OpenAI CEO Sam Altman defends large language models as more than just “stochastic parrots”
OpenAI CEO Sam Altman is pushing back against critics who argue that large language models are nothing more than “stochastic parrots” that lack true understanding and problem-solving capabilities. Altman believes that as AI models continue to grow in power, they will only improve in their abilities.
During an event at the UN General Assembly meeting in Manhattan, Altman highlighted the recent release of the o1 model, internally known as Strawberry, which he claims is better at math than all but the top few hundred students in the U.S. and works at the level of upper-echelon programmers. Altman sees this as a significant step towards AI advancing scientific discovery.
Altman referenced mathematician Terence Tao’s comparison of AI to a grad student, noting that AI has progressed from being very incompetent to now being a mediocre grad student that can be given tasks. He envisions AI becoming a unique and useful research partner in the near future.
However, not everyone shares Altman’s optimism. UN tech envoy Amandeep Singh Gill expressed skepticism about using large language models as a shortcut to solving complex global issues. Gill believes that some tasks could be better accomplished through other means, such as human interaction.
Despite differing opinions, Altman remains confident in the value of AI and its potential to create significant benefits. He emphasized that the energy use of AI relative to the value it creates is minimal and is expected to become even more efficient in the future.
Altman also hinted at more groundbreaking news from OpenAI in the near future, teasing the possibility of a new AI model like GPT-5. He also revealed that the AI-powered device he is working on with Jony Ive will not be a phone, leaving room for speculation about what the project may entail.
As the debate over the capabilities and future of AI continues, Altman’s unwavering confidence in the technology’s potential for advancement sets the stage for further innovation and discovery in the field.