June 13, 20256 min readAir-Conditioning Can Surprisingly Help the Power Grid during Extreme HeatSwitching on air-conditioning during extreme heat doesn’t have to make us feel guilty—it can actually boost power grid reliability and help bring more renewable energy onlineBy Johanna Mathieu & The Conversation US Imagedepotpro/Getty ImagesThe following essay is reprinted with permission from The Conversation, an online publication covering the latest research.As summer arrives, people are turning on air conditioners in most of the U.S. But if you’re like me, you always feel a little guilty about that. Past generations managed without air conditioning – do I really need it? And how bad is it to use all this electricity for cooling in a warming world?If I leave my air conditioner off, I get too hot. But if everyone turns on their air conditioner at the same time, electricity demand spikes, which can force power grid operators to activate some of the most expensive, and dirtiest, power plants. Sometimes those spikes can ask too much of the grid and lead to brownouts or blackouts.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.Research I recently published with a team of scholars makes me feel a little better, though. We have found that it is possible to coordinate the operation of large numbers of home air-conditioning units, balancing supply and demand on the power grid – and without making people endure high temperatures inside their homes.Studies along these lines, using remote control of air conditioners to support the grid, have for many years explored theoretical possibilities like this. However, few approaches have been demonstrated in practice and never for such a high-value application and at this scale. The system we developed not only demonstrated the ability to balance the grid on timescales of seconds, but also proved it was possible to do so without affecting residents’ comfort.The benefits include increasing the reliability of the power grid, which makes it easier for the grid to accept more renewable energy. Our goal is to turn air conditioners from a challenge for the power grid into an asset, supporting a shift away from fossil fuels toward cleaner energy.Adjustable equipmentMy research focuses on batteries, solar panels and electric equipment – such as electric vehicles, water heaters, air conditioners and heat pumps – that can adjust itself to consume different amounts of energy at different times.Originally, the U.S. electric grid was built to transport electricity from large power plants to customers’ homes and businesses. And originally, power plants were large, centralized operations that burned coal or natural gas, or harvested energy from nuclear reactions. These plants were typically always available and could adjust how much power they generated in response to customer demand, so the grid would be balanced between power coming in from producers and being used by consumers.But the grid has changed. There are more renewable energy sources, from which power isn’t always available – like solar panels at night or wind turbines on calm days. And there are the devices and equipment I study. These newer options, called “distributed energy resources,” generate or store energy near where consumers need it – or adjust how much energy they’re using in real time.One aspect of the grid hasn’t changed, though: There’s not much storage built into the system. So every time you turn on a light, for a moment there’s not enough electricity to supply everything that wants it right then: The grid needs a power producer to generate a little more power. And when you turn off a light, there’s a little too much: A power producer needs to ramp down.The way power plants know what real-time power adjustments are needed is by closely monitoring the grid frequency. The goal is to provide electricity at a constant frequency – 60 hertz – at all times. If more power is needed than is being produced, the frequency drops and a power plant boosts output. If there’s too much power being produced, the frequency rises and a power plant slows production a little. These actions, a process called “frequency regulation,” happen in a matter of seconds to keep the grid balanced.This output flexibility, primarily from power plants, is key to keeping the lights on for everyone.Finding new optionsI’m interested in how distributed energy resources can improve flexibility in the grid. They can release more energy, or consume less, to respond to the changing supply or demand, and help balance the grid, ensuring the frequency remains near 60 hertz.Some people fear that doing so might be invasive, giving someone outside your home the ability to control your battery or air conditioner. Therefore, we wanted to see if we could help balance the grid with frequency regulation using home air-conditioning units rather than power plants – without affecting how residents use their appliances or how comfortable they are in their homes.From 2019 to 2023, my group at the University of Michigan tried this approach, in collaboration with researchers at Pecan Street Inc., Los Alamos National Laboratory and the University of California, Berkeley, with funding from the U.S. Department of Energy Advanced Research Projects Agency-Energy.We recruited 100 homeowners in Austin, Texas, to do a real-world test of our system. All the homes had whole-house forced-air cooling systems, which we connected to custom control boards and sensors the owners allowed us to install in their homes. This equipment let us send instructions to the air-conditioning units based on the frequency of the grid.Before I explain how the system worked, I first need to explain how thermostats work. When people set thermostats, they pick a temperature, and the thermostat switches the air-conditioning compressor on and off to maintain the air temperature within a small range around that set point. If the temperature is set at 68 degrees, the thermostat turns the AC on when the temperature is, say, 70, and turns it off when it’s cooled down to, say, 66.Every few seconds, our system slightly changed the timing of air-conditioning compressor switching for some of the 100 air conditioners, causing the units’ aggregate power consumption to change. In this way, our small group of home air conditioners reacted to grid changes the way a power plant would – using more or less energy to balance the grid and keep the frequency near 60 hertz.Moreover, our system was designed to keep home temperatures within the same small temperature range around the set point.Testing the approachWe ran our system in four tests, each lasting one hour. We found two encouraging results.First, the air conditioners were able to provide frequency regulation at least as accurately as a traditional power plant. Therefore, we showed that air conditioners could play a significant role in increasing grid flexibility. But perhaps more importantly – at least in terms of encouraging people to participate in these types of systems – we found that we were able to do so without affecting people’s comfort in their homes.We found that home temperatures did not deviate more than 1.6 Fahrenheit from their set point. Homeowners were allowed to override the controls if they got uncomfortable, but most didn’t. For most tests, we received zero override requests. In the worst case, we received override requests from two of the 100 homes in our test.In practice, this sort of technology could be added to commercially available internet-connected thermostats. In exchange for credits on their energy bills, users could choose to join a service run by the thermostat company, their utility provider or some other third party.Then people could turn on the air conditioning in the summer heat without that pang of guilt, knowing they were helping to make the grid more reliable and more capable of accommodating renewable energy sources – without sacrificing their own comfort in the process.This article was originally published on The Conversation. Read the original article.

