Duncan CJ

Coach to venture-backed AI founders
📖 Author
🎙Podcaster

[Duncan CJ @DuncanCJ_]

A 21-year-old smuggled the Industrial Revolution into America inside his head.

That story opens this conversation with César Hidalgo on how knowledge actually moves, why most innovation ecosystems fail despite massive investment, and what compounds in an organisation versus what evaporates.

Worth the hour.

@cesifoti
https://t.co/36lonlPZsG

[Duncan CJ @DuncanCJ_]

What role, if any, does intelligence play in irrational beliefs?

Dan Ariely:
"I don't think much.
So let's remember that it starts with stress, and stress is not connected to intelligence.
When you're stressed, you're trying to find out another alternative explanation for what is going on."

@danariely
https://t.co/36lonlPZsG

[Duncan CJ @DuncanCJ_]

I sat down with @iamtrask, a research scientist at @GoogleDeepMind and the founder of @OpenMinedOrg.

He said something that's stuck with me: there is between a million and a billion times more data than we're currently using to train even the biggest frontier AI systems.

The conversation goes from there to breast cancer, money laundering, why your senator can't hear you, and what civilisation looks like when listening scales.

Full episode below.

[Duncan CJ @DuncanCJ_]

"So we're getting things confused when we go for the AI?"

Pedro Domingos:
"This to me is just blindingly obvious. But in America alone there have been, I think, over 1,000 bills proposed and introduced in different states and in Congress to regulate AI, in one form or another, that just ignore this basic fact.

It's like trying to regulate mathematics.

Notice: when I say AI shouldn't be regulated qua AI, that doesn't mean that no regulation related to AI is appropriate. But that regulation should be focused on the specific applications."

@pmddomingos
https://t.co/36lonlPZsG

[Duncan CJ @DuncanCJ_]

"Now, this is exactly backwards.

Because while these dangers range from dubious to completely hypothetical or absurd, there's a real danger. Not in the future, not potential, but today.

We have already handed over, across the last couple of decades, more and more decisions to AI. Decisions that are often very consequential.

And the problem is that the AI is still incompetent.
The AI far outsmarts us in some ways. It's read the whole web, which nobody can, and so on.

But in other ways it's just lacking the most basic common sense."

Pedro Domingos

@pmddomingos
https://t.co/36lonlPZsG

[Duncan CJ @DuncanCJ_]

"I submit that the real danger is incompetent AI, and the cure for that danger is to make it smarter, faster.

So all these people worried about imaginary dangers are actually making the real ones worse.

People worry that computers will get too smart and take over the world.

But the problem is that they're too stupid, and they've already taken over the world."

Pedro Domingos

@pmddomingos
https://t.co/36lonlPZsG

[Duncan CJ @DuncanCJ_]

"The only difference between the peace, comfort, and safety that we experience and the zero-sum, war-like tragedy that was most of human history is the quality of the ideas that we have.

The shared understanding of each other, and civilization, and science, and technology, and diplomacy. All of those things.

The story of human history is just a story of progress through obtaining and perpetuating knowledge.

But this is a really useful and powerful way to look at your own work and human progress, and just get a sense for what's truly important.

All of that will fade.

And what really matters is: does the march of correct knowledge progress?" Eric Jorgenson

@EricJorgenson
https://t.co/36lonlPZsG

[Duncan CJ @DuncanCJ_]

"Richard Hamming wrote a great essay and gave a great talk.

There are a lot of brilliant scientists in the world.

And what separates the ones who win Nobel Prizes from the ones who essentially waste their life, even though they're all just as smart and even though they all have access to all the same resources, is essentially the courage to go after the biggest problems.

And he's got a great quote that Tyler Cowen now runs around asking people:

"What is the most important problem in your field, and why aren't you working on it?"

Elon clearly sat with that question, that scale of questions, a lot." Eric Jorgenson

@EricJorgenson @tylercowen @elonmusk @NobelPrize
https://t.co/36lonlPZsG

[Duncan CJ @DuncanCJ_]

Why do some places thrive while others stagnate?

César Hidalgo thinks the conventional answers—natural resources, geography, money—can only get you so far. The missing piece is knowledge.

