183 - Guest: Oren Etzioni, AI in Science, Professor Emeritus, part 2
Dec 18, 2023
auto_awesome
Guest Oren Etzioni, Professor Emeritus of Computer Science at the University of Washington and founding CEO of the Allen Institute for Artificial Intelligence, discusses AI's impact in scientific research, the development of Aristo and scaling AI models, the tension between positive and negative aspects of AI, exploring AI technology and the nature of intelligence, and AI progress and settlement of actors' strike.
The scale of data used for training large language models like GPT-3.5 significantly impacts their performance and transformative potential in fields like artificial intelligence.
Ethical considerations and responsible deployment of AI technologies are crucial, highlighting the need for an AI Hippocratic oath and targeted regulations to address specific applications.
Deep dives
The Power of Large Language Models
The podcast episode discusses the transformative power of large language models in the field of artificial intelligence. The guest, Aaron Etsioni, shares his experience working on the Aristo project and the significant progress made in using AI to pass grade-level science tests. He emphasizes the surprising impact of scaling up models like GPT-3.5, highlighting that the scale of the data used for training makes a significant difference. While acknowledging that there are still challenges to be overcome, Etsioni expresses optimism about the potential of AI and the ongoing development of even more advanced models like GPT-5.
Ethical Considerations in AI
Etsioni discusses the importance of ethical considerations in the field of AI. He proposes the idea of an AI Hippocratic oath for practitioners, emphasizing the need for responsible and accountable development and deployment of AI technologies. He distinguishes AI as a tool rather than a weapon, highlighting the diverse range of uses and potential benefits, as well as the potential risks associated with misuse. Etsioni advocates for specific regulations targeted at particular applications, rather than broad and potentially ineffectual regulations. He also emphasizes the importance of transparency and detection in addressing the growing issue of misinformation.
Implications for the Film Industry
The podcast briefly explores the impact of AI on the film industry, specifically in relation to the use of synthetic likenesses and performances. It mentions the recent settlement between SAG-AFTRA and the Alliance of Motion Picture and Television Producers, which includes provisions for compensating actors and securing their consent when using their likeness in AI-generated content. The settlement places importance on the rights of actors and the need for ethical considerations in artificial intelligence. The podcast also hints at future discussions on AI in motion picture production.
At the intersection of scientific research and artificial intelligence lies our guest Oren Etzioni, professor emeritus of Computer Science at the University of Washington and most notably the founding CEO of the Allen Institute for Artificial Intelligence (AI2) in Seattle, founded by the late Paul Allen, co-founder of Microsoft. His awards include AAAI Fellow and Seattle’s Geek of the Year.
Oren grew the institute to a team of over 200 researchers and created singularly important tools such as the Semantic Scholar, search engine that can understand scientific literature, and Mosaic, a knowledge base formed by extracting scientific knowledge from text. This is hugely important because of just how much the rate of research paper creation now outstrips the ability of researchers to read it. AI could transform the productivity of scientific research by unprecedented measures.
In this conclusion of the interview we talk about AI2’s scientific assistance project called Aristo, Oren’s views on the concerns about AI and how to address them, and his Hippocratic Oath for AI practitioners.
All this plus our usual look at today's AI headlines.