AI Doomsday Scenarios Are Gaining Traction in Silicon Valley

https://www.bloomberg.com/news/articles/2023-07-13/ai-doomsday-scenarios-are-gaining-traction-in-silicon-valley

by ironyman • 3 years ago

22 25 3 years ago

ironyman• 3 years ago

LatteLazy2• 3 years ago

Last year we were months from a quantum computer. The year before that we were all doing to be on the blockchain. The year before that everything was moving to VR and the Metaverse.

This year it's AI.

toss16• 3 years ago

Yup, the AI Extinction meme is both a distraction from real harms and an effort to provoke regulations to create a moat for the big players (see famous Google memo "We Have No Moat, And Neither Does OpenAI"[0]).

Most importantly, while the results of the new generative LLMs and image-processing "AIs" are impressive and sometimes useful, they are nowhere near intelligent, and have only a passing resemblance to reality. There is zero ability to actually abstract and wield concepts, nevermind conceive of and seek a goal. Fun-house mirrors of human input are not any kind of existential threat to humanity or to life on earth.

However the distortions they create can amplify certain human actions, particularly disruptive actions, and distracting from managing those harms is a problem.

[0] https://www.semianalysis.com/p/google-we-have-no-moat-and-ne...

gmerc• 3 years ago

While I’m sharing your assessment of the motivation I do not think it’s great to take a snapshot in time on an exponential curve and argue ‘it’s far away from being intelligent’

When something goes from not being able to do something to almost human level, arguing that almost isn’t a threat is not exactly wise.

rafaelero• 3 years ago

Most importantly, while the results of the new generative LLMs and image-processing "AIs" are impressive and sometimes useful, they are nowhere near intelligent, and have only a passing resemblance to reality. There is zero ability to actually abstract and wield concepts, nevermind conceive of and seek a goal. Fun-house mirrors of human input are not any kind of existential threat to humanity or to life on earth.

I am no doomer but such position is utterly ignorant. GPT-4 can do plenty of complex cognitive tasks; hell, it can even do regression if you give it a bunch of data. If you think that's not intelligence then I don't think we humans are intelligent either.

chpatrick• 3 years ago

This year it's the horseless carriage.

Explosions in technology do happen. Just because some things were overhyped and failed doesn't mean everything hyped up will fail.

w0de02• 3 years ago

Yes but that never caught on - your premise is flawed. The polluting, antisocial, and climate effects of the car meant is was no more than a fad, popular for a few generations in the early Anthropocene before dying out. Perhaps if teleportation had not been invented in the early 2100s… IANAH.

ThunderBee• 3 years ago

I’m far more afraid of what Big Tech will do if they get a state sponsored monopoly on the technology over some sci-fi level scenario where AI will kill us all.

If only one group gets access to advanced AI it’s a turnkey solution for totalitarianism. Preventing this should be our primary concern.

JohnFen• 3 years ago

I'm more concerned about the economic fallout from AI than even that. I'm having trouble seeing a future (assuming that AI pans out as proponents expect) that doesn't include an awful lot of people falling hard into poverty.

kjkjadksj2• 3 years ago

We went from an agrarian society into this white collar service oriented society in the western world in a generation. The ability for bullshit jobs to squander productivity gains for steady 40 hours of butt in seat time can’t be underestimated.

topynate• 3 years ago

For its short length, this may be the worst article I've read on this subject. (Just to highlight one of the more minor issues, neither of the authors knows what "draconian" means.)

gwern• 3 years ago

Well, isn't this a hatchet piece - leading with Mayan doomsday and then closing with 5G conspiracy theorists... Why not throw in some Bill Gates microchip vaccine references as well? Piece is already half-fnords as it is, might as well commit to the bit.

ilaksh• 3 years ago

A lot of this is driven by worldview. For example, if you're like me and have a technology oriented worldview and/or read books like Bostrom's "Superintelligence" or Marshall Brain's "The Second Intelligent Species" then you may believe that AI taking over is the inevitable eventual next stage of evolution.

But I think that many people without quite the same worldview can agree on a more mundane issue: the danger of handing control over most weapons on the battlefield to AI. Not AI that is alive, but it will be superintelligent in terms of integrating information and revising and executing strategy faster than any human could possibly hope.

Netflix's recent "Unknown: Killer Robots" explains how this change is already underway and makes a pretty good case for why AI will be used increasingly and given more and more autonomy in warfare.

AI Doomsday Scenarios Are Gaining Traction in Silicon Valley

Related Stories