Latest news with #NoamBrown
India Today
22-07-2025
- Business
- India Today
OpenAI and Google both won gold at 2025 International Math Olympiad: Full story in 5 points
In a first for artificial intelligence, OpenAI and Google have announced that their AI models have scored gold medal-worthy results at the 2025 International Mathematical Olympiad (IMO), a prestigious global competition for high school students. The development is being seen as a landmark moment in the race to build AI systems that can reason like humans and solve complex academic Google won gold at 2025 International Math Olympiad: Full story in 5 points-OpenAI and Google used advanced reasoning-based AI models that worked through natural language rather than relying on traditional mathematical programming methods. Both companies' systems successfully solved five out of six problems, a score that crosses the threshold for a gold medal at the IMO. This is the first time any AI models have managed to reach that level of accuracy in the competition's history.-The 66th IMO was held on Australia's Sunshine Coast, with 630 student participants. Alongside them, Google's DeepMind AI unit officially took part with its "Gemini Deep Think" model, which had been introduced earlier at the company's I/O event in May. The model managed to work through all problems in the same 4.5-hour time frame given to human participants, using plain English to process and solve the questions. -On the other hand, OpenAI did not officially enter the contest but later shared that its own experimental model had achieved similar gold-level scores when given the same problems. OpenAI's scores were verified by three independent IMO medalists, according to the company. The model used a new method involving massively scaled-up "test-time compute", which essentially means the system was allowed to run longer and use greater computing power to think through multiple approaches in parallel. OpenAI researcher Noam Brown described the effort as computationally 'very expensive'.-While Google's DeepMind had its results verified and certified by the IMO's committee, OpenAI revealed its achievement after the official competition results were made public. Both companies respected the IMO board's condition to delay announcements until the student rankings had been confirmed.-The achievement has sparked optimism among researchers. Professor Junehyuk Jung from Brown University — himself a former IMO gold medalist — said that this progress shows how close AI is to playing a supporting role in solving high-level research problems in mathematics. According to Google, this breakthrough is not just about solving maths problems. It's about demonstrating that AI systems are now capable of applying logic and reasoning, not just in maths but potentially in fields like physics and theoretical computer science. While OpenAI confirmed it won't release such high-level mathematical tools to the public immediately, it hinted that the capabilities could soon extend beyond math.- Ends
TimesLIVE
22-07-2025
- Business
- TimesLIVE
AI systems from Google and OpenAI soar at global maths competition
OpenAI's breakthrough was achieved with a new experimental model centered on massively scaling up "test-time compute". This was done by allowing the model to "think" for longer periods and deploying parallel computing power to run many lines of reasoning simultaneously, according to Noam Brown, researcher at OpenAI. Brown declined to say how much computing power it cost OpenAI, but called it "very expensive". To OpenAI researchers, it is another clear sign AI models can command extensive reasoning capabilities that could expand into areas beyond maths. The optimism is shared by Google researchers, who believe AI models' capabilities can apply to research quandaries in other fields such as physics, said Jung, who won an IMO gold medal as a student in 2003. Of the 630 students participating in the 66th IMO on the Sunshine Coast in Queensland, Australia, 67 contestants, or about 11%, achieved gold medal scores. Google's DeepMind AI unit last year achieved a silver medal score using AI systems specialised for maths. This year, Google used a general-purpose model called Gemini Deep Think, a version of which was previously unveiled at its annual developer conference in May. Unlike previous AI attempts that relied on formal languages and lengthy computation, Google's approach this year operated entirely in natural language and solved the problems within the official 4.5-hour time limit, the company said in a blog post. OpenAI, which has its own set of reasoning models, similarly built an experimental version for the competition, according to a post by researcher Alexander Wei on social media platform X. He noted the company does not plan to release anything with this level of maths capability for several months. This year marked the first time the competition coordinated officially with some AI developers, who have for years used prominent maths competitions such as IMO to test model capabilities. IMO judges certified the results of the companies, including Google, and asked them to publish results on July 28. "We respected the IMO board's original request that all AI labs share their results only after the official results had been verified by independent experts and the students had rightly received the acclamation they deserved," Google DeepMind CEO Demis Hassabis said on X on Monday. OpenAI, which published its results on Saturday and first claimed gold medal status, said in an interview it had permission from an IMO board member to do so after the closing ceremony on Saturday. The competition on Monday allowed cooperating companies to publish results, said Gregor Dolinar, president of IMO's board.
