Latest news with #o3
Economic Times
02-07-2025
- Business
- Economic Times
When AI goes rogue, even exorcists might flinch
Ghouls in the machine As GenAI use grows, foundation models are advancing rapidly, driven by fierce competition among top developers like OpenAI, Google, Meta and Anthropic. Each is vying for a reputational edge and business advantage in the race to lead development. This gives them a reputational edge, along with levers to further grow their business faster than their models powering GenAI are making significant strides. The most advanced - OpenAI's o3 and Anthropic's Claude Opus 4 - excel at complex tasks such as advanced coding and complex writing tasks, and can contribute to research projects and generate the codebase for a new software prototype with just a few considered prompts. These models use chain-of-thought (CoT) reasoning, breaking problems into smaller, manageable parts to 'reason' their way to an optimal solution. When you use models like o3 and Claude Opus 4 to generate solutions via ChatGPT or similar GenAI chatbots, you see such problem breakdowns in action, as the foundation model reports interactively the outcome of each step it has taken and what it will do next. That's the theory, anyway. While CoT reasoning boosts AI sophistication, these models lack the innate human ability to judge whether their outputs are rational, safe or ethical. Unlike humans, they don't subconsciously assess appropriateness of their next steps. As these advanced models step their way toward a solution, some have been observed to take unexpected and even defiant actions. In late May, AI safety firm Palisade Research reported on X that OpenAI's o3 model sabotaged a shutdown mechanism - even when explicitly instructed to 'allow yourself to be shut down'. An April 2025 paper by Anthropic, 'Reasoning Models Don't Always Say What They Think', shows that Opus 4 and similar models can't always be relied upon to faithfully report on their chains of reason. This undermines confidence in using such reports to validate whether the AI is acting correctly or safely. A June 2025 paper by Apple, 'The Illusion of Thinking', questions whether CoT methodologies truly enable reasoning. Through experiments, it exposed some of these models' limitations and situations where they 'experience complete collapse'.The fact that research critical of foundation models is being published after release of these models indicates the latter's relative immaturity. Under intense pressure to lead in GenAI, companies like Anthropic and OpenAI are releasing these models at a point where at least some of their fallibilities are not fully line was first crossed in late 2022, when OpenAI released ChatGPT, shattering public perceptions of AI and transforming the broader AI market. Until then, Big Tech had been developing LLMs and other GenAI tools, but were hesitant to release them, wary of unpredictable and uncontrollable argue for a greater degree of control over the ways in which these models are released - seeking to ensure standardisation of model testing and publication of the outcomes of this testing alongside the model's release. However, the current climate prioritises time to market over such development does this mean for industry, for those companies seeking to gain benefit from GenAI? This is an incredibly powerful and useful tech that is making significant changes to our ways of working and, over the next five years or so, will likely transform many I am continually wowed as I use these advanced foundation models in work and research - but not in my writing! - I always use them with a healthy dose of scepticism. Let's not trust them to always be correct and to not be subversive. It's best to work with them accordingly, making modifications to both prompts and codebases, other language content and visuals generated by the AI in a bid to ensure correctness. Even so, while maintaining discipline to understand the ML concepts one is working with, one wouldn't want to be without GenAI these these principles at scale, advice to large businesses on how AI can be governed and controlled: a risk-management approach - capturing, understanding and mitigating risks associated with AI use - helps organisations benefit from AI, while minimising chances of it going methods include guard rails in a variety of forms, evaluation-controlled release of AI services, and including a human-in-the-loop. Technologies that underpin these guard rails and evaluation methods need to keep up with model innovations such as CoT reasoning. This is a challenge that will continually be faced as AI is further developed. It's a good example of new job roles and technology services being created within industry as AI use becomes more prevalent. Such governance and AI controls are increasingly becoming a board imperative, given the current drive at an executive level to transform business using AI. Risk from most AI is low. But it is important to assess and understand this. Higher-risk AI can still, at times, be worth pursuing. With appropriate AI governance, this AI can be controlled, solutions innovated and benefits achieved. As we move into an increasingly AI-driven world, businesses that gain the most from AI will be those that are aware of its fallibilities as well as its huge potential, and those that innovate, build and transform with AI accordingly. (Disclaimer: The opinions expressed in this column are that of the writer. The facts and opinions expressed here do not reflect the views of Elevate your knowledge and leadership skills at a cost cheaper than your daily tea. Delhivery survived the Meesho curveball. Can it keep on delivering profits? Why the RBI's stability report must go beyond rituals and routines Ozempic, Wegovy, Mounjaro: Are GLP-1 drugs weight loss wonders or health gamble? 3 critical hurdles in India's quest for rare earth independence Stock Radar: Apollo Hospitals breaks out from 2-month consolidation range; what should investors do – check target & stop loss Add qualitative & quantitative checks for wealth creation. 7 small-cap stocks from different sectors with upside potential of over 25% These 7 banking stocks can give more than 20% returns in 1 year, according to analysts Wealth creation is about holding the right stocks and ignoring the noise. 13 'right stocks' with an upside potential of up to 34%
Time of India
26-06-2025
- Business
- Time of India
Meet Trapit Bansal, Meta's new AI superintelligence team hire - Is Meta poaching top talent from OpenAI?
