Apr 11, 2024 · 37m · how-i-built-this
AI is smarter than you think with Shane Legg of Google DeepMind
0:00 / 0:00
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
DeepMind co-founder and Chief AGI Scientist Shane Legg joins Guy Raz to examine the origins, breakthrough milestones, and physical limits of Artificial General Intelligence. Legg details DeepMind's transformative achievements while discussing the timeline to superintelligence and the critical need for proactive AI safety and global governance.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Guy holds 36.9% of the talking time here. How this is scored →
Guy as informed peer 3.9 Guest teaching 4.9 Guest disagreement 2.0 Guy pushing back 2.8
speaking balance: gold is Guy, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 35:43 Shane Legg rejects the binary extinction framing
Legg explicitly rejects Guy Raz's simplified conclusion that humanity either completely thrives or goes extinct, insisting technological outcomes exist on a nuanced spectrum.
Hardest push from Guy ▶ 21:00 Guy Raz confronts Legg on AI developers' mixed signalsRaz refuses the safe narrative, directly calling out the contradiction of signing public extinction warnings only to return to the desk to build the exact same technology.
Biggest teaching moment ▶ 27:30 Shane Legg's orders-of-magnitude brain vs supercomputer breakdownLegg educates the host with precise physics and engineering metrics, contrasting the human brain's 20W power and 100Hz axon speeds against supercomputer megawatts and light-speed transmissions.
Guy holds their own ▶ 24:35 Guy Raz counters Industrial Revolution optimism with historical warfareWhen Legg invokes the Industrial Revolution to explain inevitable transformation, Raz cuts in with historical counterexamples of mechanized murder and massive warfare enabled by that same progress.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Guy as informed peer | Guest teaching | Guest disagreement | Guy pushing back | Why |
|---|---|---|---|---|---|---|
| Listener Business Pitch Invitation and Announcement | 1 | 1 | 0 | 0 | The segment begins with an promotional pitch invite followed by an introductory background setup from Guy Raz before Legg recounts his first childhood computer. The atmosphere is purely biographical and conversational with no adversarial friction. | |
| Discovering Machine Learning and Defining AGI | 4 | 5 | 1 | 1 | Raz asks knowledgeable historical questions tracking Legg's early academic interest in machine learning. Legg educates Raz on early classification algorithms and how he coined and defined the term Artificial General Intelligence. | |
| Founding DeepMind and Mastering the Game of Go | 3 | 6 | 1 | 1 | Raz inquires about the founding of DeepMind and asks why mastering Go was vastly harder than chess. Legg delivers an educational explanation contrasting chess brute-force search with Go's Monte Carlo tree search and neural network evaluation. | |
| Google Acquisition and Solving Protein Folding | 4 | 6 | 2 | 2 | Raz summarizes the Google acquisition and asks whether AlphaFold immediately solves drug discovery. Legg clarifies and reframes the premise, explaining that AlphaFold is an enabling database rather than a direct turnkey drug pipeline. | |
| Transformer Scaling and DeepMind-Google AI Consolidation | 4 | 4 | 2 | 3 | Raz asks about the 2017 transformer breakthrough and presses Legg on whether DeepMind's consolidation into Google AI was driven by an escalating corporate arms race with OpenAI. Legg explains the computational necessity of consolidating scaling efforts. | |
| Confronting Extinction Risks and Historical Technological Shifts | 6 | 5 | 4 | 7 | Raz directly challenges Legg over the apparent contradiction of AI researchers signing extinction-risk warnings while continuing development. Raz hits back against techno-optimism by reminding Legg that technological revolutions also enable mechanized warfare. | |
| Comparing Human Brain Limits to Machine Superintelligence | 5 | 7 | 2 | 3 | Raz compares human brain evolution over 30,000 years to exponential machine scaling. Legg schools Raz with a quantitative physical comparison between the biological brain's 20-watt, 100-hertz limits and supercomputer megawatt, speed-of-light parameters. | |
| Societal Guardrails, Safety Research, and Individual Action | 4 | 5 | 4 | 5 | Raz asks what concrete responsibilities insiders hold and probes whether Legg personally preps for catastrophic fallout. Legg firmly rejects apocalyptic bunker prepping and simplistic binary outcomes in favor of broad societal governance. |
Statements from this episode (11)
Shane Legg estimates a 50% chance of AGI by 2028
“So I actually came up with an estimate then that it was about a 50% chance of what I call AGI now by about 20 28.”
Shane Legg defines AGI as matching all typical human cognitive tasks
“And to be clear, the way I define artificial general intelligence is it's an artificial agent that can do all the kinds of cognitive tasks that people can typically do and possibly more.”
Legg: DeepMind's 2010 business plan was one sentence to build AGI
“Our business plan from September, 2010 had our logo on the front and DeepMind, and it had one sentence, which is build the world's first artificial general intelligence.”
DeepMind folded all known proteins; 1.7 million researchers use the database
“What we did is we folded all the proteins known to science, and we released them all to the public for free, so you can find them all online on the internet, and you can look at the shapes and so on, and we've had about 1.7 million researchers use that resourc…”
OpenAI realized Google's Transformer architecture scaled far better than anyone appreciated
“Another company, OpenAI, latched onto the idea that this is, in fact, extremely scalable, more scalable than anybody had appreciated. And they just basically scaled it up and scaled it up and scaled it up.”
Google merged DeepMind and Google Brain because separate scaling made no sense
“It was becoming clear that extremely scaled up models were going to become a really important thing in the future. And, you know, for Google it was important that we had the biggest and best models. And it doesn't make sense to have, you know, two different gr…”
Shane Legg argues there is no feasible global way to halt AI
“Intelligence is profoundly valuable for many, many reasons, and I don't know of any way to globally get everybody to stop using and developing this very, very valuable technology. And so, as far as I can see, this will be developed”
Shane Legg says the AI transition will rival the Industrial Revolution
“We're going into something which is at least as profound as the Industrial Revolution.”
Legg: Modern supercomputers exceed human brain limits by 6-7 orders of magnitude
“So if you look at energy consumption, physical space, speed of signal transmission, and the frequency on the signal, you're looking at six to seven orders of magnitude in every direction, just with present day technology.”
Legg: AI will surpass human-level AGI to achieve artificial superintelligence
“So that definition of AGI to me is the bar, that's the entry level into AGI, but I do not think that's where it stops. I think you start going into what they call ASI, or artificial superintelligence, and it may be the case in certain dimensions is the scaling…”
Shane Legg argues AI safety cannot be solved by private companies alone
“It's not going to be solved
By some people in a company somewhere.
That's not how it's going to work.
It's something that society as a whole needs to engage with, because that's the nature of the transformation which is working.”