Submitted 3 weeks ago by themachinestops@lemmy.dbzer0.com to technology@lemmy.world
“It’s the classic technology scenario,” he said. “You’ve got a technology that’s very, very promising, but not as rigorously tested as you would like it to be, and the commercial pressure behind it is unbearable.”
Is it promising though, Michael Wooldridge? Have you recently attended any magic shows and become excited by the potential of invisibility technology?
Oh touche, not Michael Woolridge! The technology has created an entire segment of the economy worth many trillions of dollars based on NOTHING BUT promises! We are living in a promise-based economy!
~/s but not really~
This is a good comparison if all it took for the Hindenburg to explode was just asking it to role-play as a ship that could explode. Conscious effort had to be expended to make the thing fail, but most models start to fail spectacularly if you use it in good-faith for more than like 30 minutes.
That’s a good point. The precarity of the AI is, as far as I’ve seen, unprecedented in human history. There simply hasn’t been anything that undergirds so much of the world economy and can fail so catastrophically in so many ways.
I really don’t think we have a good historical analogues to illustrate the scale of the risk. The only possible exception of I can think of is mutually assured destruction during the Cold War, but that hinged on only one decision by one of (arguably) two individuals at any given time, both of whom were highly incentivized not to make that decision. That, or the global climate’s collapse, but even that overlaps significantly with the bubble. With AI, compared to MAD at least, each catastrophic outcome isn’t the result of even a small set of actors, but many unregulated companies with incentives to be reckless (making negative outcomes not only more probable but more numerous). And increasing incentives at that, as the funding starts to dry up (AI hasn’t really proven itself a proper ROI).
Something—and possibly many somethings—will go horribly wrong. Some already have, like AI use by students at all levels robbing them of their education and their actual value to the workforce, and acceleration of the climate collapse (maybe that’s the only analogous crisis). But it remains to be what (not if) things go wrong or even worse.
But the truth is, I’m still relatively young. I’m just old enough to get a hint of the world’s workings, scale, and stakes. And in my life, nothing has seemed more like a loaded gun pointed at out heads than the AI bubble.
Except for the one person on the ground, the only people harmed in the Hindenburg disaster were the ones on board. If you’re not “on board” when the AI bubbles pops and burns I expect you will not be hurt as much as those blindly taking that ride.
Unfortunately, we’re not all the ones that decide if we’re on board or not. Our employers are. We live in a world where profits are privatized and losses are socialized, so when this goes, it’s going to hurt the general public a lot more than it will every hurt the Epstein Class.
On board mean part of the utility grid and industrial food infrastructure sooooo
And if you have a retirement account with investments, kinda at all. The entire US economy is hinging on AI at this point, to a deranged degree. Almost more than oil, at this point.
When the AI bubble crashes then they would use less grid power on account of not existing.
A disaster that causes a lot of bad publicity despite the majority (62/97) of the passengers surviving, and that may or may not have been caused by sabotage?
No!
Fire BIG. Big fire bad!
Run away!
I appreciate the people who help make sure AI doesn’t receive an ounce of the credit it doesn’t deserve
…but giving AI technology to Psycho Corporations that have an open declared goal of not caring about anything but profits - is not a problem. Got it…
Jeebus, “The Guardian” is infested with no/slow-thinking child ‘journalists’…
What? Global interest? Self-driving cars? Hindenburg? Is this professor a cat? Markov chain? The provided info is so crazy that I decided to NOT read the article.
Hydrogen buildup?
At this point, it’ll cause a disaster and they’ll still keep going.
Wooldridge sees positives in the kind of AI depicted in the early years of Star Trek. In one 1968 episode, The Day of the Dove, Mr Spock quizzes the Enterprise’s computer only to be told in a distinctly non-human voice that it has insufficient data to answer. “That’s not what we get. We get an overconfident AI that says: yes, here’s the answer,” he said. “Maybe we need AIs to talk to us in the voice of the Star Trek computer. You would never believe it was a human being.”
Hmm. That’s probably a pretty straightforward modification for existing LLMs, at least at the token level.
You can obtain token probabilities, so you can give some estimate out-of-band confidence in a response, down to the token level. Don’t really need to change anything for that, just expose some data.
And you could make the AI aware of its own neural net’s confidence level, feed the confidence back into the neural net for subsequent tokens, see if you can get it to take that information into account.
en.wikipedia.org/wiki/Recurrent_neural_network
In artificial neural networks, recurrent neural networks (RNNs) are designed for processing sequential data, such as text, speech, and time series,[1] where the order of elements is important. Unlike feedforward neural networks, which process inputs independently, RNNs utilize recurrent connections, where the output of a neuron at one time step is fed back as input to the network at the next time step. This enables RNNs to capture temporal dependencies and patterns within sequences.
You can obtain token probabilities, so you can give some estimate out-of-band confidence in a response, down to the token level.
That means literally nothing. You can get wrong answer with 100% token confidence, and correct one with 0.000001% confidence.
You can get wrong answer with 100% token confidence, and correct one with 0.000001% confidence.
If everything that I’ve seen in the past has said that 1+1 is 4, then sure — I’m going to say that 1+1 is 4. I will say that 1+1 is 4 and be confident in that.
But if I’ve seen multiple sources of information that state differing things — say, half of the information that I’ve seen says that 1+1 is 4 and the other half says that 1+1 is 2, then I can expose that to the user.
I do think that Aceticon does raise a fair point, that fully capturing uncertainty probably needs a higher level of understanding than an LLM directly generating text from its knowledge store is going to have. For example, having many ways of phrasing a response will also reduce confidence in the response, even if both phrasings are semantically compatible. Being on the edge between saying that, oh…an object is “white” or “eggshell” will also reduce the confidence in token probability, even if the two responses are both semantically more-or-less identical in the context of the given conversation.
The problem is that LLMs don’t generate “an answer” as a whole, they just generate tokens (generally word-sized, but not always) for the next text element given the context of all the text elements (the whole conversation) so far and the confidence level is per-token.
Further, the confidence level is not about logical correctness, it’s about “how likely is this token to appear in this context”.
So even if you try using token confidence you still end up stuck due to the underlying problem that the LLMs architecture is that of a “realistic text generator” and hence that confidence level is all about “what text comes next” and not at all about the logical elements conveyed via text such as questions and answers.
And now we hear stories about how easy it is to hack systems with built in LLM’s and when you think about it, they are basically trained to be as helpful and forthcoming as possible, and then we give them the keys to the system!
Tangent: that pic reminds me of the terrorizing tit in Everything You Always Wanted to Know About Sex* (*But Were Afraid to Ask)
Honestly first thing that came to mind was booby.
Hindenburg was a hiccup in history relative to the fallout from an AI bust.
UnspecificGravity@piefed.social 3 weeks ago
The difference being that the Hindenburg was a perfectly functioning rigid airship that had a lot of inherent risks due to the nature of its design.
AI isn’t good enough at its actual job to be in this position. The risk of AI is people pretending that it works when it doesn’t. It would be like if you made a blimp and filled it with carbon dioxide and people kept buying tickets and just sitting there waiting for it to take off.
criss_cross@lemmy.world 3 weeks ago
With society insisting that it’ll take off in the future and only suckers would leave.