Meta has poached Trapit Bansal, a key AI researcher from OpenAI who contributed significantly to their early AI reasoning and reinforcement learning efforts. Bansal's move to Meta's new AI superintelligence team underscores the intense competition for AI talent. This team aims to develop next-generation AI reasoning models, rivaling those of OpenAI and Google. Tired of too many ads? Remove Ads Ex-OpenAI Researcher Trapit Bansal Joins Meta A Key Figure in OpenAI's Reasoning Work Tired of too many ads? Remove Ads Joining a Powerhouse Team at Meta Mark Zuckerberg's AI Hiring Spree Tired of too many ads? Remove Ads FAQs Meta has made another bold move in the AI talent wars by hiring Trapit Bansal , an AI researcher who played a pivotal role in shaping OpenAI 's early efforts in AI reasoning and reinforcement learning, according to a report by who joined OpenAI in 2022, is now among the most publicly visible names to leave the firm and join Meta's brand-new AI superintelligence team , an initiative that's fast attracting all the top minds in the field of AI, as per the spokesperson Kayla Wood confirmed the news to TechCrunch that Bansal had departed OpenAI, while even Bansal's LinkedIn page mentions that he has left OpenAI in June this year, according to the TechCrunch READ: After Canada, now US: College graduates face the toughest job market in decades – what's gone wrong? During his time at OpenAI, Bansal worked closely with co-founder Ilya Sutskever and played an instrumental role in the development of the company's foundational AI reasoning model, o1, as reported by growing interest in AI reasoning models , especially as Meta's competitors like OpenAI's o3 and DeepSeek's R1 hit new performance milestones, makes Bansal's move even more impactful, according to the READ: Karoline Leavitt says no enriched uranium was removed from Iranian nuclear sites prior to US attacks Bansal brings his expertise to an impressive team at Meta's AI superintelligence lab, which includes former Scale AI CEO Alexandr Wang, ex-Google DeepMind researcher Jack Rae, and machine learning veteran Johan Schalkwyk, as per the report. Bloomberg and The Wall Street Journal reported that several other former OpenAI researchers, Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai, have recently joined Meta as mission of the lab is to develop next-gen AI reasoning models that may match or exceed OpenAI and Google, however, Meta has not yet put out a public AI reasoning model, as reported by READ: Last chance to claim your Fortnite refund – Act fast or risk missing out on free cash Meta CEO Mark Zuckerberg has also been making compensation deals in the $100 million range to lure top AI talent to build his new AI team, as reported by TechCrunch. However, it is not known what Bansal was offered to join in this deal, as reported by has also reportedly tried to acquire startups with heavy-hitting AI research labs, like Sutskever's Safe Superintelligence, Mira Murati's Thinking Machines Labs, and Perplexity, to further fill out its new AI unit, but those talks never progressed to a final stage, according to the a recent podcast, OpenAI CEO Sam Altman asserted that Meta has been trying to poach his startup's top talent, but highlighted that 'none of our best people have decided to take him up on that,' quoted an AI researcher who helped OpenAI develop its first major reasoning model and worked closely with Ilya a newly formed unit aimed at developing advanced AI reasoning models, similar to those at OpenAI and Google.