India Today
21-07-2025
- Science
- India Today
OpenAI won gold at the world's toughest math exam. Why the Olympiad gold matters
In a jaw-dropping achievement for the world of artificial intelligence, OpenAI's latest experimental model has scored at the gold medal level at the International Mathematical Olympiad (IMO) -- one of the toughest math exams on the is the same event held on the Sunshine Coast in Australia where India won six medals this year and ranked 7th amongst 110 participating HITS GOLD IN THE WORLD'S TOUGHEST MATH TESTThe IMO is no ordinary competition. Since its launch in 1959 in Romania, it has become the gold standard for testing mathematical genius among high school students globally. Over two intense days, participants face a gruelling four-and-a-half-hour paper with only three questions each day. These are not your average exam questions -- they demand deep logic, creativity and problem-solving that, OpenAI's model solved five out of six questions correctly -- under the same testing conditions as human DOUBTED AI COULD DO THIS -- UNTIL NOWEven renowned mathematician Terence Tao -- an IMO gold medallist himself -- had doubts. In a podcast in June, he suggested that AI wasn't yet ready for the IMO level and should try simpler math contests first. But OpenAI has now proven otherwise."Also this model thinks for a *long* time. o1 thought for seconds. Deep Research for minutes. This one thinks for hours. Importantly, it's also more efficient with its thinking," Noam Brown from OpenAI wrote on LinkedIn."It's worth reflecting on just how fast AI progress has been, especially in math. In 2024, AI labs were using grade school math (GSM8K) as an eval in their model releases. Since then, we've saturated the (high school) MATH benchmark, then AIME, and now are at IMO gold," he THIS IS A BIG DEAL FOR GENERAL AIThis isn't just about math. OpenAI says this shows their AI model is breaking new ground in general-purpose reasoning. Unlike Google DeepMind's AlphaGeometry -- built just for geometry -- OpenAI's model is a general large language model that happens to be great at math too."Typically for these AI results, like in Go/Dota/Poker/Diplomacy, researchers spend years making an AI that masters one narrow domain and does little else. But this isn't an IMO-specific model. It's a reasoning LLM that incorporates new experimental general-purpose techniques," Brown explained in his Sam Altman called it 'a dream' when OpenAI began. 'This is a marker of how far AI has come in a decade.'advertisementBut before you get your hopes up, this high-performing AI isn't going public just yet. Altman confirmed it'll be 'many months' before this gold-level model is STILL REMAINNot everyone is fully convinced. AI expert Gary Marcus called the model's results 'genuinely impressive' -- but raised fair questions about training methods, how useful this is for the average person, and how much it all the win marks a huge leap in what artificial intelligence can do -- and how fast it's improving.- EndsMust Watch
India Today
11-06-2025
- Business
- India Today
OpenAI adds o3 Pro to ChatGPT and drops o3 price by 80 per cent, but open-source AI is delayed
OpenAI has been busy. This morning, we woke up to a bunch of updates from the AI company. To begin with, OpenAI has released o3 Pro, its most advanced reasoning model to date, for Pro and Team users of ChatGPT. The release of the new model also comes with a dramatic 80 per cent price cut for the standard o3 model, which makes it more accessible for developers and researchers around the world. However, OpenAI CEO Sam Altman has also announced that the company has had to push back the release of its open-source model until later this Pro lands in ChatGPTOpenAI has launched a new o3 Pro model, which it says is its most capable AI model yet. o3 Pro is available to Pro and Team users in ChatGPT and through OpenAI's API. It replaces the earlier o1 Pro model and is designed to deliver even more reliable and accurate answers across complex domains like science, programming, mathematics, and education. OpenAI described o3 Pro as a 'version of our most intelligent model, o3, designed to think longer and provide the most reliable responses.' The company notes that in internal and external academic evaluation, o3 Pro has consistently outperformed both o3 and its predecessors, especially in clarity, accuracy, instruction-following, and comprehensiveness. advertisement Although o3 Pro uses the same underlying model as o3, it is optimised for dependability. OpenAI says reviewers have ranked the model higher across key tasks in science and data analysis. As OpenAI puts it, 'We recommend using it for challenging questions where reliability matters more than speed, and waiting a few minutes is worth the tradeoff.' Notably, o3 Pro is also equipped with tools that enhance its capabilities, including the ability to browse the web, analyse documents, interpret visual inputs, run Python code, and use memory for more personalised interactions. However, these advanced tools mean its responses typically take longer to generate than those of o1 the rollout, a few features remain unavailable in o3 Pro for now. Temporary chats are currently disabled due to a technical issue, and the model cannot yet generate images or use the Canvas interface. Users looking for image generation are advised to stick with GPT-4o, o3, or and Edu customers will gain access to o3 Pro the following week. A massive price cut for o3Alongside the launch of o3 Pro, OpenAI has also slashed the cost of using the o3 model by 80 per cent – a move that makes the model more accessible and affordable to developers. The cost of using o3 now stands at $2 per million input tokens and $8 per million output tokens, with additional discounts for cached (previously seen) prompts. This is a major reduction from the previous rates of $10 and $40 respectively. As OpenAI researcher Noam Brown pointed out on X, this update represents a