Former Google CEO launches AI+Science moonshot plan to achieve OpenAI's goals

Former Google CEO Eric Schmidt is launching a massive AI science non-profit startup aimed at using artificial intelligence to address challenges facing scientific research

Picture

He invited two outstanding scientists to lead this non-profit initiative:

Francis Crick Institute Application Samuel Rodriguez, founder of the Biotechnology Laboratory, and Andrew White, a professor at the University of Rochester and a pioneer in the use of artificial intelligence in chemistry. They are all relatively young academic stars in their respective fields who have already achieved outstanding achievements

Schmidt, Rodriguez and White all believe that artificial intelligence will change the future of scientific research

In an article titled "How AI Will Change the Way Science is Done" published in "MIT Technology Review", Schmidt expressed his vision

With the advent of artificial intelligence, science will become more exciting and, in some ways, unrecognizable. The impacts of this shift will reach far beyond the confines of the laboratory, they will affect us all

Pictures

at On the other hand, Rodriques and White have put forward predictions and assumptions that artificial intelligence will subvert science in their own experimental websites or public speeches

Rodriques said: "We need a core AI A team of researchers and core scientists who will work together and employ rapid iteration cycles to build tools that leverage cutting-edge technologies and bring real value to scientists."

Jim ·Fan believes this company has great potential. If LLM and intelligent robots become the infrastructure of future scientific research, experiments like LK-99 will no longer remain at the level of manual alchemy

Picture

According to people familiar with the matter, the work Schmidt is doing is based on OpenAI as a template, but the funding comes from Schmidt Futures, which Schmidt co-founded with his wife Wendy. Basically, the funds for the activities were paid by Schmidt personally

Jim also expressed concerns about whether Schmidt’s organization can continue

AI Science Moonshot

Determining the basis for forming scientific insights and theories is how to collect, transform and understand data

Among them, the collection of data and analysis are the basis of scientific understanding and discovery.

In the 1950s, the introduction of digital technology paved the way for the widespread use of computers in scientific research

Since 2010, The rise of deep learning enables artificial intelligence to provide valuable guidance by identifying scientifically relevant patterns from large data sets. This significantly expands the scope and ambition of the scientific discovery process

Scientific discovery is a multifaceted process involving several interrelated stages, including hypothesis formulation, experimental design, data collection and Analysis

Although there are differences in scientific practices and procedures at different stages of scientific research, artificial intelligence algorithms have the ability to span traditionally isolated disciplines

Artificial intelligence (AI) is increasingly used in the integration of massive data sets across disciplines and fields, precise measurements, experimental guidance, and exploration of data-compatible theoretical spaces, while also providing actionable and Reliable models and integration with scientific workflows

The application of AI can enhance the design and execution of scientific research. It can collect, visualize and process data by optimizing parameters and functions, automating procedures to explore a large number of candidate hypotheses to form theoretical perspectives, and generate hypotheses and estimate their uncertainties to make recommendations for relevant experiments

Picture

Science in the AI ​​Era

However, using artificial intelligence to conduct scientific research does not mean that it can be done casually

One of the biggest challenges is the huge hypothesis space in scientific questions. This makes systematic exploration unfeasible

In the field of biochemistry, it is estimated that the number of drug molecules that need to be explored is about 10 to the 60th power

While AI systems can revolutionize scientific workflows by speeding up processes and providing predictions with near-experimental accuracy

Obtaining reliably annotated datasets is a challenge for AI models This is a rather large project that may require a lot of time and resources for experiments and simulations

In recent developments, AlphaFold was developed by Google DeepMind and successfully solved a 50-year-old problem. Protein Folding Puzzle

Molecular system simulations of millions of particles16 powered by AI-powered AlphaFold demonstrate the potential of artificial intelligence in solving challenging scientific problems

One problem is that people’s opacity about the internal operation of AI will reduce trust in the prediction results and also limit its applicability in certain fields

For example, the output of the model must conform to real-life conditions before practical application. For example, human space exploration and areas that inform policymaking, such as climate science, etc.

Future Scientific Research

For AI majors Looking into the future, the demand for knowledge will be affected by two forces

First, one of the areas that will benefit from AI applications is autonomous driving. Secondly, the introduction of AI smart tools will enhance state-of-the-art technology and create new opportunities, such as applications in biological, chemical or physical processes, such as using AI to study nuclear fusion reactions, etc.

The composition of the future research team will change based on these two forces, including Al experts, software and hardware engineers, and new forms of cooperation involving all levels of government, educational institutions and enterprises

As state-of-the-art models continue to scale, energy consumption increases and computational costs become increasingly expensive. As a result, large technology companies are investing in computing infrastructure and cloud services, constantly pushing the limits of scale and efficiency. This means that both for-profit and non-academic organizations will use large-scale The computing infrastructure

# is necessary so that higher education institutions can better integrate multiple disciplines. In addition, academic institutions often possess unique historical databases and measurement techniques that may not exist elsewhere but are essential to Al Science

These complementary assets It will promote a new model of industry-university cooperation and have an impact on the selection of research questions

Reference:

//m.sbmmt.com /link/db261d4f615f0e982983be499e57ccda

The above is the detailed content of Former Google CEO launches AI+Science moonshot plan to achieve OpenAI's goals. For more information, please follow other related articles on the PHP Chinese website!