Posted on : June 13, 2025 By Tech World Times Uncategorized  Rate this post Sorting is important in programming. It helps organize data. Sorting improves performance in searching, analysis, and reporting. There are many sorting algorithms. One of the simplest is Insertion Sort. In this article, we will learn how to implement Insertion Sort in Java. We will explain each step in simple words. You will see examples and understand how it works. What Is Insertion Sort? Insertion Sort is a simple sorting algorithm. It works like how you sort playing cards. You take one card at a time and place it in the right position. It compares the current element with those before it. If needed, it shifts elements to the right. Then, it inserts the current element at the correct place. How Insertion Sort Works Let’s understand with a small list: Example List: [8, 3, 5, 1] Steps: First element (8) is already sorted. Compare 3 with 8. Move 8 right. Insert 3 before it → [3, 8, 5, 1] Compare 5 with 8. Move 8 right. Insert 5 after 3 → [3, 5, 8, 1] Compare 1 with 8, 5, 3. Move them right. Insert 1 at start → [1, 3, 5, 8] Now the list is sorted! Why Use Insertion Sort? Insertion Sort is simple and easy to code. It works well for: Small datasets Nearly sorted lists Educational purposes and practice However, it is not good for large datasets. It has a time complexity of O(n²). Time Complexity of Insertion Sort Best Case (already sorted): O(n) Average Case: O(n²) Worst Case (reversed list): O(n²) It performs fewer steps in nearly sorted data. How to Implement Insertion Sort in Java Now let’s write the code for Insertion Sort in Java. We will explain each part. Step 1: Define a Class javaCopyEditpublic class InsertionSortExample { // Code goes here } We create a class named InsertionSortExample. Step 2: Create the Sorting Method javaCopyEditpublic static void insertionSort(int[] arr) { int n = arr.length; for (int i = 1; i < n; i++) { int key = arr[i]; int j = i - 1; while (j >= 0 && arr[j] > key) { arr[j + 1] = arr[j]; j = j - 1; } arr[j + 1] = key; } } Let’s break it down: arr[i] is the current value (called key). j starts from the previous index. While arr[j] > key, shift arr[j] to the right. Insert the key at the correct position. This logic sorts the array step by step. Step 3: Create the Main Method Now we test the code. javaCopyEditpublic static void main(String[] args) { int[] numbers = {9, 5, 1, 4, 3}; System.out.println("Before sorting:"); printArray(numbers); insertionSort(numbers); System.out.println("After sorting:"); printArray(numbers); } This method: Creates an array of numbers Prints the array before sorting Calls the sort method Prints the array after sorting Step 4: Print the Array Let’s add a helper method to print the array. javaCopyEditpublic static void printArray(int[] arr) { for (int number : arr) { System.out.print(number + " "); } System.out.println(); } Now you can see how the array changes before and after sorting. Full Code Example javaCopyEditpublic class InsertionSortExample { public static void insertionSort(int[] arr) { int n = arr.length; for (int i = 1; i < n; i++) { int key = arr[i]; int j = i - 1; while (j >= 0 && arr[j] > key) { arr[j + 1] = arr[j]; j = j - 1; } arr[j + 1] = key; } } public static void printArray(int[] arr) { for (int number : arr) { System.out.print(number + " "); } System.out.println(); } public static void main(String[] args) { int[] numbers = {9, 5, 1, 4, 3}; System.out.println("Before sorting:"); printArray(numbers); insertionSort(numbers); System.out.println("After sorting:"); printArray(numbers); } } Sample Output yamlCopyEditBefore sorting: 9 5 1 4 3 After sorting: 1 3 4 5 9 This confirms that the sorting works correctly. Advantages of Insertion Sort in Java Easy to implement Works well with small inputs Stable sort (keeps equal items in order) Good for educational use When Not to Use Insertion Sort Avoid Insertion Sort when: The dataset is large Performance is critical Better algorithms like Merge Sort or Quick Sort are available Real-World Uses Sorting small records in a database Teaching algorithm basics Handling partially sorted arrays Even though it is not the fastest, it is useful in many simple tasks. Final Tips Practice with different inputs Add print statements to see how it works Try sorting strings or objects Use Java’s built-in sort methods for large arrays Conclusion Insertion Sort in Java is a great way to learn sorting. It is simple and easy to understand. In this guide, we showed how to implement it step-by-step. We covered the logic, code, and output. We also explained when to use it. Now you can try it yourself. Understanding sorting helps in coding interviews and software development. Keep practicing and exploring other sorting methods too. The more you practice, the better you understand algorithms. Tech World TimesTech World Times (TWT), a global collective focusing on the latest tech news and trends in blockchain, Fintech, Development & Testing, AI and Startups. If you are looking for the guest post then contact at techworldtimes@gmail.com

Transcript [MUSIC]      [BOOK PASSAGE]   PETER LEE: “In ‘The Little Black Bag,’ a classic science fiction story, a high-tech doctor’s kit of the future is accidentally transported back to the 1950s, into the shaky hands of a washed-up, alcoholic doctor. The ultimate medical tool, it redeems the doctor wielding it, allowing him to practice gratifyingly heroic medicine. … The tale ends badly for the doctor and his treacherous assistant, but it offered a picture of how advanced technology could transform medicine—powerful when it was written nearly 75 years ago and still so today. What would be the Al equivalent of that little black bag? At this moment when new capabilities are emerging, how do we imagine them into medicine?”   [END OF BOOK PASSAGE]     [THEME MUSIC]     This is The AI Revolution in Medicine, Revisited. I’m your host, Peter Lee.    Shortly after OpenAI’s GPT-4 was publicly released, Carey Goldberg, Dr. Zak Kohane, and I published The AI Revolution in Medicine to help educate the world of healthcare and medical research about the transformative impact this new generative AI technology could have. But because we wrote the book when GPT-4 was still a secret, we had to speculate. Now, two years later, what did we get right, and what did we get wrong?     In this series, we’ll talk to clinicians, patients, hospital administrators, and others to understand the reality of AI in the field and where we go from here.   [THEME MUSIC FADES] The book passage I read at the top is from “Chapter 10: The Big Black Bag.”  In imagining AI in medicine, Carey, Zak, and I included in our book two fictional accounts. In the first, a medical resident consults GPT-4 on her personal phone as the patient in front of her crashes. Within seconds, it offers an alternate response based on recent literature. In the second account, a 90-year-old woman with several chronic conditions is living independently and receiving near-constant medical support from an AI aide.    In our conversations with the guests we’ve spoken to so far, we’ve caught a glimpse of these predicted futures, seeing how clinicians and patients are actually using AI today and how developers are leveraging the technology in the healthcare products and services they’re creating. In fact, that first fictional account isn’t so fictional after all, as most of the doctors in the real world actually appear to be using AI at least occasionally—and sometimes much more than occasionally—to help in their daily clinical work. And as for the second fictional account, which is more of a science fiction account, it seems we are indeed on the verge of a new way of delivering and receiving healthcare, though the future is still very much open.  As we continue to examine the current state of AI in healthcare and its potential to transform the field, I’m pleased to welcome Bill Gates and Sébastien Bubeck.   Bill may be best known as the co-founder of Microsoft, having created the company with his childhood friend Paul Allen in 1975. He’s now the founder of Breakthrough Energy, which aims to advance clean energy innovation, and TerraPower, a company developing groundbreaking nuclear energy and science technologies. He also chairs the world’s largest philanthropic organization, the Gates Foundation, and focuses on solving a variety of health challenges around the globe and here at home.  Sébastien is a research lead at OpenAI. He was previously a distinguished scientist, vice president of AI, and a colleague of mine here at Microsoft, where his work included spearheading the development of the family of small language models known as Phi. While at Microsoft, he also coauthored the discussion-provoking 2023 paper “Sparks of Artificial General Intelligence,” which presented the results of early experiments with GPT-4 conducted by a small team from Microsoft Research.    [TRANSITION MUSIC]   Here’s my conversation with Bill Gates and Sébastien Bubeck.  LEE: Bill, welcome.  BILL GATES: Thank you.  LEE: Seb …  SÉBASTIEN BUBECK: Yeah. Hi, hi, Peter. Nice to be here.  LEE: You know, one of the things that I’ve been doing just to get the conversation warmed up is to talk about origin stories, and what I mean about origin stories is, you know, what was the first contact that you had with large language models or the concept of generative AI that convinced you or made you think that something really important was happening?  And so, Bill, I think I’ve heard the story about, you know, the time when the OpenAI folks—Sam Altman, Greg Brockman, and others—showed you something, but could we hear from you what those early encounters were like and what was going through your mind?   GATES: Well, I’d been visiting OpenAI soon after it was created to see things like GPT-2 and to see the little arm they had that was trying to match human manipulation and, you know, looking at their games like Dota that they were trying to get as good as human play. And honestly, I didn’t think the language model stuff they were doing, even when they got to GPT-3, would show the ability to learn, you know, in the same sense that a human reads a biology book and is able to take that knowledge and access it not only to pass a test but also to create new medicines.  