Google DeepMind co-founder Demis Hassabis’s Nobel Prize win shows how promising AI is in science

AI is making a splash at the Nobel Prizes this week. AI pioneers John Hopfield and Geoffrey Hinton won the 2024 Physics Prize for their breakthroughs in machine learning that led to today’s AI boom. Yesterday, Demis Hassabis and John Jumper of Google DeepMind, along with David Baker, a professor of biochemistry at the University of Washington, received the chemistry prize for discovering techniques to predict and develop novel proteins that could transform the way therapeutic drugs are made.

A 50-year dream

Heiner Linke, chairman of the Nobel Committee for Chemistry, said in a press release that the researchers who take home the chemistry prize “are fulfilling.”[ed] a 50-year-old dream.” This dream was actually the vision of a previous Nobel Prize winner, the chemist Christian Anfinsen, who postulated back in 1973 that it was possible to predict the shape of a protein based solely on knowledge of its DNA sequence. A consequence of this idea was that it would be possible to manipulate DNA to design proteins, which are the building blocks and engines of life, with specific functions, since mostly the shape of a protein determines what it does.

Progress toward realizing this dream began in 2003, when Baker used the 20 different amino acids found in proteins to create new proteins unlike any other. Then in 2020, Hassabis and Jumper made a stunning breakthrough with their machine learning model AlphaFold2, allowing them to predict the structure of virtually all 200 million proteins that researchers have identified. According to the press release, the model has since been used by more than two million people from 190 countries.

The award brings Hassabis full circle, as he began his AI endeavors by teaching computers to master games like Go, but always dreamed bigger. Furthermore, it represents the best that AI can offer humanity.

From a distant vision to the highest prize in science

More than a decade ago, Hassabis envisioned a future in which AI models would achieve monumental scientific breakthroughs. In 2014, when his AI lab DeepMind was still largely focused on teaching machines to play games, and shortly after it was sold to Google, Hassabis said MIT Technology Review about his vision for “AI Scientists”.

“But Hassabis sounds more enthusiastic when he talks about going beyond simply optimizing the algorithms behind today’s products,” the article said, after mentioning how AI is being used to refine YouTube recommendations or improve enterprise search could. “He dreams of creating ‘AI scientists’ who could, for example, generate and test new hypotheses about diseases in the laboratory.”

DeepMind began working on protein folding in 2016 and won awards in 2018 for the first version of AlphaFold. The company followed up two years later with AlphaFold 2 and announced in July 2022 that it had successfully predicted virtually all known proteins. Earlier this year, the research lab, now known as Google DeepMind, introduced AlphaFold 3, which it claims can predict the interactions of proteins with DNA, RNA and various other molecules, offering significant accuracy improvements over the previous model.

Overall, it’s an amazing achievement for Hassabis – and clear evidence of how quickly AI is evolving and improving. Ten years ago it was all just a vision. This week the breakthrough is real, its impact is being felt around the world, and it was just awarded science’s highest prize.

The most positive impact of AI

While the use of AI models is controversial in many industries and some executives question the value of today’s AI software in achieving financial returns, AI’s impact on scientific discovery is already beginning to bear fruit, as breakthroughs like AlphaFold make clear. I’m often asked about the positive impact AI can have on humanity or what I think is the most exciting way AI is being used. Scientific research and medical breakthroughs are always my answer.

This summer, an AI model developed by Cambridge scientists showed 82% accuracy in predicting the progression of Alzheimer’s disease, outperforming clinical tests. Several AI-discovered drugs have made it into Phase I and Phase II testing, including just last week a cancer drug from Recursion.

The success of AI in areas like drug discovery and medicine is by no means guaranteed or free from problems like bias, but it is clearly a worthwhile endeavor with some early successes worth celebrating.

And with that, there’s even more AI news here.

Sage Lazzaro
[email protected]
sagelazzaro.com

AI IN THE NEWS

Amazon uses AI to show drivers which packages they need to take with them for delivery. The new technology, called Vision-Assisted Package Retrieval (VAPR), will be installed in the company’s delivery trucks and will mark packages with a green or red light to indicate which ones are destined for delivery at the current stop. Additionally, there is an audible alert to let drivers know they have selected the correct package. Amazon hopes the technology will make the process easier and more efficient, eliminating the need for drivers to shuffle through their trucks. You can read more here TechCrunch.