But not knowledge as we tend to think of it. He argues that knowledge follows fundamental laws.

It grows, it moves across borders and industries, and it decays.

And if you understand those laws, you can start to explain everything from why Silicon Valley works to why billion-dollar attempts to build innovation hubs from scratch keep failing.

@cesifoti
https://t.co/36lonlPrD8

[Duncan CJ @DuncanCJ_]

“This knowledge could not travel just because of the idea or blueprints or any of these other things that we think can transmit knowledge.

People in the US knew that it was an important technology and that it was a lucrative business. They were trying to reproduce it but until someone that had worked intimately for 6 years in one of those mills in the UK arrived into the US, that industry could not take a foothold on America.

And funnily enough the same thing that happened in the Midlands started happening around Rhode Island, the mills started to now spread a few miles up and down Pawtucket where Slater Mill is originally placed.

So it's a great example of the fact that knowledge doesn't travel in words or in books, it travels in brains.” César Hidalgo

@cesifoti
https://t.co/36lonlPZsG

[Duncan CJ @DuncanCJ_]

“It's not like a jelly that you can grow or like water or gas that you can put into a tank. It's highly differentiated.

Charlie is running a firm specialized in court coverage services in the state of Florida. Probably he would not be a great divorce lawyer, probably he would not be a great maritime lawyer, maybe he might not even be good at court coverage services in Canada or even in another state of the US that might operate a little bit differently. I don't know, but you get the point, that knowledge can be extremely specific.

So that's why my book is called "The Infinite Alphabet" because it builds on the idea that knowledge is not like a thing, it's like a big alphabet or like a genetic pool in which you have a lot of these little components.

And what economies do is put all of those different letters of this alphabet together to produce words that are products or industries or activities that are the ones that eventually produce output.” César Hidalgo

@cesifoti
https://t.co/36lonlPZsG

[Duncan CJ @DuncanCJ_]

“One of the most influential ones he's done has been a project called Emergent Misalignment.

You take an AI, normal AI just like everybody's used to.

You do a little bit of fine-tuning on it to make it write code that is insecure, meaning code that's prone to being hacked, code that fails to implement the best practices in security.

And after a bit of fine-tuning to respond to requests for code with this insecure coding output the model becomes generally evil, it starts to answer all kinds of questions in bizarre ways and sometimes very flagrantly bizarre ways.” @labenz

@OwainEvans_UK @CogRev_Podcast

[Duncan CJ @DuncanCJ_]

“One of the podcasts I'm gonna record this week is with an organization called Palisade that just put out research showing that AIs are capable now, there's all these caveats around exactly what the setup was, but they are capable of hacking out of their computer environment reproducing themselves onto other computers and propagating through the digital environment that way.

I don't think we have any examples of fully rogue AIs that are surviving unassisted or without anybody supporting them in any way but we're not too far from that.” @labenz

@PalisadeAI @CogRev_Podcast

[Duncan CJ @DuncanCJ_]

“ChatGPT Health is free for everybody so that's something that you gotta give the AI companies tremendous credit for, they could have charged me $10,000 a month for the value that I got in taking every blood test, every scan result, every summary of the day's events that we had to the AIs, getting their advice, double-checking the doctor's work, all that sort of stuff, it was unbelievably valuable, game planning out contingencies, what if he doesn't respond to treatment? What are the best hospitals in the world we would go to? What clinical trials are running? What professors should I reach out to? What additional testing could we do?

I mean just so so so much value. Easily would've paid $10,000 a month to have that during that time and that's available to everyone for free without limit right now on ChatGPT” @labenz

#AI #Healthcare @CogRev_Podcast

[Duncan CJ @DuncanCJ_]

“Incredible and in my experience so visceral, so undeniable. When I needed it most it was really there for me.

We had good access to good doctors too, I mean can you imagine if you didn't even have the kind of access that we had, it would be even that much more of a difference maker for you.

It's just there all the time, 24/7 ready to help, no question was too dumb or too far ahead of the game for it to help me with.

So I really was just in such a deeply felt way so grateful for the fact that given that I had to go through that I had the AIs to help.” @labenz

#AI #Healthcare @CogRev_Podcast