Open AI / ChatGPT CEO, Sam Altman of St. Louis recently visited his school, John Burroughs High School, to talk to students about new developments in Artificial intelligence and his experiences of transitioning from a scholar to an entrepreneur.
He visited it only a day after the latest release of the newest model of ChatGPT, referred to as OpenAI o1 on the previous Thursday. A new model is described for its capabilities to handle more complicated problems in science, computing, and Mathematics.
As per the understanding of Altman, the developments which led to the new model took several years and according to him the new model is good for the field of artificial intelligence.
ChatGPT founder Sam Altman Statement
Reported , Sam Altman said , “We launched a model for the first time that can reason, sort of like a human model. It scores at a benchmark called GPQA, which is sort of hard questions and a bunch of topics. It does better than human PhD students. It does great at math competitions, and great at programming competitions. It’s the first time, at least for me using the model, where I’m like, ‘Ah,’ this is really, really smart.”
Stating this during a fireside chat at the recent T-Mobile Capital Markets Day, Altman lauded the o1 models and their capacity to ’reason.’ This, he said, will open up a new world of opportunities in AI that was not possible with the GPT class of models that preceded the o1 models.
According to Altman, such reasoning models will make it possible to realize level 3 AI, the kind of AI dubbed agentic systems.
During the fireside, Altman also focused on the fact that the current GPT series of models is quite effective. This has been made possible by GPT-4o which supports such models as ChatGPT and Advanced Voice. It’s natively multimodal but functions as any prior AI – on a token-by-token basis.
Altman said “The GPT series of models were amazing at ‘system-1’ type thinking, but what we wanted was systems that could reason. There is so much value if AI can reason across problems. o1 is the first system that can do complex reasoning, and if you give it a challenging problem you can get extraordinary results.”
Here, system 1 can be described as rapid, holistic, and automatic thinking. this system 2, which is where the models are going with o1, means more slow, conscious thinking but at the same time having more logical reasoning.
OpenAI o1
As reported, OpenAI o1 is a new category of a large language model; OpenAI o1 is an entirely new class. Prior generations and strategies, including the GPT family, answer the user’s attention token by token, which causes hallucinations or even bizarre and unsubstantiated facts.
There are several ways of avoiding this kind of difficulty such as using wider context windows where the particular error and others marked by the AI can be corrected and seen or using memory functions similar to the former but across different chats. But these are just sticking plasters and a change was needed at a paradigm level.
The world needs AI to invent new things, develop weapons against diseases, and ponder over better means of generating energy and others. Their strategy with o1 was different and it shifted to a concept of a chain of thought, where once you feed the AI model with a prompt, it disappears. The AI then works at solving in the same progressive way a human would do before presenting the result.
Sam Altman’s approach to Open o1
With o1, OpenAI reversed the process, shifting to a chain-of-thought model where it insulates the AI model once one provides it with a prompt for the model to work on before it unearths the solution, in a manner akin to how a human might analyze a problem before arriving at the answer.
The following is a piece of assurance that all students who have been to school must have lost marks for not showing their working out properly: Well, now even AI has to too.
When talking to T-Mobile, Altman identified healthcare and education as being two good examples of where reasoning models similar to o1 might have a lot of impact. The idea that “if you imagine every student getting personalized tutoring, built for them, along with other learning experiences” is considerable.
His biggest wish is that AI will be able to advance the sciences and related areas. Alibaba: “If AI can help us invent new things, find cures for diseases, think of new power sources and other creativities, then it has won.”
Lastly, implying that Deep Learning would lead to AGI and more, Altman affirmed OpenAI’s strategy and the company’s willingness to change what it learned.