And so my challenge to them was that if their LLM could get a five on the advanced placement biology test, then I would say, OK, it took biologic knowledge and encoded it in an accessible way and that I didn’t expect them to do that very quickly but it would be profound.   And it was only about six months after I challenged them to do that, that an early version of GPT-4 they brought up to a dinner at my house, and in fact, it answered most of the questions that night very well. The one it got totally wrong, we were … because it was so good, we kept thinking, Oh, we must be wrong. It turned out it was a math weakness [LAUGHTER] that, you know, we later understood that that was an area of, weirdly, of incredible weakness of those early models. But, you know, that was when I realized, OK, the age of cheap intelligence was at its beginning.  LEE: Yeah. So I guess it seems like you had something similar to me in that my first encounters, I actually harbored some skepticism. Is it fair to say you were skeptical before that?  GATES: Well, the idea that we’ve figured out how to encode and access knowledge in this very deep sense without even understanding the nature of the encoding, …  LEE: Right.   GATES: … that is a bit weird.   LEE: Yeah.  GATES: We have an algorithm that creates the computation, but even say, OK, where is the president’s birthday stored in there? Where is this fact stored in there? The fact that even now when we’re playing around, getting a little bit more sense of it, it’s opaque to us what the semantic encoding is, it’s, kind of, amazing to me. I thought the invention of knowledge storage would be an explicit way of encoding knowledge, not an implicit statistical training.  LEE: Yeah, yeah. All right. So, Seb, you know, on this same topic, you know, I got—as we say at Microsoft—I got pulled into the tent. [LAUGHS]  BUBECK: Yes.   LEE: Because this was a very secret project. And then, um, I had the opportunity to select a small number of researchers in MSR [Microsoft Research] to join and start investigating this thing seriously. And the first person I pulled in was you.  BUBECK: Yeah.  LEE: And so what were your first encounters? Because I actually don’t remember what happened then.  BUBECK: Oh, I remember it very well. [LAUGHS] My first encounter with GPT-4 was in a meeting with the two of you, actually. But my kind of first contact, the first moment where I realized that something was happening with generative AI, was before that. And I agree with Bill that I also wasn’t too impressed by GPT-3.  I though that it was kind of, you know, very naturally mimicking the web, sort of parroting what was written there in a nice way. Still in a way which seemed very impressive. But it wasn’t really intelligent in any way. But shortly after GPT-3, there was a model before GPT-4 that really shocked me, and this was the first image generation model, DALL-E 1.  So that was in 2021. And I will forever remember the press release of OpenAI where they had this prompt of an avocado chair and then you had this image of the avocado chair. [LAUGHTER] And what really shocked me is that clearly the model kind of “understood” what is a chair, what is an avocado, and was able to merge those concepts.  So this was really, to me, the first moment where I saw some understanding in those models.   LEE: So this was, just to get the timing right, that was before I pulled you into the tent.  BUBECK: That was before. That was like a year before.  LEE: Right.   BUBECK: And now I will tell you how, you know, we went from that moment to the meeting with the two of you and GPT-4.  So once I saw this kind of understanding, I thought, OK, fine. It understands concept, but it’s still not able to reason. It cannot—as, you know, Bill was saying—it cannot learn from your document. It cannot reason.   So I set out to try to prove that. You know, this is what I was in the business of at the time, trying to prove things in mathematics. So I was trying to prove that basically autoregressive transformers could never reason. So I was trying to prove this. And after a year of work, I had something reasonable to show. And so I had the meeting with the two of you, and I had this example where I wanted to say, there is no way that an LLM is going to be able to do x.  And then as soon as I … I don’t know if you remember, Bill. But as soon as I said that, you said, oh, but wait a second. I had, you know, the OpenAI crew at my house recently, and they showed me a new model. Why don’t we ask this new model this question?   LEE: Yeah. BUBECK: And we did, and it solved it on the spot. And that really, honestly, just changed my life. Like, you know, I had been working for a year trying to say that this was impossible. And just right there, it was shown to be possible.   LEE: [LAUGHS] One of the very first things I got interested in—because I was really thinking a lot about healthcare—was healthcare and medicine.  And I don’t know if the two of you remember, but I ended up doing a lot of tests. I ran through, you know, step one and step two of the US Medical Licensing Exam. Did a whole bunch of other things. I wrote this big report. It was, you know, I can’t remember … a couple hundred pages.   And I needed to share this with someone. I didn’t … there weren’t too many people I could share it with. So I sent, I think, a copy to you, Bill. Sent a copy to you, Seb.   I hardly slept for about a week putting that report together. And, yeah, and I kept working on it. But I was far from alone. I think everyone who was in the tent, so to speak, in those early days was going through something pretty similar. All right. So I think … of course, a lot of what I put in the report also ended up being examples that made it into the book.  But the main purpose of this conversation isn’t to reminisce about [LAUGHS] or indulge in those reminiscences but to talk about what’s happening in healthcare and medicine. And, you know, as I said, we wrote this book. We did it very, very quickly. Seb, you helped. Bill, you know, you provided a review and some endorsements.  But, you know, honestly, we didn’t know what we were talking about because no one had access to this thing. And so we just made a bunch of guesses. So really, the whole thing I wanted to probe with the two of you is, now with two years of experience out in the world, what, you know, what do we think is happening today?  You know, is AI actually having an impact, positive or negative, on healthcare and medicine? And what do we now think is going to happen in the next two years, five years, or 10 years? And so I realize it’s a little bit too abstract to just ask it that way. So let me just try to narrow the discussion and guide us a little bit.   Um, the kind of administrative and clerical work, paperwork, around healthcare—and we made a lot of guesses about that—that appears to be going well, but, you know, Bill, I know we’ve discussed that sometimes that you think there ought to be a lot more going on. Do you have a viewpoint on how AI is actually finding its way into reducing paperwork?  GATES: Well, I’m stunned … I don’t think there should be a patient-doctor meeting where the AI is not sitting in and both transcribing, offering to help with the paperwork, and even making suggestions, although the doctor will be the one, you know, who makes the final decision about the diagnosis and whatever prescription gets done.   It’s so helpful. You know, when that patient goes home and their, you know, son who wants to understand what happened has some questions, that AI should be available to continue that conversation. And the way you can improve that experience and streamline things and, you know, involve the people who advise you. I don’t understand why that’s not more adopted, because there you still have the human in the loop making that final decision.  But even for, like, follow-up calls to make sure the patient did things, to understand if they have concerns and knowing when to escalate back to the doctor, the benefit is incredible. And, you know, that thing is ready for prime time. That paradigm is ready for prime time, in my view.  LEE: Yeah, there are some good products, but it seems like the number one use right now—and we kind of got this from some of the previous guests in previous episodes—is the use of AI just to respond to emails from patients. [LAUGHTER] Does that make sense to you?  BUBECK: Yeah. So maybe I want to second what Bill was saying but maybe take a step back first. You know, two years ago, like, the concept of clinical scribes, which is one of the things that we’re talking about right now, it would have sounded, in fact, it sounded two years ago, borderline dangerous. Because everybody was worried about hallucinations. What happened if you have this AI listening in and then it transcribes, you know, something wrong?  Now, two years later, I think it’s mostly working. And in fact, it is not yet, you know, fully adopted. You’re right. But it is in production. It is used, you know, in many, many places. So this rate of progress is astounding because it wasn’t obvious that we would be able to overcome those obstacles of hallucination. It’s not to say that hallucinations are fully solved. In the case of the closed system, they are.   Now, I think more generally what’s going on in the background is that there is something that we, that certainly I, underestimated, which is this management overhead. So I think the reason why this is not adopted everywhere is really a training and teaching aspect. People need to be taught, like, those systems, how to interact with them.  And one example that I really like, a study that recently appeared where they tried to use ChatGPT for diagnosis and they were comparing doctors without and with ChatGPT (opens in new tab). And the amazing thing … so this was a set of cases where the accuracy of the doctors alone was around 75%. ChatGPT alone was 90%. So that’s already kind of mind blowing. But then the kicker is that doctors with ChatGPT was 80%.   Intelligence alone is not enough. It’s also how it’s presented, how you interact with it. And ChatGPT, it’s an amazing tool. Obviously, I absolutely love it. But it’s not … you don’t want a doctor to have to type in, you know, prompts and use it that way.  