud CEO Eliah Seton.The terms that SoundCloud is currently using were updated in February last year with text including this passage:In the absence of a separate agreement that states otherwise, You explicitly agree that your Content may be used to inform, train, develop or serve as input to artificial intelligence or machine intelligence technologies or services as part of and for providing the services.”But Seton says that “in the coming weeks,” that line will be replaced with this:We will not use Your Content to train generative AI models that aim to replicate or synthesize your voice, music, or likeness without your explicit consent, which must be affirmatively provided through an opt-in mechanism.Seton reiterates that SoundCloud has never used member content to train AI, including large language models, for music creation or to mimic or replace members’ work. And, echoing what a SoundCloud spokesperson told The Verge in an email over the weekend, Seton says if the company does use generative AI, it “may make this opportunity available to our human artists with their explicit consent, via an opt-in mechanism.”Ed Newton-Rex, the tech ethicist who first discovered the change, isn’t satisfied with the changes. In an X post, he says the tweaked language could still allow for “models trained on your work that might not directly replicate your style but that still compete with you in the market.” According to Rex, “If they actually want to address concerns, the change required is simple. It should just read “We will not use Your Content to train generative AI models without your explicit consent.” SoundCloud did not immediately respond to The Verge’s request for comment.See More:

Real-time 3D (RT3D) technology is transforming the way we work, learn, and socialize. Amid this transformation, the real-time 3D job market is growing rapidly, driven largely by increasing demand for immersive technologies in areas such as health care, gaming, architecture, and education.The growth of the real-time 3D industry has created a demand for talent, leading to a challenge for institutions and training providers to adequately prepare learners for future careers in this field. For example, there are still gaps in access and equity, particularly for those in underrepresented and low socioeconomic backgrounds. Factors such as training costs can make it difficult for learners to gain the skills needed for real-time 3D jobs, creating a demand for programs that can address these challenges.To support this, Unity Social Impact has partnered with Jobs for the Future (JFF), a leading nonprofit organization that works to create economic opportunities for underserved populations through innovative education and workforce development initiatives.Unity Social Impact and JFF are excited to announce the Unity Workforce Grant: Preparing future creators for in-demand careers. This open call for proposals will award funding to a limited number of higher education institutions, nonprofit organizations, and companies running workforce development programs to support the next generation of real-time 3D creators in accordance with the terms, conditions, and eligibility requirements described.If your organization does one or more of the following, then we encourage you to apply:Runs programs, courses, and offerings to equip learners with the skills needed to enter the real-time 3D job marketDiversifies and increases the global talent pool for real-time 3D jobsDrives training programs or designs educational content serving underrepresented, low-income, and/or underserved learner populations“3D artist. Avatar designer. Extended reality (XR) simulation developer. Technologies like virtual and augmented reality are opening exciting new pathways to quality jobs and careers – and effective workforce training will be essential for learners and workers to seize opportunities that lead to economic advancement,” says Kristina Francis, executive director of JFFLabs. “We’re delighted to partner with Unity to support innovative organizations, ensuring that the potential of training in XR/real-time development skills is truly accessible to all.”“We’re delighted to partner with Unity to support innovative organizations, ensuring that the potential of training in XR/real-time development skills is truly accessible to all.”The Unity Workforce Grant will reward up to US$200,000 to organizations training learners for the purpose of job readiness and career placement in the realm of immersive technologies and real-time 3D.“Rapidly upskilling into high-demand jobs, like real-time 3D, is crucial to building a thriving economy and society,” adds Jessica Lindl, vice president of Unity Social Impact. “It’s our responsibility to ensure that everyone, regardless of their background, has access to economic opportunities. The Unity Workforce Grant, in partnership with Jobs for the Future, is an important step in making this happen.”Grant applications will be accepted from April 17 to June 2, 2023 at 11:59 PM Pacific Time. Learn more about the grant and eligibility requirements.This grant initiative is part of the Unity Charitable Fund, which previously funded efforts such as the Higher Ed XR Innovation Open Call and EdTech Creator Challenge.Still have questions that you’d like answered live? Register to join us for a one-hour webinar on Monday, April 24 or one of two live Q&A sessions on either May 15 or May 22.