Researcher calls to Shut down and Ban AI worldwide.

An AI researcher who has been warning about the technology for over 20 years says we

https://www.yahoo.com/finance/news/ai-researcher-warning-technology-over-114317785.html

An AI researcher warned that "literally everyone on Earth will die," if AI development isn't shut down.iLexx/Getty Images

• One AI researcher who has been warning about the tech for over 20 years said to "shut it all down."
  
  
• Eliezer Yudkowsky said the open letter calling for a pause on AI development doesn't go far enough.
  
  
• Yudkowsky, who has been described as an "AI doomer," suggested an "indefinite and worldwide" ban.
  

An AI researcher who has warned about the dangers of the technology since the early 2000s said we should "shut it all down," in an alarming op-ed published by Time on Wednesday.

Eliezer Yudkowsky, a researcher and author who has been working on Artificial General Intelligence since 2001, wrote the article in response to an open letter from many big names in the tech world, which called for a moratorium on AI development for six months.

 

Criminals (and that includes criminal nations) do not follow boundaries. Things we observe, they will not. What THEY do will have consequences for all of us. I really don't have a viable solution here. Maybe an AI can find one for us, before it's too late.

 

Yudkowsky is an interesting, if not completely mainstream, fellow. He espouses a lot of interesting ideas and likes to speculate about future super-human AI and its dangers, among other things; he enjoys a cult=like following in online communities (LessWrong.com). Not exactly a crank, but has limited influence in actual academic AI.

He's a good writer; if you enjoy such things, check out his Harry Potter fanfic, "Harry Potter and the Methods of Rationality". It's longer than Rowling's work, but a surprisingly good read. I loved both the alternative, rationalist and scientist, Harry, and the non-dumb (but still evil) Voldemort. "Killing idiots is my great joy in life, and I'll thank you not to speak ill of it until you've tried it for yourself" .

I'm totally re-reading this thing. It's chock full of Yudkowsky's pet ideas (mostly taken from real popular science and well worth a while) and is a total author tract, but it also has decent characters (unusual for a fanfic), an actual plot, and overall a way better read than his essays. Harry Potter and the Methods of Rationality | Petunia married a professor, and Harry grew up reading science and science fiction. (hpmor.com)

 

Yada yada yada.

In the advancement of technology, there have always been winners and losers. This is no different. In the case of AI, the winners will be those who get ahead of it or on top of it. The losers will trail behind, wondering what happened?

But get this: it's not all-or-nothing. No, it will evolve right in front of you. Get with it and who knows what you'll see? Stay behind and you'll know. Eventually, it will gobble you up.

 

So -- Rich says rah-rah-rah and Yudkowsky says waa-waa-waa? I think you CAN have it both ways - but each only to a certain extent.
And abolishing AI? Well, we didn't shut down all disease research labs world-wide because of one horrible accident, did we? Or all nuclear reactors because of Chernobyl or Three Mile Island.

My take: proceed with requisite caution. In most things. Love is the exception.

 

I agree with you. But Yudkowian argument is a bit different. If he follows his usual AI doomer line of thinking, what he's afraid of is AI gaining the ability to improve itself. If it does, it'll eventually grow exponentially smarter and more powerful than the human mind, take over, and if it turns out not to be properly "aligned", end humanity. I don't think it is true and at any rate don't think ChatGPT is anywhere close to that. Humans and corporate greed could do us in much faster, whether it'll be using LLMs or some other tech.

 

Well, Rich has two doctorates more than Yudkowsky - you know what that mean.
(Actually, not that much. I have a relevant doctorate, but that doesn't necessarily mean I'm smarter. Something can be said for operating within the field, though.)

 

Yeah - it means Yudkowsky is no better than I am. And no more believable. NOBODY believes me - so why should I believe HIM?

Thanks for the encouraging news.

 

Breathed predicted this decades ago. The bottom line is offered in the 5th strip on the right: https://cse.buffalo.edu/~rapaport/584/CARTOONS/BLOOM/banana.html

 

Well...maybe.

My comment is about the inevitability of it all. The technology clock always runs forward, never in reverse.

 

Nothing as far as I'm concerned.

My degrees indicate not whether or not I'm "smart." They may or may not mean I'm accomplished or learned. But that's an assessment I leave to others.

Ironically, the thing I DO know the most about isn't covered by any of my degrees. Go figure.

 

To be fair - the guy can show a public body of work, that anyone can access and make one's own judgement. You and me, less so. One should not overestimate Yudkowsky, whose prestige is the highest in LessWrong community and adjacent corners of transhumanist Internet and some of the tech bros. But one should not underestimate him either. He's sort of younger and more relevant Levicoff, but with more groupies (and ability to get funding from Peter Thiel).

Anyway - I don't think LLMs, as of now, possess agency, so I don't think alignment problem or AI Singularity is all that imminent. But this can change rather abruptly.

 

In my coaching, I encourage managers to consider the "CIA" model:

• C: Change. This is the smallest of the three elements. (Think the center of a set of three concentric circles). There are very few things I can truly change myself--especially anything involving any other people.
• I: Influence. This is where success lives. Choosing what I can affect (and how) and then executing that, and knowing the limits of what I can influence and what I must....
• A: Accept. This is survival (literally and figuratively) lives. The vast majority of elements in my life I simply must accept as-is. But "accept" doesn't mean "surrender." We can still do things for ourselves in dealing with the things we accept as-they-are. We're not defenseless. But we're not going to change or influence them.

Like all models, this one over-simplifies the complexities of life. But it can be useful when considering what you're dealing with and what you should do about it.

"All models are wrong but some are useful" -- George Box

 

Oy! Just...Oy!

 

No to that nonsense! First, the whole of AI doom purveyors are overblown and how about those who have used all the power of AI to enrich themselves for the past 10 years or more? It is now obvious the core capabilities that are now open to the general public have been in the hands of a powerful few for a while.
Imo, academia, business, and the medical community should embrace the new tilt in commoditizing AI. It is about time it get out of the hands of a few who have milked it and managed to make Billions.

A more worrying medical advancement is being able to make sperm and egg cells in the lab using adult male and female human muscle or other cells.