Chinese mathematician Terence Tao leads the White House Generative AI Working Group, and Li Feifei will speak at the group

·The Generative AI Working Group established by the President’s Council of Advisors on Science and Technology is designed to help assess key opportunities and risks in the field of artificial intelligence and provide advice to the President of the United States on ensuring that these technologies are developed and deployed as fairly, safely, and responsibly as possible .

·AMD CEO Lisa Su and Google Cloud Chief Information Security Officer Phil Venables are also members of this working group.

Chinese-American mathematician and Fields Medal winner Terence Tao.

On May 13, local time, Chinese-American mathematician and Fields Medal winner Terence Tao announced that he and physicist Laura Greene will co-lead the generative program of the U.S. President’s Council of Advisors on Science and Technology (PCAST). Artificial Intelligence Working Group. The group will hold a public meeting during the PCAST ​​conference on May 19, local time. Demis Hassabis, the father of AlphaGo and founder of DeepMind, and Stanford University professor Li Feifei will give speeches at the meeting.

The U.S. President’s Council of Advisors on Science and Technology (PCAST) Generative Artificial Intelligence Working Group will hold a public meeting on May 19, local time.

Chinese-American mathematician and Fields Medal winner Terence Tao announced the news on his blog.

According to Tao Zhexuan's blog, the generative artificial intelligence working group mainly studies the impact of generative artificial intelligence technology in the scientific and social fields, including large-scale text-based language models such as ChatGPT, image generation such as DALL-E 2 and Midjourney processors, as well as models for scientific applications such as protein design or weather forecasting. It is worth mentioning that AMD CEO Lisa Su and Google Cloud Chief Information Security Officer Phil Venables are also members of this working group.

Member of the U.S. President’s Council of Advisors on Science and Technology (PCAST) Generative Artificial Intelligence Working Group.

According to an article published by the White House on its official website, the Generative Artificial Intelligence Working Group established by PCAST ​​aims to help assess key opportunities and risks in the field of artificial intelligence, and to ensure that these are developed and deployed as fairly, safely, and responsibly as possible. Technology advises the President of the United States.

The article defines generative artificial intelligence as a type of AI system that, after training on a large data set, can be used to generate text, images, videos, or other output based on given prompts. These technologies are developing rapidly and have the potential to revolutionize many aspects of modern life. These tools are used in scientific research to design new drugs, proteins, and materials and are expected to facilitate the process of scientific discovery. In medicine, generative AI has the potential to provide recommendations to healthcare professionals. These tools can speed up writing computer code, writing presentations, and summaries in the workplace.

The article also points out that generative artificial intelligence models may also be used for malicious behaviors, such as creating false information, spreading misinformation, and impersonating personal identities. If used without safeguards, generative AI could polarize society, exacerbate prejudice and inequality in society, and more generally threaten democracy by making it difficult for citizens to understand what is happening in the world. Additionally, generative AI systems may violate privacy and destroy intellectual property.

Advising the President is the PCAST ​​Generative Artificial Intelligence Working Group, which identifies additional needs and opportunities and proposes the best ways to address them.

In the meeting held on the 19th, two expert groups from the Generative Artificial Intelligence Working Group will introduce the current situation in this field. The speakers include:

AI Enabling Science:

Anima Anandkumar (Director of Machine Learning Research at NVIDIA, Professor of Computer and Mathematical Sciences at Caltech)

Demis Hassabis (Founder of DeepMind)

Li Feifei (Professor of Computer Science at Stanford University)

Al Impacts on Society:

Sendhil Mullainathan (Professor of Computing and Behavioral Sciences, University of Chicago Booth School of Business)

Daron Acemoglu (Professor of Economics, MIT)

Sarah Kreps (Professor of Government, Cornell University)

Tao Zhexuan said that the working group is soliciting public input to determine how to identify and promote beneficial deployment of generative artificial intelligence and find the best way to reduce risks. “Our initial focus is on the challenging topic of how to detect, counter, and mitigate AI-generated disinformation and deepfakes without sacrificing the free speech and public engagement with elected officials needed for a healthy democracy to function; going forward We may also make further requests around other aspects of generating artificial intelligence."

The emergence of ChatGPT has attracted great attention from top mathematicians, such as Terence Tao, who began to study how to use artificial intelligence to assist in completing their work. In an article in "Nature" magazine "How does AI change mathematics?" "The rise of chatbots highlights the importance of conversation", Andrew Granville, a number theorist at the University of Montreal in Canada, also said, "We are looking at a very specific question: Will machines change mathematics?" Mathematics Economist Kevin Buzzard agrees, “The fact that big mathematicians like Fields Medalists and other very famous mathematicians are now interested in this field shows that it is changing in a way that it has never been before. Got 'hot'."

Previously, Tao Zhexuan wrote on the decentralized social network Mastodon, "I can clearly say that today is the first day that GPT-4 saved me a lot of tedious work." In the attempt, Tao Zhexuan discovered ChatGPT ChatGPT has many hidden functions, such as searching formulas, parsing documents with code format, rewriting sentences in academic papers, etc. Sometimes ChatGPT can also perform semantic searches on unfinished math problems to generate some tips.

He also compared the comparative advantages between himself and current generative artificial intelligence tools:

I have developed enough technology to optimize my workflow for the tasks I perform every day, and AI tools don’t add much value to me. Most obviously in studying math, but also writing emails. I installed a plugin that lets GPT-4 write an email reply when I click a button, but I almost never use it because after decades of practice I've become able to write appropriate email responses quickly.

For those tasks for which I have some expertise but little practice, AI tools are helpful: often I can use them to create a first draft of the output, which I can then validate and revise, or at least use as a source of inspiration. Sometimes inspiration comes from the AI's flaws, which aligns with the idea of ​​Cunningham's Law, but is more effective than trying to solve the problem myself. Examples of this type cover data processing, translation, and text formats that I rarely use, such as public speaking and rules documents.

For those tasks for which I have little expertise and do not require extreme quality and reliability, I can simply ask the AI ​​tool and more or less follow its suggestions. Here, AI functions as a slightly more convenient version of a traditional search engine.

Finally, for those tasks for which I have no expertise but require quality and reliability that neither AI nor myself can solve, I must consult human experts, such as repairing a complex, expensive, delicate device.

An interesting attempt is that Tao Zhexuan asked ChatGPT to identify "Kummer's theorem" (used to describe how to find the optimal solution under constraints) from the description. Although ChatGPT cannot give the correct answer, it can still be based on Legendre's formula gives an approximate answer. Regarding the performance of ChatGPT, Tao Zhexuan said that artificial intelligence plays a role in mathematics, providing a preliminary approximate answer, and then combined with a traditional search engine, the correct answer can be easily found.

The above is the detailed content of Chinese mathematician Terence Tao leads the White House Generative AI Working Group, and Li Feifei will speak at the group. For more information, please follow other related articles on the PHP Chinese website!