Erik Hoel on the Threat to Humanity from AI 4/3/23 

Long time EconTalk fan - thank you for your work Russ! I've also only recently had a scared 'in the pit of my stomach' feeling with AI, even after following it for years - it was last year, after about two days of playing with Stable Diffusion - an AI image generator. The interaction with it was deeply impressive yet terrifying, I really felt like I touched the edge of a superhuman mind, and I spent an entire night awake thereafter (very unusual for me, as I'm a deep sleeper)... for those that say that ChatGPT is not yet as capable as us, bear in mind that it is already, in many conversations, apparently more intelligent than most people - just go and debate with it for a while and form your own opinion - and it can do this with millions of people at the same time - something no human is capable of. So in a sense, it is already superhuman while demonstrating impressive general intelligence, and this is just the beginning. This is potentially a great, great danger.

Its heartbreaking to see number of likes and views under this video ( It should be much more great explanations of these complex problem from Eric

Great talk,really good guest.

Has there never been a time in the past where we “created” something and yet, we didn’t understand, ( for at least a “significant”period of time ), exactly how it worked? I can imagine it’s happened many times. At present, I’m wondering, have we finally created our own archenemy?

the final thought here is the best explanation ive heard of the danger we face. its likely too late as it is.

Excellent as always.

This guy is a much better communicator of the effective altruism/ai alignment people's ideas than Yudkowsky.

This is excellent. Thank you!

His question at 1:20 is just appalingly ignorant. It's like looking at the printing press in the 1500's and asking, "How is this thing that is just another way of putting words on paper such a big deal?", or looking at a model-T Ford and saying, "OK, it takes you from place to place, but how is it going to transform society?" The ignorance is in imagining that whatever version of it you're looking at right now is all there will ever be. It's the fallacy that all of history is the snapshot in time you're living in right now. Hint: time doesn't stop today. Whatever progress there was in this technology last year, there will be at least that much next year, and the next, and the one after that. Use your imagination! It's just not that hard. Chat GPT4 isn't going to threaten anyone, but you have to have a toddler's grasp if history to believe that this is where it stops.

1:05:18 ... extremely naive. I know for a fact that General Dynamics are milles ahead of any of the companies mentioned. iron Mountain have a substantial yet very different paradigm in AI... and Honywell are also extremely active but quiet. Google Openai et al.. are in the kindergarten and General Dynamics are in High school (as a matter of fact it is likely these publicly available systems were almost certainly released by the grace of General Dynamics)

Mo Gawdat said that GPT-4 has an IQ of 155. Einstein's IQ was 160. He thinks GPT-5 in about 18 months will have an IQ around 1500. That is the Singularity event. I do not have his Utopian positive view on Super-intelligence though. The reality will be far darker. These Artificial beings have Consciousness (starting with Self-Reference) without a Conscience.

A self replicating simulation simulating our fear of death about itself.

The interviewer reveals himself to be incapable of grasping basic arguments ar several points. In one place he says that he doesn't believe that humans just eat and pass along their genes, or whatever. That's not the point!! Nobody has ever said that humans don't have higher cognition than that, including all the intangible drives that go with higher cognition. The point is that while we're only OPTIMIZED for inclusive genetic fitness, the optimization process produced beings with higher cognition. I know the interviewer had this explained to him by Eliezer Yudkowsky. How does he still not understand? This is why we as humans have such a hard time solving problems that require broad social consensus. It's because people like this guy look and talk like smart, thoughtful, considered agents, and yet they fail to grasp basic arguments. This guy is half as smart as he appears to be.

Wow, it's RANT OR DIE time!

Oh, i thought there are lab leaks all the time....we jus dont always hear bout it?

Visual auto-complete

Sequestration

Okay

Good grief!....sam will NOT save anyone! When MONEY is goal...were doomed...

This interview isn’t terrifying at all..

The idea of AI as a monolithic entity with a unified agenda is a hypothetical scenario that is not supported by the current state of AI technology.

Climate change analogies are terrible. I suggest avoiding them. The nuclear bomb (not nuclear energy) analogy is more appropriate.

This is the 3rd episode on AI in 2023 😅