It should be, as Bill was saying, kind of running continuously in the background, sending you notifications. And you have to be really careful of the rate at which those notifications are being sent. Because if they are too frequent, then the doctor will learn to ignore them. So you have to … all of those things matter, in fact, at least as much as the level of intelligence of the machine.  LEE: One of the things I think about, Bill, in that scenario that you described, doctors do some thinking about the patient when they write the note. So, you know, I’m always a little uncertain whether it’s actually … you know, you wouldn’t necessarily want to fully automate this, I don’t think. Or at least there needs to be some prompt to the doctor to make sure that the doctor puts some thought into what happened in the encounter with the patient. Does that make sense to you at all?  GATES: At this stage, you know, I’d still put the onus on the doctor to write the conclusions and the summary and not delegate that.  The tradeoffs you make a little bit are somewhat dependent on the situation you’re in. If you’re in Africa, So, yes, the doctor’s still going to have to do a lot of work, but just the quality of letting the patient and the people around them interact and ask questions and have things explained, that alone is such a quality improvement. It’s mind blowing.   LEE: So since you mentioned, you know, Africa—and, of course, this touches on the mission and some of the priorities of the Gates Foundation and this idea of democratization of access to expert medical care—what’s the most interesting stuff going on right now? Are there people and organizations or technologies that are impressing you or that you’re tracking?  GATES: Yeah. So the Gates Foundation has given out a lot of grants to people in Africa doing education, agriculture but more healthcare examples than anything. And the way these things start off, they often start out either being patient-centric in a narrow situation, like, OK, I’m a pregnant woman; talk to me. Or, I have infectious disease symptoms; talk to me. Or they’re connected to a health worker where they’re helping that worker get their job done. And we have lots of pilots out, you know, in both of those cases.   The dream would be eventually to have the thing the patient consults be so broad that it’s like having a doctor available who understands the local things.   LEE: Right.   GATES: We’re not there yet. But over the next two or three years, you know, particularly given the worsening financial constraints against African health systems, where the withdrawal of money has been dramatic, you know, figuring out how to take this—what I sometimes call “free intelligence”—and build a quality health system around that, we will have to be more radical in low-income countries than any rich country is ever going to be.   LEE: Also, there’s maybe a different regulatory environment, so some of those things maybe are easier? Because right now, I think the world hasn’t figured out how to and whether to regulate, let’s say, an AI that might give a medical diagnosis or write a prescription for a medication.  BUBECK: Yeah. I think one issue with this, and it’s also slowing down the deployment of AI in healthcare more generally, is a lack of proper benchmark. Because, you know, you were mentioning the USMLE [United States Medical Licensing Examination], for example. That’s a great test to test human beings and their knowledge of healthcare and medicine. But it’s not a great test to give to an AI.  It’s not asking the right questions. So finding what are the right questions to test whether an AI system is ready to give diagnosis in a constrained setting, that’s a very, very important direction, which to my surprise, is not yet accelerating at the rate that I was hoping for.  LEE: OK, so that gives me an excuse to get more now into the core AI tech because something I’ve discussed with both of you is this issue of what are the right tests. And you both know the very first test I give to any new spin of an LLM is I present a patient, the results—a mythical patient—the results of my physical exam, my mythical physical exam. Maybe some results of some initial labs. And then I present or propose a differential diagnosis. And if you’re not in medicine, a differential diagnosis you can just think of as a prioritized list of the possible diagnoses that fit with all that data. And in that proposed differential, I always intentionally make two mistakes.  I make a textbook technical error in one of the possible elements of the differential diagnosis, and I have an error of omission. And, you know, I just want to know, does the LLM understand what I’m talking about? And all the good ones out there do now. But then I want to know, can it spot the errors? And then most importantly, is it willing to tell me I’m wrong, that I’ve made a mistake?   That last piece seems really hard for AI today. And so let me ask you first, Seb, because at the time of this taping, of course, there was a new spin of GPT-4o last week that became overly sycophantic. In other words, it was actually prone in that test of mine not only to not tell me I’m wrong, but it actually praised me for the creativity of my differential. [LAUGHTER] What’s up with that?  BUBECK: Yeah, I guess it’s a testament to the fact that training those models is still more of an art than a science. So it’s a difficult job. Just to be clear with the audience, we have rolled back that [LAUGHS] version of GPT-4o, so now we don’t have the sycophant version out there.  Yeah, no, it’s a really difficult question. It has to do … as you said, it’s very technical. It has to do with the post-training and how, like, where do you nudge the model? So, you know, there is this very classical by now technique called RLHF [reinforcement learning from human feedback], where you push the model in the direction of a certain reward model. So the reward model is just telling the model, you know, what behavior is good, what behavior is bad.  But this reward model is itself an LLM, and, you know, Bill was saying at the very beginning of the conversation that we don’t really understand how those LLMs deal with concepts like, you know, where is the capital of France located? Things like that. It is the same thing for this reward model. We don’t know why it says that it prefers one output to another, and whether this is correlated with some sycophancy is, you know, something that we discovered basically just now. That if you push too hard in optimization on this reward model, you will get a sycophant model.  So it’s kind of … what I’m trying to say is we became too good at what we were doing, and we ended up, in fact, in a trap of the reward model.  LEE: I mean, you do want … it’s a difficult balance because you do want models to follow your desires and …  BUBECK: It’s a very difficult, very difficult balance.  LEE: So this brings up then the following question for me, which is the extent to which we think we’ll need to have specially trained models for things. So let me start with you, Bill. Do you have a point of view on whether we will need to, you know, quote-unquote take AI models to med school? Have them specially trained? Like, if you were going to deploy something to give medical care in underserved parts of the world, do we need to do something special to create those models?  GATES: We certainly need to teach them the African languages and the unique dialects so that the multimedia interactions are very high quality. We certainly need to teach them the disease prevalence and unique disease patterns like, you know, neglected tropical diseases and malaria. So we need to gather a set of facts that somebody trying to go for a US customer base, you know, wouldn’t necessarily have that in there.  Those two things are actually very straightforward because the additional training time is small. I’d say for the next few years, we’ll also need to do reinforcement learning about the context of being a doctor and how important certain behaviors are. Humans learn over the course of their life to some degree that, I’m in a different context and the way I behave in terms of being willing to criticize or be nice, you know, how important is it? Who’s here? What’s my relationship to them?   Right now, these machines don’t have that broad social experience. And so if you know it’s going to be used for health things, a lot of reinforcement learning of the very best humans in that context would still be valuable. Eventually, the models will, having read all the literature of the world about good doctors, bad doctors, it’ll understand as soon as you say, “I want you to be a doctor diagnosing somebody.” All of the implicit reinforcement that fits that situation, you know, will be there. LEE: Yeah. GATES: And so I hope three years from now, we don’t have to do that reinforcement learning. But today, for any medical context, you would want a lot of data to reinforce tone, willingness to say things when, you know, there might be something significant at stake.  LEE: Yeah. So, you know, something Bill said, kind of, reminds me of another thing that I think we missed, which is, the context also … and the specialization also pertains to different, I guess, what we still call “modes,” although I don’t know if the idea of multimodal is the same as it was two years ago. But, you know, what do you make of all of the hubbub around—in fact, within Microsoft Research, this is a big deal, but I think we’re far from alone—you know, medical images and vision, video, proteins and molecules, cell, you know, cellular data and so on.  BUBECK: Yeah. OK. So there is a lot to say to everything … to the last, you know, couple of minutes. Maybe on the specialization aspect, you know, I think there is, hiding behind this, a really fundamental scientific question of whether eventually we have a singular AGI [artificial general intelligence] that kind of knows everything and you can just put, you know, explain your own context and it will just get it and understand everything.  