OpenAI predicts that the company could lose up to $14 billion in 2026 and will not be profitable until 2029. That’s according to an article in The Information that cited OpenAI financial documents the company said it had seen. The company also predicts it will spend more than $200 billion by the end of the decade, with much of that amount going toward training new AI models. Total losses could reach $44 billion between 2023 and 2028. The document also shows that the OpenAI project will generate $100 billion in revenue by 2029, with ChatGPT continuing to account for the majority of revenue. However, the company’s current cash burn is portrayed as lower than some previous news reports had suggested, as the company only spent about $340 million in the first half of 2024 and the balance sheet before that still shows a cash balance of $1 billion reported its most recent $6.6 billion funding round valued the company at $157 billion.

Anthropic halves batch processing costs with a new API. The Message Batch API allows developers to send batches of up to 10,000 queries per batch, costing 50% less than standard API calls, according to Anthropic. Processing takes place within 24 hours, offering the opportunity to save costs on tasks that are not time-critical. The new API is now available for Claude 3.5 Sonnet, Claude 3 Opus and Claude 3 Haiku and provides a solution to one of the main problems of large language models – the high cost of inference. For more information, see VentureBeat.

Advertisers are tempering their expectations of AI as early efforts fail. After a lot of initial hype, skepticism is growing in the industry and people are starting to believe that many of the new AI tools for advertisers aren’t offering as big a leap forward as they had hoped. Business Insider The shift was reportedly evident at New York Advertising Week, where “industry insiders appeared to be talking as much about the limitations as the promises of AI.” Major agencies continue to plan significant investments in AI.

The AI ​​implementation was a huge mess, or at least an extraordinary challenge. That’s according to a Feature I wrote this week for Mercury’s Meridian magazine, where I take an in-depth look at the many challenges companies face in implementing AI into their products and internal processes. To name a few, companies are struggling to cut through the hype, figure out what use cases AI is good for, navigate fast-moving regulation, protect their IT stacks from the spread of AI, deal with hallucinations, and… dealing with a variety of complicated copyright and copyright infringement cases. Security, privacy and compliance concerns. And that doesn’t even include the technical challenges.

LUCKY ON AI

Exclusive: Zoom’s future isn’t video, it’s AI for work, says CEO Eric Yuan – but can it challenge Microsoft and Google? – by Sharon Goldman

The USA wants to prevent Google from monopolizing the emerging AI search market – by David Meyer

New Nobel laureate and AI godfather Geoffrey Hinton says he is proud that his student fired OpenAI boss Sam Altman – by Christiaan Hetzner

Wimbledon will ban linesmen from its tennis matches after 147 years – and turn to AI instead – by Prarthana Prakash

Whirlpool’s CIO says lessons from the IoT hype cycle are applicable to generative AI – by John Kell

AI CALENDAR

22-23 Oct.: TedAI, San Francisco

28-30 Oct.: Voice & AI, Arlington, Virginia.

19–22 November: Microsoft Ignite, Chicago

2nd-6th December: AWS re:Invent, Las Vegas

8th–12th December: Neural Information Processing Systems (Neurips) 2024, Vancouver, British Columbia

9th–10th December: Fortune Brainstorm AI, San Francisco (register Here)

EYE ON AI NUMBERS

20,000 to 34,000

That’s how many users interact with the Rabbit R1 — the $200 AI gadget that can act as a digital assistant and perform actions like calling an Uber — on a daily basis, according to an interview with CEO Jesse Lyu on the Decoder podcast. During the episode, Lyu denied previous reports of Fast company However, the company only has 5,000 daily active users and told the publication that 5,000 people use the R1 at any given time, rather than per day. Fast company has corrected his article. The entire interview is an interesting – and sometimes heated – conversation about this new type of device, where AI is going, and what happens if the services the R1 connects users to (Spotify, Uber, etc.) decide they won’t I want the company to act as a middleman.