OpenAI's Advancements: From Simplified AI Language to Potential Chip Development

OpenAI has trained AI models to produce text that humans can easily understand, here's how - Times of India

Tech companies like Google and ChatGPT maker OpenAI are pushing innovation to enhance the output provided by chatbots. By harnessing the power of AI, machine learning and advanced data analytics, they are optimising language models to improve the responses. In one such case, OpenAI has reported a breakthrough in which the company trained strong language models to produce text that is easy for weak language models to verify.The Microsoft-backed company found that this training also made the text easier for humans to evaluate. "Making sure that language models produce understandable text is crucial to making them helpful for people, especially when dealing with complex tasks like solving math problems," OpenAI said in a blog this week. The company said that it found that despite its AI models spitting out factually correct answers, they were harder to understand. "When we asked human evaluators with limited time to assess these highly optimised solutions, they made nearly twice as many errors compared to when they evaluated less optimised solutions. This finding highlights the importance of not just correctness, but also clarity and ease of verification in AI-generated text," it added. How OpenAI found a solution to this problem OpenAI trained advanced language models to create text that weaker models can easily verify, which humans could also evaluate more effectively - a process called improving legibility. The company deployed a 'prover-verifier game' wherein two players: a "prover" that generates a solution and a "verifier" that checks it for accuracy. "To improve legibility, we optimise chain-of-thoughts on grade-school math problems to be verifiable by weaker models and study whether this makes them more legible to humans. This training procedure is inspired by the Prover-Verifier Game, a game-theoretic framework to encourage learning agents to solve decision problems in a verifiable manner," OpenAI said. OpenAI says it used a training scheme in which a strong model produced solutions that a much weaker model, the verifier, can easily verify - large and small models from the GPT-4 family. The TOI Tech Desk is a dedicated team of journalists committed to delivering the latest and most relevant news from the world of technology to readers of The Times of India. TOI Tech Desk's news coverage spans a wide spectrum across gadget launches, gadget reviews, trends, in-depth analysis, exclusive reports and breaking stories that impact technology and the digital universe. Be it how-tos or the latest happenings in AI, cybersecurity, personal gadgets, platforms like WhatsApp, Instagram, Facebook and more; TOI Tech Desk brings the news with accuracy and authenticity.

ChatGPT-maker OpenAI wants to make its own AI chips - Times of India

OpenAI, the company behind ChatGPT, is in talks with semiconductor designers, including Broadcom, to develop a new AI chip, as part of a broader strategy to expand its computing capacity and reduce reliance on Nvidia's GPUs. The discussions, which are still in early stages, come amid OpenAI CEO Sam Altman's ambitious plans to raise up to $7 trillion for a massive chip-making project. According to The Information, OpenAI is exploring the possibility of producing its own AI chips to overcome the shortage of expensive graphics processing units (GPUs) crucial for developing AI models like ChatGPT, GPT-4, and DALL-E3. The company has reportedly been hiring former Google employees who worked on the tech giant's tensor processing unit. The Financial Times reports that these talks are part of a larger effort led by Altman to bolster the supply of components and infrastructure needed to run increasingly powerful AI models. "The limiting factor of AI is capacity: chip capacity, energy capacity, compute capacity. [OpenAI] is not just going to sit back and let others build [that] when they are on the front line," a source familiar with OpenAI's plans told the Financial Times. Altman's vision extends beyond mere chip development. The Wall Street Journal had earlier reported that he aims to raise between $5 trillion and $7 trillion to boost global chip-building capacity. This amount surpasses the current size of the global chip market and exceeds the combined market capitalizations of Apple and Microsoft. The OpenAI CEO has been engaging with various stakeholders, including chipmakers, partners like Microsoft, government bodies, and financial backers. He has reportedly met with high-profile figures such as UAE's top security official Sheikh Tahnoun bin Zayed al Nahyan and US Commerce Secretary Gina Raimondo to discuss the project. While OpenAI's plans are ambitious, they are in line with what other players have been doing recently. The scarcity of Nvidia's $40,000 H100 chips, considered the most powerful AI chips on the market, has prompted tech giants like Microsoft, Google and Meta to develop their own in-house solutions. Meanwhile, US government has also announced a $5 billion investment, which has prompted major chipmakers to invest in the US, helping companies to start off their in-house chip ambitions. The TOI Tech Desk is a dedicated team of journalists committed to delivering the latest and most relevant news from the world of technology to readers of The Times of India. TOI Tech Desk's news coverage spans a wide spectrum across gadget launches, gadget reviews, trends, in-depth analysis, exclusive reports and breaking stories that impact technology and the digital universe. Be it how-tos or the latest happenings in AI, cybersecurity, personal gadgets, platforms like WhatsApp, Instagram, Facebook and more; TOI Tech Desk brings the news with accuracy and authenticity.