That’s one vision. I have to say, I don’t particularly believe in this vision. In fact, we humans are not like that at all. I think, hopefully, we are general intelligences, yet we have to specialize a lot. And, you know, I did myself a lot of RL, reinforcement learning, on mathematics. Like, that’s what I did, you know, spent a lot of time doing that. And I didn’t improve on other aspects. You know, in fact, I probably degraded in other aspects. [LAUGHTER] So it’s … I think it’s an important example to have in mind.  LEE: I think I might disagree with you on that, though, because, like, doesn’t a model have to see both good science and bad science in order to be able to gain the ability to discern between the two?  BUBECK: Yeah, no, that absolutely. I think there is value in seeing the generality, in having a very broad base. But then you, kind of, specialize on verticals. And this is where also, you know, open-weights model, which we haven’t talked about yet, are really important because they allow you to provide this broad base to everyone. And then you can specialize on top of it.  LEE: So we have about three hours of stuff to talk about, but our time is actually running low. BUBECK: Yes, yes, yes.   LEE: So I think I want … there’s a more provocative question. It’s almost a silly question, but I need to ask it of the two of you, which is, is there a future, you know, where AI replaces doctors or replaces, you know, medical specialties that we have today? So what does the world look like, say, five years from now?  GATES: Well, it’s important to distinguish healthcare discovery activity from healthcare delivery activity. We focused mostly on delivery. I think it’s very much within the realm of possibility that the AI is not only accelerating healthcare discovery but substituting for a lot of the roles of, you know, I’m an organic chemist, or I run various types of assays. I can see those, which are, you know, testable-output-type jobs but with still very high value, I can see, you know, some replacement in those areas before the doctor.   The doctor, still understanding the human condition and long-term dialogues, you know, they’ve had a lifetime of reinforcement of that, particularly when you get into areas like mental health. So I wouldn’t say in five years, either people will choose to adopt it, but it will be profound that there’ll be this nearly free intelligence that can do follow-up, that can help you, you know, make sure you went through different possibilities.  And so I’d say, yes, we’ll have doctors, but I’d say healthcare will be massively transformed in its quality and in efficiency by AI in that time period.  LEE: Is there a comparison, useful comparison, say, between doctors and, say, programmers, computer programmers, or doctors and, I don’t know, lawyers?  GATES: Programming is another one that has, kind of, a mathematical correctness to it, you know, and so the objective function that you’re trying to reinforce to, as soon as you can understand the state machines, you can have something that’s “checkable”; that’s correct. So I think programming, you know, which is weird to say, that the machine will beat us at most programming tasks before we let it take over roles that have deep empathy, you know, physical presence and social understanding in them.  LEE: Yeah. By the way, you know, I fully expect in five years that AI will produce mathematical proofs that are checkable for validity, easily checkable, because they’ll be written in a proof-checking language like Lean or something but will be so complex that no human mathematician can understand them. I expect that to happen.   I can imagine in some fields, like cellular biology, we could have the same situation in the future because the molecular pathways, the chemistry, biochemistry of human cells or living cells is as complex as any mathematics, and so it seems possible that we may be in a state where in wet lab, we see, Oh yeah, this actually works, but no one can understand why.  BUBECK: Yeah, absolutely. I mean, I think I really agree with Bill’s distinction of the discovery and the delivery, and indeed, the discovery’s when you can check things, and at the end, there is an artifact that you can verify. You know, you can run the protocol in the wet lab and see [if you have] produced what you wanted. So I absolutely agree with that.   And in fact, you know, we don’t have to talk five years from now. I don’t know if you know, but just recently, there was a paper that was published on a scientific discovery using o3- mini (opens in new tab). So this is really amazing. And, you know, just very quickly, just so people know, it was about this statistical physics model, the frustrated Potts model, which has to do with coloring, and basically, the case of three colors, like, more than two colors was open for a long time, and o3 was able to reduce the case of three colors to two colors.   LEE: Yeah.  BUBECK: Which is just, like, astounding. And this is not … this is now. This is happening right now. So this is something that I personally didn’t expect it would happen so quickly, and it’s due to those reasoning models.   Now, on the delivery side, I would add something more to it for the reason why doctors and, in fact, lawyers and coders will remain for a long time, and it’s because we still don’t understand how those models generalize. Like, at the end of the day, we are not able to tell you when they are confronted with a really new, novel situation, whether they will work or not.  Nobody is able to give you that guarantee. And I think until we understand this generalization better, we’re not going to be willing to just let the system in the wild without human supervision.  LEE: But don’t human doctors, human specialists … so, for example, a cardiologist sees a patient in a certain way that a nephrologist …  BUBECK: Yeah. LEE: … or an endocrinologist might not. BUBECK: That’s right. But another cardiologist will understand and, kind of, expect a certain level of generalization from their peer. And this, we just don’t have it with AI models. Now, of course, you’re exactly right. That generalization is also hard for humans. Like, if you have a human trained for one task and you put them into another task, then you don’t … you often don’t know. LEE: OK. You know, the podcast is focused on what’s happened over the last two years. But now, I’d like one provocative prediction about what you think the world of AI and medicine is going to be at some point in the future. You pick your timeframe. I don’t care if it’s two years or 20 years from now, but, you know, what do you think will be different about AI in medicine in that future than today?  BUBECK: Yeah, I think the deployment is going to accelerate soon. Like, we’re really not missing very much. There is this enormous capability overhang. Like, even if progress completely stopped, with current systems, we can do a lot more than what we’re doing right now. So I think this will … this has to be realized, you know, sooner rather than later.  And I think it’s probably dependent on these benchmarks and proper evaluation and tying this with regulation. So these are things that take time in human society and for good reason. But now we already are at two years; you know, give it another two years and it should be really …   LEE: Will AI prescribe your medicines? Write your prescriptions?  BUBECK: I think yes. I think yes.  LEE: OK. Bill?  GATES: Well, I think the next two years, we’ll have massive pilots, and so the amount of use of the AI, still in a copilot-type mode, you know, we should get millions of patient visits, you know, both in general medicine and in the mental health side, as well. And I think that’s going to build up both the data and the confidence to give the AI some additional autonomy. You know, are you going to let it talk to you at night when you’re panicked about your mental health with some ability to escalate? And, you know, I’ve gone so far as to tell politicians with national health systems that if they deploy AI appropriately, that the quality of care, the overload of the doctors, the improvement in the economics will be enough that their voters will be stunned because they just don’t expect this, and, you know, they could be reelected [LAUGHTER] just on this one thing of fixing what is a very overloaded and economically challenged health system in these rich countries.  You know, my personal role is going to be to make sure that in the poorer countries, there isn’t some lag; in fact, in many cases, that we’ll be more aggressive because, you know, we’re comparing to having no access to doctors at all. And, you know, so I think whether it’s India or Africa, there’ll be lessons that are globally valuable because we need medical intelligence. And, you know, thank god AI is going to provide a lot of that.  LEE: Well, on that optimistic note, I think that’s a good way to end. Bill, Seb, really appreciate all of this.   I think the most fundamental prediction we made in the book is that AI would actually find its way into the practice of medicine, and I think that that at least has come true, maybe in different ways than we expected, but it’s come true, and I think it’ll only accelerate from here. So thanks again, both of you.  [TRANSITION MUSIC]  GATES: Yeah. Thanks, you guys.  BUBECK: Thank you, Peter. Thanks, Bill.  LEE: I just always feel such a sense of privilege to have a chance to interact and actually work with people like Bill and Sébastien.    With Bill, I’m always amazed at how practically minded he is. He’s really thinking about the nuts and bolts of what AI might be able to do for people, and his thoughts about underserved parts of the world, the idea that we might actually be able to empower people with access to expert medical knowledge, I think is both inspiring and amazing.   And then, Seb, Sébastien Bubeck, he’s just absolutely a brilliant mind. He has a really firm grip on the deep mathematics of artificial intelligence and brings that to bear in his research and development work. And where that mathematics takes him isn’t just into the nuts and bolts of algorithms but into philosophical questions about the nature of intelligence.   