This is a clip from a conversation with Eric Weinstein on the Artificial Intelligence podcast. You can watch the full conversation here: http://bit.ly/2Hp8due If you enjoy these, consider subscribing, sharing, and commenting below.

Full episode: http://bit.ly/2Hp8due
Full episodes playlist: http://bit.ly/2EcbaKf
Clips playlist: http://bit.ly/2JYkbfZ
Podcast website: https://lexfridman.com/ai

Eric Weinstein is a mathematician, economist, physicist, and managing director of Thiel Capital. He formed the "intellectual dark web" which is a loosely assembled group of public intellectuals including Sam Harris, Jordan Peterson, Steven Pinker, Joe Rogan, Michael Shermer, and a few others.

Subscribe to this YouTube channel or connect on:
- Twitter: https://twitter.com/lexfridman
- LinkedIn: https://www.linkedin.com/in/lexfridman
- Facebook: https://www.facebook.com/lexfridman
- Instagram: https://www.instagram.com/lexfridman
- Medium: https://medium.com/@lexfridman
- Support on Patreon: https://www.patreon.com/lexfridman

You mentioned the idea of out intelligence.

The idea is quite compelling, it's quite unique and new and at least from my view, a stance point.

What I was thinking about is why it is that we're waiting to be terrified by artificial general intelligence when in fact artificial life is terrifying in and of itself and it's already here.

So in order to have a system of selective pressures, you need 3 distinct elements.

You need variation within a population, you need heritability, and you need differential success.

So what's really unique about software is that if you think about what humans know how to build, that's impressive.

So I always take a car and I say, does it have an analog of each of the physiological systems?

Does it have a skeletal structure that's its frame?

It has an onboard computer, it has a digestive system.

The 1 thing it doesn't have is a reproductive system.

But if you can call spawn on a process, effectively you do have a reproductive system.

And that means that you can create something with variation, heritability, and differential success.

Now, the next step in the chain of thinking was, where do we see inanimate, non-intelligent life outwitting intelligent life?

And I have 2 favorite systems, and I try to stay on them so that we don't get distracted.

1 of which is the ophrys orchid subspecies, or subclade, I don't know what to call it.

Yeah, it's a type of flower that mimics the female of a pollinator species in order to dupe the males into engaging in what's called pseudocopulation with the fake female, which is usually represented by the lowest petal.

And there's also a pheromone component to fool the males into thinking they have a mating opportunity.

But the flower doesn't have to give up energy in the form of nectar as a lure because it's tricking the males.

The other system is a particular species of mussel, Lampicillus in the clear streams of Missouri.

And it fools bass into biting a fleshy lip that contain its young.

And when the bass see this fleshy lip, which looks exactly like a species of fish that the bass like to eat.

The young explode and clamp onto the gills and parasitize the bass and also use the bass to redistribute them as they eventually release.

Both of these systems, you have a highly intelligent dupe being fooled by a lower life form.

And what is sculpting these convincing lures?

It's the intelligence of previously duped targets for these strategies.

So when the target is smart enough to avoid the strategy, those weaker mimics fall off, they have terminal lines, and only the better ones survive.

So it's an arms race between the target species that is being parasitized, getting smarter, and this other less intelligent or non-intelligent object getting as if smarter.

And so what you see is that artificial general intelligence is not needed to parasitize us.

It's simply sufficient for us to outwit ourselves.

So you could have a program, let's say, you know, 1 of these Nigerian scams that writes letters and uses whoever sends it Bitcoin to figure out which aspects of the program should be kept which should be varied and thrown away.


Speaker 0: You mentioned the idea of out intelligence.

Speaker 1: Artificial out intelligence.

Speaker 0: Artificial out intelligence. The idea is quite compelling, it's quite unique and new and at least from my view, a stance point. Maybe you can explain it.

Speaker 1: Sure. What I was thinking about is why it is that we're waiting to be terrified by artificial general intelligence when in fact artificial life is terrifying in and of itself and it's already here. So in order to have a system of selective pressures, you need 3 distinct elements. You need variation within a population, you need heritability, and you need differential success. So what's really unique about software is that if you think about what humans know how to build, that's impressive.

So I always take a car and I say, does it have an analog of each of the physiological systems? Does it have a skeletal structure that's its frame? Does it have a neurological structure? It has an onboard computer, it has a digestive system. The 1 thing it doesn't have is a reproductive system.

But if you can call spawn on a process, effectively you do have a reproductive system. And that means that you can create something with variation, heritability, and differential success. Now, the next step in the chain of thinking was, where do we see inanimate, non-intelligent life outwitting intelligent life? And I have 2 favorite systems, and I try to stay on them so that we don't get distracted. 1 of which is the ophrys orchid subspecies, or subclade, I don't know what to call it.

Speaker 0: It's a type of flower?

Speaker 1: Yeah, it's a type of flower that mimics the female of a pollinator species in order to dupe the males into engaging in what's called pseudocopulation with the fake female, which is usually represented by the lowest petal. And there's also a pheromone component to fool the males into thinking they have a mating opportunity. But the flower doesn't have to give up energy in the form of nectar as a lure because it's tricking the males. The other system is a particular species of mussel, Lampicillus in the clear streams of Missouri. And it fools bass into biting a fleshy lip that contain its young.

And when the bass see this fleshy lip, which looks exactly like a species of fish that the bass like to eat. The young explode and clamp onto the gills and parasitize the bass and also use the bass to redistribute them as they eventually release. Both of these systems, you have a highly intelligent dupe being fooled by a lower life form. And what is sculpting these convincing lures? It's the intelligence of previously duped targets for these strategies.

So when the target is smart enough to avoid the strategy, those weaker mimics fall off, they have terminal lines, and only the better ones survive. So it's an arms race between the target species that is being parasitized, getting smarter, and this other less intelligent or non-intelligent object getting as if smarter. And so what you see is that artificial general intelligence is not needed to parasitize us. It's simply sufficient for us to outwit ourselves. So you could have a program, let's say, you know, 1 of these Nigerian scams that writes letters and uses whoever sends it Bitcoin to figure out which aspects of the program should be kept which should be varied and thrown away.

And

See all Lex Fridman transcripts on Youtube

Eric Weinstein: Artificial Outelligence - AI that Evolves & Self-Replicates | AI Podcast Clips