significant shift in affordability for CEO Sam Altman confirmed the price drop on X, posting, 'we dropped the price of o3 by 80%!! excited to see what people will do with it now. think you'll also be happy with o3-pro pricing for the performance :)' advertisementThe new pricing places OpenAI in sharper competition with rivals like Google DeepMind's Gemini 2.5 Pro, which charges between $1.25 and $2.50 for input and up to $15 for output tokens, depending on prompt size. Gemini also offers Google Search integration, but with extra fees beyond a certain usage Claude Opus 4 remains the priciest competitor, charging $15 per million input tokens and $75 for output. DeepSeek's Reasoner and Chat models, on the other hand, lead the low-cost tier with rates as low as $0.035 for cached inputs during off-peak model postponedWhile the launch of o3 Pro and the price drop for the o3 model are widely welcomed, OpenAI's open-source model has hit a delay. Originally expected in June 2025, the model will now arrive later, following a surprise development by the research team. The exact timeline for the release has not yet been disclosed. In a post on X, Altman explained, 'We are going to take a little more time with our open-weights model, i.e. expect it later this summer but not June.' He added that the delay is due to the team achieving something 'unexpected and quite amazing,' which still needs more time to open-source model is expected to rival high-performing open reasoning models such as DeepSeek R1, with OpenAI aiming to raise the bar for open-access large language models. OpenAI CEO reveals cost of a single ChatGPT queryAlongside these product updates, Altman has also addressed questions about the environmental cost of using ChatGPT. In a blog post published on Tuesday, he shared that each ChatGPT query consumes about 0.34 watt-hours, which is roughly what an oven uses in a second or a high-efficiency lightbulb in a couple of terms of water usage, Altman revealed that the average query consumes 0.000085 gallons of water, or 'roughly one fifteenth of a teaspoon.' These figures, he explained, are part of a broader vision where 'the cost of intelligence should eventually converge to near the cost of electricity.'
Economic Times
22-05-2025
- Business
- Economic Times
Techies quickly brush up on AI, here's the whopping amount OpenAI, Google and xAI are shelling out and it's in millions
Since ChatGPT launched in late 2022, tech companies are fighting hard to hire top AI researchers, it's like a battle to get superstars, as per reports. Hiring in AI labs is like playing chess, each person is a strategic move, and companies want people with the right skills, says Ariel Herbert-Voss, ex-OpenAI researcher and CEO of RunSybil. These top people are called 'ICs', and their work can decide a company's future. When Noam Brown, a top OpenAI researcher, was job hunting in 2023, he got lunch with Google founder Sergey Brin, poker game at OpenAI CEO Sam Altman's house, and a private jet visit from a rich investor, as mentioned in the report by Reuters. Even Musk, the Tesla boss and owner of xAI, is mesmerized by the researchers. So much so that he himself pitches his company and the job offer to these tech geeks. Noam Brown finally chose to stay at OpenAI because they supported his work, not just with money, but also people and computing power. He said it wasn't even the best financial offer, but doing meaningful work mattered more to him. Still, money is flowing big time, some top researchers at OpenAI were offered $2 million bonuses and $20 million in equity to stop them from leaving for SSI, a company by OpenAI's ex-chief scientist Ilya Sutskever, says Reuters report. Full bonuses run in abundance among these rare breeds of researchers, with most of them required to stay just for a year to lap up the bonus paychecks. OpenAI researchers who got offers from Eleven Labs were given $1 million plus to stay. Many OpenAI researchers get over $10 million per year in total compensation, says Google DeepMind, researchers are being offered $20 million/year packages, extra equity grants outside normal cycles, and a stock vesting reduced to 3 years instead of 4, as per reports. In comparison, regular top tech engineers make around $280K in salary and another $260K in stock each year, way less than what star AI researchers are getting. In the AI world, there are only a few dozen to a thousand top AI researchers, very rare and super valuable, as mentioned in the report by Reuters. These rare experts are believed to be behind major progress in large language models. Open AI supremo Sam Altman, in a post on his now X but then Twitter handle, in 2023, said that 10x engineers are cool, but 10,000x researchers are even Murati, who used to be the chief tech officer at OpenAI, left the company in September and started her own AI company.* She was known for her leadership and ability to get things done at OpenAI. By February, she had already hired 20 people from OpenAI, and now her team is around 60 people strong. Even though her company doesn't have a product yet, it's about to raise a record-breaking seed funding round, mostly because of how strong her team is, as per the report by Reuters. Because good AI talent is so rare, companies are getting creative. A company called Zeki Data is using sports-style scouting methods like the 'Moneyball' movie to find under-the-radar AI talent. Other AI companies are hiring people with quantum computing Bubeck, who left Microsoft to join OpenAI, said AI is now attracting top talent from all fields, especially math. He told Reuters that some of these new people are super smart and really making a differenceQ1. Why are tech companies paying so much to AI researchers?Because great AI researchers are rare and can help companies build better AI faster.Q2. How is hiring in AI different from other tech jobs?In AI, hiring is super competitive and focused on just a few top experts who can make a big impact.