One of the things that Sébastien brought up was the state of evaluation of AI systems. And indeed, he was fairly critical in our conversation. But of course, the world of AI research and development is just moving so fast, and indeed, since we recorded our conversation, OpenAI, in fact, released a new evaluation metric that is directly relevant to medical applications, and that is something called HealthBench. And Microsoft Research also released a new evaluation approach or process called ADeLe.   HealthBench and ADeLe are examples of new approaches to evaluating AI models that are less about testing their knowledge and ability to pass multiple-choice exams and instead are evaluation approaches designed to assess how well AI models are able to complete tasks that actually arise every day in typical healthcare or biomedical research settings. These are examples of really important good work that speak to how well AI models work in the real world of healthcare and biomedical research and how well they can collaborate with human beings in those settings.  You know, I asked Bill and Seb to make some predictions about the future. You know, my own answer, I expect that we’re going to be able to use AI to change how we diagnose patients, change how we decide treatment options.   If you’re a doctor or a nurse and you encounter a patient, you’ll ask questions, do a physical exam, you know, call out for labs just like you do today, but then you’ll be able to engage with AI based on all of that data and just ask, you know, based on all the other people who have gone through the same experience, who have similar data, how were they diagnosed? How were they treated? What were their outcomes? And what does that mean for the patient I have right now? Some people call it the “patients like me” paradigm. And I think that’s going to become real because of AI within our lifetimes. That idea of really grounding the delivery in healthcare and medical practice through data and intelligence, I actually now don’t see any barriers to that future becoming real.  [THEME MUSIC]  I’d like to extend another big thank you to Bill and Sébastien for their time. And to our listeners, as always, it’s a pleasure to have you along for the ride. I hope you’ll join us for our remaining conversations, as well as a second coauthor roundtable with Carey and Zak.   Until next time.   [MUSIC FADES]

DJI Osmo Action 4. Adrian Kingsley-Hughes/ZDNETMultiple DJI Osmo Action 4 packages are on sale at Amazon. Both the Essential and Standard Combos have been discounted to $249, while the Adventure Combo has dropped to $349.DJI might not be the first name on people's lips when it comes to action cameras, but the company that's better known for its drones also has a really solid line of action cameras. And its latest device, the Osmo Action 4 camera, has some very impressive tricks up its sleeve.Also: One of the most versatile cameras I've used is not from Sony or Canon and it's on saleSo, what sets this action camera apart from the competition? Let's take a look. details View at Amazon First off, this is not just an action camera -- it's a pro-grade action camera.From a hardware point of view, the Osmo Action 4 features a 1/1.3-inch image sensor that can record 4K at up to 120 frames per second (fps). This sensor is combined with a wide-angle f/2.8 aperture lens that provides an ultra-wide field of view of up to 155°. And that's wide. Build quality and fit and finish are second to none. Adrian Kingsley-Hughes/ZDNETFor when the going gets rough, the Osmo Action 4 offers 360° HorizonSteady stabilization modes, including RockSteady 3.0/3.0+ for first-person video footage and HorizonBalancing/HorizonSteady modes for horizontal shots. That's pro-grade hardware right there.Also: This new AI video editor is an all-in-one production service for filmmakers - how to try itThe Osmo Action 4 also features a 10-bit D-Log M color mode. This mode allows the sensor to record over one billion colors and offers a wider dynamic range, giving you a video that is more vivid and that offers greater detail in the highlights and shadows. This mode, combined with an advanced color temperature sensor, means that the colors have a true-to-life feel regardless of whether you're shooting outdoors, indoors, or even underwater. The DJI Osmo Action 4 ready for action. Adrian Kingsley-Hughes/ZDNETI've added some video output from the Osmo Action 4 below. There are examples in both 1080p and 4K. To test the stabilization, I attached the camera to the truck and took it on some roads, some of which are pretty rough. The Osmo Action 4 had no problem with that terrain. I also popped the camera into the sea, just because. And again, no problem.I've also captured a few time-lapses with the camera -- not because I like clouds (well, actually, I do like clouds), but pointing a camera at a sky can be a good test of how it handles changing light. Also: I recommend this action camera to beginners and professional creators. Here's whyTimelapses with action cameras can suffer from unsightly exposure changes that cause the image to pulse, a condition known as exposure pumping. This issue can also cause the white balance to change noticeably in a video, but the Osmo Action 4 handled this test well.All the footage I've shot is what I've come to expect from a DJI camera, whether it's from an action camera or drone -- crisp, clear, vivid, and also nice and stable.The Osmo Action 4 is packed with various electronic image-stabilization (EIS) tech to ensure that your footage is smooth and on the horizon. It's worth noting the limitations of EIS -- it's not supported in slow-motion and timelapse modes, and the HorizonSteady and HorizonBalancing features are only available for video recorded at 1080p (16:9) or 2.7K (16:9) with a frame rate of 60fps or below. On the durability front, I've no concerns. I've subjected the Osmo Action 4 to a hard few days of testing, and it's not let me down or complained once. It takes impacts like a champ, and being underwater or in dirt and sand is no problem at all. Also: I'm a full-time Canon photographer, but this Nikon camera made me wonder if I'm missing outYou might think that this heavy-duty testing would be hard on the camera's tiny batteries, but you'd be wrong. Remember I said the Osmo Action 4 offered hours of battery life? Well, I wasn't kidding.  The Osmo Action 4's ultra-long life batteries are incredible.  Adrian Kingsley-Hughes/ZDNETDJI says that a single battery can deliver up to 160 minutes of 1080p/24fps video recording (at room temperature, with RockSteady on, Wi-Fi off, and screen off). That's over two and a half hours of recording time. In the real world, I was blown away by how much a single battery can deliver. I shot video and timelapse, messed around with a load of camera settings, and then transferred that footage to my iPhone, and still had 16% battery left.No action camera has delivered so much for me on one battery. The two extra batteries and the multifunction case that come as part of the Adventure Combo are worth the extra $100. Adrian Kingsley-Hughes/ZDNETAnd when you're ready to recharge, a 30W USB-C charger can take a battery from zero to 80% in 18 minutes. That's also impressive.What's more, the batteries are resistant to cold, offering up to 150 minutes of 1080p/24fps recording in temperatures as low as -20°C (-4°F). This resistance also blows the competition away.Even taking into account all these strong points, the Osmo Action 4 offers even more.The camera has 2x digital zoom for better composition, Voice Prompts that let you know what the camera is doing without looking, and Voice Control that lets you operate the device without touching the screen or using the app. The Osmo Action 4 also digitally hides the selfie stick from a variety of different shots, and you can even connect the DJI Mic to the camera via the USB-C port for better audio capture.Also: Yes, an Android tablet finally made me reconsider my iPad Pro loyaltyAs for price, the Osmo Action 4 Standard Combo bundle comes in at $399, while the Osmo Action 4 Adventure Combo, which comes with two extra Osmo Action Extreme batteries, an additional mini Osmo Action quick-release adapter mount, a battery case that acts as a power bank, and a 1.5-meter selfie stick, is $499.I'm in love with the Osmo Action 4. It's hands down the best, most versatile, most powerful action camera on the market today, offering pro-grade features at a price that definitely isn't pro-grade.  Everything included in the Action Combo bundle. DJIDJI Osmo Action 4 tech specsDimensions: 70.5×44.2×32.8mmWeight: 145gWaterproof: 18m, up to 60m with the optional waterproof case Microphones: 3Sensor 1/1.3-inch CMOSLens: FOV 155°, aperture f/2.8, focus distance 0.4m to ∞Max Photo Resolution: 3648×2736Max Video Resolution: 4K (4:3): 3840×2880@24/25/30/48/50/60fps and 4K (16:9): 3840×2160@24/25/30/48/50/60/100/120fpsISO Range: 100-12800Front Screen: 1.4-inch, 323ppi, 320×320Rear Screen: 2.25-inch, 326ppi, 360×640Front/Rear Screen Brightness: 750±50 cd/m² Storage: microSD (up to 512GB)Battery: 1770mAh, lab tested to offer up to 160 minutes of runtime (tested at room temperature - 25°C/77°F - and 1080p/24fps, with RockSteady on, Wi-Fi off, and screen off)Operating Temperature: -20° to 45° C (-4° to 113° F)This article was originally published in August of 2023 and updated in March 2025.Featured reviews

