Google DeepMind CEO on Drug Discovery, Hype, Isomorphic

Google DeepMind CEO and Co-Founder Demis Hassabis speaks with Bloomberg's Tom Mackenzie about the latest version of the company's AlphaFold AI system (AlphaFold 3), intended to tackle problems in biology including disease protection and treatment. Hassabis also discusses the hype cycle in AI, as well as the timeframe for AI-developed drugs and why he thinks Alphabet subsidiary Isomorphic Labs could be a $100 billion company.

00:00 - Intro
00:32 - AlphaFold 3, understanding dynamic picture of protein interactions
01:28 - Timeline for AI derived drugs. Partnerships with Eli Lilly and Novartis
02:03 - Isomorphic Labs, drug discovery
02:05 - Potential for mRNA vaccines
03:25 - Competitive threat from OpenAI
04:24 - Sustainability of AI
06:10 - Rationalization in AI sector, more closures and failures
07:14 - UK AI safety summit
08:47 - Big election year - are Labour as invested in AI?
09:20 - What are the big things coming this year?
10:31 - Isomorphic Labs, under pressure to commercialize?
12:20 - What keeps you up at night?
--------
More on Bloomberg Television and Markets


Connect with Bloomberg Television on:

Connect with Bloomberg Business on:

More from Bloomberg:


Watch more on YouTube:

[MrSchweppes]

Demis Hassabis deserves Nobel Prize in medicine for the invention of the AlphaFold system!

[jackignatius]

Tom Mackenzie comes to this interview with the most penetrating question set I've seen over the last year, and Dmis Hassabis ably responds. This is a valuable interview.

[sidnath]

One thing that stands out with Demis is that his drive for success was never about to 'outcompete' others - he aims to find out the truth and help push research and innovation to get there, for everyone, and not just the few.

[PeterSedesse]

I have a degree in biochemistry back from 1992... and we actually talked about this in some classes. Imagine a steel factory where you need to precisely bend a one mile long rod of metal 10s of thousands of times and put each bend at a precise position. Actually figuring out where each bend needs to be is a huge task all on it's own, but then figuring out the factory configuration to where you can do all of those bends, while still allowing the rest of the rod to fit move through the factory is a Herculean task. We knew back then that someday computers would not only be able to design the drug, but also be able to describe the process we needed to use to make that drug.

[jonnysolaris]

He went to my highschool. And back then he was already a genius, made millions out of video games as a teenager

[litbmeinnick]

This is what I like about Google's Deepmind. They still have their super cool and humble boss in place. Not fighting like the OpenAI folks.

[cguyglu]

The interviewer seemed well prepared. Good interview

[CalumnMcAulay]

Great to see him involved in drug discovery, we are getting ever more closer to age reversal, a great future of possibilities exist!

[hugopennmir]

Demis is in another league, a genius, the real AI GOAT

[nickb]

dam the universe did good to make someone like demis

[sdkfgnrjdi]

As a computational biologist from a small biotech company, I am scared that these bigtech AI companies will overtake my job. But as a human being, I hope they make a huge progress to find better drugs in more efficient way.
I am also thankful that Demis is interested in biology and various science fields rather than just making targeted advertisement.

[lemuhuru]

Ai is already used extensively in the most advanced chip designs as Jensen Huang stated for Nvidia. It's only natural that we see it play a major role in drug design, this is going to be an incredible next decade.

[Superteastain]

He's my fav AI oddball. Like, if you dropped a tenner and he saw I reckon he'd pick it up and notify you.

[etienneekpo]

Great work! More spotlight should be placed on such advancement and research. A clear economic model should emerge on its application with the right measures to prevent its misuse.

[ReflectionOcean]

00:00:04 Alpha Fold three predicts biomolecule structures.
00:00:25 Understanding protein interactions for drug discovery.
00:02:50 Potential implications for RNA vaccines and biologics.
00:03:32 AI advancements equip Alphabet for competitive challenges.
00:04:44 Sustainability challenges outweighed by AI benefits.
00:06:28 Rationalization expected in the AI industry.
00:07:18 UK's pace in AI development and infrastructure.
00:08:29 AI's potential benefits and responsibilities for society.
00:10:45 Revolutionizing drug discovery for societal and commercial value.
00:11:39 AI-generated drugs may impact drug approval processes.

[danecjensen]

Chapters (Powered by ChapterMe) -
00:00 - Alpha Fold 3 Dynamic protein folding, implications for drug discovery
01:17 - Alpha Fold 3 predicts protein binding and drug discovery
01:39 - Generative AI drug development partnerships with Eli Lilly and Novartis
02:36 - Alpha Fold 3 opens up new vaccine opportunities
04:36 - Challenges and benefits of generative AI
07:41 - UK planning laws hinder AI growth
08:07 - Economic and political opportunities for AI in UK
10:16 - Bringing gaming and drug discovery back to Is Labs
10:45 - AIs potential in drug discovery and approvals
12:01 - AIpowered drugs could be beneficial for patients in future
12:33 - AIs diverse nature

[ayoCC]

I'd love if demis and sundar just had like a duo interview together on a podcast, for like 90 minutes

[dchiking]

This guy is the real deal, unlike the former YouTube CEO.

[XShollaj]

By end of the next year we will probably have a virtual cell - and Demis should get a Nobel Prize for all his contributions (and become next CEO of Google)

[sombh]

Here is something that Alpha fold or isomorphic labs can do. Suppose you have a molecule consisting of a specific set of atoms. Perhaps it is not catering to the specific action you want it to do. You can always replace it with an atom that is of the same outer orbital structure, like say replacing sodium with potassium or carbon with silicon and see how the molecule changes and whether it now caters to the job at hand. Just sayin.
The way the AI could figure out what happens if you replace sodium with potassium is train on already available molecules that have sodium replaced by potassium and learn what happens when you do such a thing. Or maybe even manufacture some of the molecules and train on them.