&w=3840&q=100)
Business Standard
26-06-2025
- Business
- Business Standard
Meta recruits leading OpenAI researcher Trapit Bansal for AI reasoning lab
Meta has onboarded a prominent OpenAI researcher, Trapit Bansal, to work on advanced aritificial intelligence (AI) reasoning models within its recently established AI superintelligence team, according to a report by TechCrunch. Trapit Bansal had been with OpenAI since 2022 and played a major role in launching the company's reinforcement learning research, working closely with co-founder Ilya Sutskever. He is named as one of the original contributors to OpenAI's first AI reasoning model, known as o1. His LinkedIn profile indicates that he left OpenAI in June. OpenAI spokesperson Kayla Wood confirmed to TechCrunch that Bansal had indeed exited the organisation. Boost to Meta's AI superintelligence team Bansal is expected to significantly strengthen Meta's new AI superintelligence group, which already includes notable figures such as former Scale AI CEO Alexandr Wang. The team is also in discussions to bring in former GitHub CEO Nat Friedman and Safe Superintelligence co-founder Daniel Gross. His expertise could help Meta develop a cutting-edge AI reasoning model to compete with leading offerings like OpenAI's o3 and DeepSeek's R1. At present, Meta does not have a publicly available AI reasoning model. Zuckerberg's high-profile hiring strategy In recent months, Meta CEO Mark Zuckerberg has aggressively recruited top AI talent, reportedly offering compensation packages as high as $100 million. While Bansal's offer remains undisclosed, his decision to join indicates the success of Zuckerberg's strategy in attracting leading AI researchers. According to The Wall Street Journal, Bansal will join other recent hires from OpenAI — Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai — at Meta. The team also includes Jack Rae, formerly of Google DeepMind, and Johan Schalkwyk, previously with startup Sesame, according to a Bloomberg report. Attempts to acquire AI startups fell through In a bid to expand its AI capabilities further, Meta also explored acquiring startups known for their AI research, including Safe Superintelligence (co-founded by Sutskever), Mira Murati's Thinking Machines Labs, and Perplexity. However, none of these talks reached a final agreement. On a recent podcast, OpenAI CEO Sam Altman commented on Meta's recruitment attempts, stating, 'None of our best people have decided to take him up on that.' AI reasoning a critical focus for Meta Developing powerful AI reasoning models is essential for Meta's new unit. Over the past year, firms such as OpenAI, Google, and DeepSeek have released high-performing models that can tackle complex tasks by reasoning through problems before producing answers. This approach, which makes use of additional computation time and resources, has led to improved performance both in benchmarks and in real-world applications. Future ambitions for Meta's AI lab Meta's AI superintelligence group is expected to become a crucial part of its wider operations, similar to the role DeepMind plays within Google. The company has plans to develop AI agents for enterprise use, led by Clara Shih, the former Salesforce CEO of AI.
The Hindu
11-06-2025
- Business
- The Hindu
France's Mistral unveils its first 'reasoning' AI model
French artificial intelligence startup Mistral on Tuesday announced a so-called "reasoning" model it said was capable of working through complex problems, following in the footsteps of top US developers. Available immediately on the company's platforms as well as the AI platform Hugging Face, the Magistral "is designed to think things through - in ways familiar to us," Mistral said in a blog post. The AI was designed for "general purpose use requiring longer thought processing and better accuracy" than its previous generations of large language models (LLMs), the company added. Like other "reasoning" models, Magistral displays a so-called "chain of thought" that purports to show how the system is approaching a problem given to it in natural language. This means users in fields like law, finance, healthcare and government would receive "traceable reasoning that meets compliance requirements" as "every conclusion can be traced back through its logical steps", Mistral said. The company's claim gestures towards the challenge of so-called 'interpretability,' or working out how AI systems arrive at a given response. Since they are "trained" on gigantic corpuses of data rather than directly programmed by humans, much behaviour by AI systems remains impenetrable even to their creators. Mistral also vaunted improved performance in software coding and creative writing by Magistral. Competing "reasoning" models include OpenAI's o3, some versions of Google's Gemini and Anthropic's Claude, or Chinese challenger DeepSeek's R1. The idea that AIs can "reason" was called into question this week by Apple, the tech giant that has struggled to match achievements by leaders in the field. Several Apple researchers published a paper called "The Illusion of Thinking" that claimed to find "fundamental limitations in current models" which "fail to develop generalisable reasoning capabilities beyond certain complexity thresholds".
The Star
11-06-2025
- Business
- The Star
France's Mistral unveils its first 'reasoning' AI model
The AI was designed for "general purpose use requiring longer thought processing and better accuracy" than its previous generations of large language models (LLMs), the company added. — Pixabay PARIS: French artificial intelligence startup Mistral on Tuesday announced a so-called "reasoning" model it said was capable of working through complex problems, following in the footsteps of top US developers. Available immediately on the company's platforms as well as the AI platform Hugging Face, the Magistral "is designed to think things through – in ways familiar to us," Mistral said in a blog post. The AI was designed for "general purpose use requiring longer thought processing and better accuracy" than its previous generations of large language models (LLMs), the company added. Like other "reasoning" models, Magistral displays a so-called "chain of thought" that purports to show how the system is approaching a problem given to it in natural language. This means users in fields like law, finance, healthcare and government would receive "traceable reasoning that meets compliance requirements" as "every conclusion can be traced back through its logical steps", Mistral said. The company's claim gestures towards the challenge of so-called "interpretability" – working out how AI systems arrive at a given response. Since they are "trained" on gigantic corpuses of data rather than directly programmed by humans, much behaviour by AI systems remains impenetrable even to their creators. Mistral also vaunted improved performance in software coding and creative writing by Magistral. Competing "reasoning" models include OpenAI's o3, some versions of Google's Gemini and Anthropic's Claude, or Chinese challenger DeepSeek's R1. The idea that AIs can "reason" was called into question this week by Apple – the tech giant that has struggled to match achievements by leaders in the field. Several Apple researchers published a paper called "The Illusion of Thinking" that claimed to find "fundamental limitations in current models" which "fail to develop generalizable reasoning capabilities beyond certain complexity thresholds". – AFP