The Inefficiency of Static Chain-of-Thought Reasoning in LRMs Recent LRMs achieve top performance by using detailed CoT reasoning to solve complex tasks. However, many simple tasks they handle could be solved by smaller models with fewer tokens, making such elaborate reasoning unnecessary. This echoes human thinking, where we use fast, intuitive responses for easy problems and slower, analytical thinking for complex ones. While LRMs mimic slow, logical reasoning, they generate significantly longer outputs, thereby increasing computational cost. Current methods for reducing reasoning steps lack flexibility, limiting models to a single fixed reasoning style. There is a growing need for adaptive reasoning that adjusts effort according to task difficulty.  Limitations of Existing Training-Based and Training-Free Approaches Recent research on improving reasoning efficiency in LRMs can be categorized into two main areas: training-based and training-free methods. Training strategies often use reinforcement learning or fine-tuning to limit token usage or adjust reasoning depth, but they tend to follow fixed patterns without flexibility. Training-free approaches utilize prompt engineering or pattern detection to shorten outputs during inference; however, they also lack adaptability. More recent work focuses on variable-length reasoning, where models adjust reasoning depth based on task complexity. Others study “overthinking,” where models over-reason unnecessarily. However, few methods enable dynamic switching between quick and thorough reasoning—something this paper addresses directly.  Introducing OThink-R1: Dynamic Fast/Slow Reasoning Framework Researchers from Zhejiang University and OPPO have developed OThink-R1, a new approach that enables LRMs to switch between fast and slow thinking smartly, much like humans do. By analyzing reasoning patterns, they identified which steps are essential and which are redundant. With help from another model acting as a judge, they trained LRMs to adapt their reasoning style based on task complexity. Their method reduces unnecessary reasoning by over 23% without losing accuracy. Using a loss function and fine-tuned datasets, OThink-R1 outperforms previous models in both efficiency and performance on various math and question-answering tasks.  System Architecture: Reasoning Pruning and Dual-Reference Optimization The OThink-R1 framework helps LRMs dynamically switch between fast and slow thinking. First, it identifies when LRMs include unnecessary reasoning, like overexplaining or double-checking, versus when detailed steps are truly essential. Using this, it builds a curated training dataset by pruning redundant reasoning and retaining valuable logic. Then, during fine-tuning, a special loss function balances both reasoning styles. This dual-reference loss compares the model’s outputs with both fast and slow thinking variants, encouraging flexibility. As a result, OThink-R1 can adaptively choose the most efficient reasoning path for each problem while preserving accuracy and logical depth.  Empirical Evaluation and Comparative Performance The OThink-R1 model was tested on simpler QA and math tasks to evaluate its ability to switch between fast and slow reasoning. Using datasets like OpenBookQA, CommonsenseQA, ASDIV, and GSM8K, the model demonstrated strong performance, generating fewer tokens while maintaining or improving accuracy. Compared to baselines such as NoThinking and DualFormer, OThink-R1 demonstrated a better balance between efficiency and effectiveness. Ablation studies confirmed the importance of pruning, KL constraints, and LLM-Judge in achieving optimal results. A case study illustrated that unnecessary reasoning can lead to overthinking and reduced accuracy, highlighting OThink-R1’s strength in adaptive reasoning.  Conclusion: Towards Scalable and Efficient Hybrid Reasoning Systems In conclusion, OThink-R1 is a large reasoning model that adaptively switches between fast and slow thinking modes to improve both efficiency and performance. It addresses the issue of unnecessarily complex reasoning in large models by analyzing and classifying reasoning steps as either essential or redundant. By pruning the redundant ones while maintaining logical accuracy, OThink-R1 reduces unnecessary computation. It also introduces a dual-reference KL-divergence loss to strengthen hybrid reasoning. Tested on math and QA tasks, it cuts down reasoning redundancy by 23% without sacrificing accuracy, showing promise for building more adaptive, scalable, and efficient AI reasoning systems in the future.  Check out the Paper and GitHub Page. All credit for this research goes to the researchers of this project. Also, feel free to follow us on Twitter and don’t forget to join our 100k+ ML SubReddit and Subscribe to our Newsletter. Sana HassanSana Hassan, a consulting intern at Marktechpost and dual-degree student at IIT Madras, is passionate about applying technology and AI to address real-world challenges. With a keen interest in solving practical problems, he brings a fresh perspective to the intersection of AI and real-life solutions.Sana Hassanhttps://www.marktechpost.com/author/sana-hassan/Building AI-Powered Applications Using the Plan → Files → Code Workflow in TinyDevSana Hassanhttps://www.marktechpost.com/author/sana-hassan/MemOS: A Memory-Centric Operating System for Evolving and Adaptive Large Language ModelsSana Hassanhttps://www.marktechpost.com/author/sana-hassan/Google AI Unveils a Hybrid AI-Physics Model for Accurate Regional Climate Risk Forecasts with Better Uncertainty AssessmentSana Hassanhttps://www.marktechpost.com/author/sana-hassan/Run Multiple AI Coding Agents in Parallel with Container-Use from Dagger

Join the event trusted by enterprise leaders for nearly two decades. VB Transform brings together the people building real enterprise AI strategy. Learn more When DeepSeek released its R1 model this January, it wasn’t just another AI announcement. It was a watershed moment that sent shockwaves through the tech industry, forcing industry leaders to reconsider their fundamental approaches to AI development. What makes DeepSeek’s accomplishment remarkable isn’t that the company developed novel capabilities; rather, it was how it achieved comparable results to those delivered by tech heavyweights at a fraction of the cost. In reality, DeepSeek didn’t do anything that hadn’t been done before; its innovation stemmed from pursuing different priorities. As a result, we are now experiencing rapid-fire development along two parallel tracks: efficiency and compute.  As DeepSeek prepares to release its R2 model, and as it concurrently faces the potential of even greater chip restrictions from the U.S., it’s important to look at how it captured so much attention. Engineering around constraints DeepSeek’s arrival, as sudden and dramatic as it was, captivated us all because it showcased the capacity for innovation to thrive even under significant constraints. Faced with U.S. export controls limiting access to cutting-edge AI chips, DeepSeek was forced to find alternative pathways to AI advancement. While U.S. companies pursued performance gains through more powerful hardware, bigger models and better data, DeepSeek focused on optimizing what was available. It implemented known ideas with remarkable execution — and there is novelty in executing what’s known and doing it well. This efficiency-first mindset yielded incredibly impressive results. DeepSeek’s R1 model reportedly matches OpenAI’s capabilities at just 5 to 10% of the operating cost. According to reports, the final training run for DeepSeek’s V3 predecessor cost a mere $6 million — which was described by former Tesla AI scientist Andrej Karpathy as “a joke of a budget” compared to the tens or hundreds of millions spent by U.S. competitors. More strikingly, while OpenAI reportedly spent $500 million training its recent “Orion” model, DeepSeek achieved superior benchmark results for just $5.6 million — less than 1.2% of OpenAI’s investment. If you get starry eyed believing these incredible results were achieved even as DeepSeek was at a severe disadvantage based on its inability to access advanced AI chips, I hate to tell you, but that narrative isn’t entirely accurate (even though it makes a good story). Initial U.S. export controls focused primarily on compute capabilities, not on memory and networking — two crucial components for AI development. That means that the chips DeepSeek had access to were not poor quality chips; their networking and memory capabilities allowed DeepSeek to parallelize operations across many units, a key strategy for running their large model efficiently. This, combined with China’s national push toward controlling the entire vertical stack of AI infrastructure, resulted in accelerated innovation that many Western observers didn’t anticipate. DeepSeek’s advancements were an inevitable part of AI development, but they brought known advancements forward a few years earlier than would have been possible otherwise, and that’s pretty amazing. Pragmatism over process Beyond hardware optimization, DeepSeek’s approach to training data represents another departure from conventional Western practices. Rather than relying solely on web-scraped content, DeepSeek reportedly leveraged significant amounts of synthetic data and outputs from other proprietary models. This is a classic example of model distillation, or the ability to learn from really powerful models. Such an approach, however, raises questions about data privacy and governance that might concern Western enterprise customers. Still, it underscores DeepSeek’s overall pragmatic focus on results over process. The effective use of synthetic data is a key differentiator. Synthetic data can be very effective when it comes to training large models, but you have to be careful; some model architectures handle synthetic data better than others. For instance, transformer-based models with mixture of experts (MoE) architectures like DeepSeek’s tend to be more robust when incorporating synthetic data, while more traditional dense architectures like those used in early Llama models can experience performance degradation or even “model collapse” when trained on too much synthetic content. This architectural sensitivity matters because synthetic data introduces different patterns and distributions compared to real-world data. When a model architecture doesn’t handle synthetic data well, it may learn shortcuts or biases present in the synthetic data generation process rather than generalizable knowledge. This can lead to reduced performance on real-world tasks, increased hallucinations or brittleness when facing novel situations.  Still, DeepSeek’s engineering teams reportedly designed their model architecture specifically with synthetic data integration in mind from the earliest planning stages. This allowed the company to leverage the cost benefits of synthetic data without sacrificing performance. Market reverberations Why does all of this matter? Stock market aside, DeepSeek’s emergence has triggered substantive strategic shifts among industry leaders. Case in point: OpenAI. Sam Altman recently announced plans to release the company’s first “open-weight” language model since 2019. This is a pretty notable pivot for a company that built its business on proprietary systems. It seems DeepSeek’s rise, on top of Llama’s success, has hit OpenAI’s leader hard. Just a month after DeepSeek arrived on the scene, Altman admitted that OpenAI had been “on the wrong side of history” regarding open-source AI.  With OpenAI reportedly spending $7 to 8 billion annually on operations, the economic pressure from efficient alternatives like DeepSeek has become impossible to ignore. As AI scholar Kai-Fu Lee bluntly put it: “You’re spending $7 billion or $8 billion a year, making a massive loss, and here you have a competitor coming in with an open-source model that’s for free.” This necessitates change. This economic reality prompted OpenAI to pursue a massive $40 billion funding round that valued the company at an unprecedented $300 billion. But even with a war chest of funds at its disposal, the fundamental challenge remains: OpenAI’s approach is dramatically more resource-intensive than DeepSeek’s. Beyond model training Another significant trend accelerated by DeepSeek is the shift toward “test-time compute” (TTC). As major AI labs have now trained their models on much of the available public data on the internet, data scarcity is slowing further improvements in pre-training. To get around this, DeepSeek announced a collaboration with Tsinghua University to enable “self-principled critique tuning” (SPCT). This approach trains AI to develop its own rules for judging content and then uses those rules to provide detailed critiques. The system includes a built-in “judge” that evaluates the AI’s answers in real-time, comparing responses against core rules and quality standards. The development is part of a movement towards autonomous self-evaluation and improvement in AI systems in which models use inference time to improve results, rather than simply making models larger during training. DeepSeek calls its system “DeepSeek-GRM” (generalist reward modeling). But, as with its model distillation approach, this could be considered a mix of promise and risk. For example, if the AI develops its own judging criteria, there’s a risk those principles diverge from human values, ethics or context. The rules could end up being overly rigid or biased, optimizing for style over substance, and/or reinforce incorrect assumptions or hallucinations. Additionally, without a human in the loop, issues could arise if the “judge” is flawed or misaligned. It’s a kind of AI talking to itself, without robust external grounding. On top of this, users and developers may not understand why the AI reached a certain conclusion — which feeds into a bigger concern: Should an AI be allowed to decide what is “good” or “correct” based solely on its own logic? These risks shouldn’t be discounted. At the same time, this approach is gaining traction, as again DeepSeek builds on the body of work of others (think OpenAI’s “critique and revise” methods, Anthropic’s constitutional AI or research on self-rewarding agents) to create what is likely the first full-stack application of SPCT in a commercial effort. This could mark a powerful shift in AI autonomy, but there still is a need for rigorous auditing, transparency and safeguards. It’s not just about models getting smarter, but that they remain aligned, interpretable, and trustworthy as they begin critiquing themselves without human guardrails. Moving into the future So, taking all of this into account, the rise of DeepSeek signals a broader shift in the AI industry toward parallel innovation tracks. While companies continue building more powerful compute clusters for next-generation capabilities, there will also be intense focus on finding efficiency gains through software engineering and model architecture improvements to offset the challenges of AI energy consumption, which far outpaces power generation capacity.  Companies are taking note. Microsoft, for example, has halted data center development in multiple regions globally, recalibrating toward a more distributed, efficient infrastructure approach. While still planning to invest approximately $80 billion in AI infrastructure this fiscal year, the company is reallocating resources in response to the efficiency gains DeepSeek introduced to the market. Meta has also responded, With so much movement in such a short time, it becomes somewhat ironic that the U.S. sanctions designed to maintain American AI dominance may have instead accelerated the very innovation they sought to contain. By constraining access to materials, DeepSeek was forced to blaze a new trail. Moving forward, as the industry continues to evolve globally, adaptability for all players will be key. Policies, people and market reactions will continue to shift the ground rules — whether it’s eliminating the AI diffusion rule, a new ban on technology purchases or something else entirely. It’s what we learn from one another and how we respond that will be worth watching. Jae Lee is CEO and co-founder of TwelveLabs. Daily insights on business use cases with VB Daily If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI. Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

Things are warming up as Remedy’s FBC: Firebreak approaches its June 17 launch on PlayStation 5 as part of the PlayStation Plus Game Catalog. We chatted with Communications Director Thomas Puha, Lead Level Designer Teemu Huhtiniemi, Lead Designer/Lead Technical Designer Anssi Hyytiainen, and Game Director/Lead Writer Mike Kayatta about some of the fascinating and often hilarious development secrets behind the first-person shooter. PlayStation Blog: First, what PS5 and PS5 Pro features did you utilize? Thomas Puha: We’ll support 3D Audio, and we’re prioritising 60 FPS on both formats. We’re aiming for FSR2 with an output resolution of 2560 x 1440 (1440p) on PS, and PSSR with an output resolution of 3840×2160 (4K) on PS5 Pro. Some of the DualSense wireless controller’s features are still a work in progress, but we’re looking to use haptic feedback in a similar way to our previous titles, such as Control and Alan Wake 2. For example, we want to differentiate the weapons to feel unique from each other using the adaptive triggers. Going into the game itself, were there any other influences on its creation outside of Control? Mike Kayatta: We looked at different TV shows that had lots of tools for going into a place and dealing with a crisis. One was a reality show called Dirty Jobs, where the host Mike Rowe finds these terrible, dangerous, or unexpected jobs that you don’t know exist, like cleaning out the inside of a water tower. We also looked at PowerWash Simulator. Cleaning dirt is oddly meditative and really fulfilling. It made me wish a zombie attacked me to break the Zen, and then I’d go right back to cleaning. And we were like, that would be pretty fun in the game. Play Video Were there specific challenges you faced given it’s your first multiplayer game and first-person shooter? Anssi Hyytiainen: It’s radically different from a workflow point of view. You can’t really test it alone, necessarily, which is quite a different experience. And then there are times when one player is missing things on their screen that others are seeing. It was like, “What are you shooting at?” What’s been your favorite moments developing the game so far? Teemu Huhtiniemi: There were so many. But I like when we started seeing all of these overlapping systems kind of click, because there’s a long time in the development where you talk about things on paper and have some prototypes, but you don’t really see it all come together until a point. Then you start seeing the interaction between the systems and all the fun that comes out of that. Kayatta: I imagine there’s a lot of people who probably are a little skeptical about Remedy making something so different. Even internally, when the project was starting. And once we got the trailer out there, everyone was so nervous, but it got a pretty positive reaction. Exposing it to the public is very motivating, because with games, for a very long time, there is nothing, or it is janky and it’s ugly and you don’t find the fun immediately. Were there any specific ideals you followed while you worked on the game? Kayatta: Early on we were constantly asking ourselves, “Could this only happen in Control or at Remedy?” Because the first thing you hear is, “Okay, this is just another co-op multiplayer shooter” – there’s thousands of them, and they’re all good. So what can we do to make it worth playing our game? We were always saying we’ve got this super weird universe and really interesting studio, so we’re always looking at what we could do that nobody else can. Huhtiniemi: I think for me it was when we chose to just embrace the chaos. Like, that’s the whole point of the game. It’s supposed to feel overwhelming and busy at times, so that was great to say it out loud. Kayatta: Yeah, originally we had a prototype where there were only two Hiss in the level, but it just didn’t work, it wasn’t fun. Then everything just accidentally went in the opposite direction, where it was super chaos. At some point we actually started looking at Overcooked quite a bit, and saying, “Look, just embrace it. It’s gonna be nuts.” How did you finally decide on the name FBC: Firebreak, and were there any rejected, alternate, or working titles? Kayatta: So Firebreak is named after real world firebreaks, where you deforest an area to prevent a fire from spreading, but firebreaks are also topographical features of the Oldest House. And so we leaned into the term being a first responder who stops fires from spreading. The FBC part came from not wanting to put ‘Control’ in the title, so Control players wouldn’t feel like they had to detour to this before Control 2, but we didn’t want to totally detach from it either as that felt insincere. An external partner pitched a title. They were very serious about talking up the game being in the Oldest House, and then dramatically revealed the name: Housekeepers. I got what they were going for, but I was like, we cannot call it this. It was like you were playing as a maid!   FBC: Firebreak launches on PS5 June 17 as a day on PlayStation Plus Game Catalog title.