AI 'godfather' Yoshua Bengio feels 'lost' over life's work | Hacker News Reader

AI 'godfather' Yoshua Bengio feels 'lost' over life's work

bbc.com

17 pointsalasr12 comments

This takes a special kind of arrogance, or at least self centered view of the world. LeCun has emerged as the only sane voice of the OG AI crowd. Although I haven't heard from Schmidhuber.

It's not like he climbed a bell tower with a megaphone to announce on his own initiative that he "feels lost over his life's work."

A reporter for the BBC contacted him.

> Prof Bengio admitted those concerns were taking a personal toll on him, as his life's work, which had given him direction and a sense of identity, was no longer clear to him.

> "You could say I feel lost. But you have to keep going and you have to engage, discuss, encourage others to think with you."

If you read the article, it's almost like he's even politely disagreeing with the reporter Zoe Kleinman who appears to have asked him something on a zoom call like maybe asked if he "feels lost". I don't know exactly what their conversation was. I don't have the whole video of their conversation.

he wrote more about ai risk here https://news.ycombinator.com/item?id=36042126

If you listen to dismissive HN commenters, having just learned about LLM from ChatGPT 6 months ago, over the AI experts of Bengio, Hinton, Altman, and hundreds of research professors... well, you're just the new type of climate change denier. Now you know how they feel.

'"It's never too late to improve," says Prof Bengio of AI's current state. "It's exactly like climate change.'

> "Bengio, Hinton, Altman"

I know this isn't your point but one of those isn't like the others. Altman is a Musk type power seeking executive, whereas Bengio and Hinton are AI boffins.

Sutskever then, his name was just less prominent. But you’re wrong if you don’t think Altman understands the field and implications at a deep level. Remember, he is the executive that first productized LLMs.

I think this climate change analogy is interesting, but maybe not in the way they meant.

The thing I think is interesting is the trajectory of the 'skeptics' or 'deniers'.

'Climate change skeptic' used to mean they didn't think a systematic change in climate was happening. Then 'climate change skeptic' shifted to mean that of course a systematic change in climate is happening but it's not due to people. Now 'climate change skeptic' means something more like of course people have caused a systematic change in climate but the woke liberals are trying to use it as a hook to force me to change my way of life and adopt their woke policies of eating bugs and sleeping in pods and make my children change their genders.

'LLM skeptic' is following an interestingly similar path. You can see skeptics like Gary Marcus in particular following the path. Originally 'LLM skeptic' meant that these non-mechanistic non-symbolic word predictors were stochastic parrots or blurry jpegs of the web with not even a theoretical prospect of performing anything like reason or approaching the realm of human cognition. Now 'LLM skeptic' means something more like we shouldn't let these dangerous, unreliable, deceptive, and uninterpretable algorithms become certified as lawyers or doctors and we shouldn't let them hold judicial or political offices.

What an utterly absurd statement to make. It is embarrassing honestly and not worth more of a response.

I give you a couple months like the White House Press Sec.

https://twitter.com/kristjanmoore/status/1663860424100413440

The greenhouse effect is measurable, the AI risk debate is a lot of hypothesizing and vibes. I'm not saying there's not risk — I'm just saying you can't directly compare the two.

Also, credentialism is really tiresome. There's plenty of really smart people in adjacent fields that have valuable opinions on things.

The reverse of this is that if you think ai will bring doom then you’re just like those folks that believed 5g brings covid. The people in control of ai are indeed dangerous, they want your data, your ip, and your job. And to get those they want to use laws made as a result of the fear they induce by launching conspiracy theories and anthropomorphize software: “it’s coming for you and it cant be controlled, unless of course you let us ‘tame’ it”.

You know some of the people actively dismissing them ARE ai researchers? I'd put the ratio 10 to 1 in their favor from meetings and that most of the people panicking are administrator types who over estimate and need to hawk for money.

Their is a stink of ignorance from Bengio and Hinton over what threats are already on the table. They seemed to have blissfully ignored them for their careers. It shows up in their proposed solutions/fears.

It's like bengio complaining about computer science needing ethics as a solution. I did take ethics and I still find his arguments wrong. I would argue that he is overestimating risk and panicking, which will create more damage than good. No, he's just trying to claim some mantel of authority by saying he's the only moral one in the room and castigate his opposition unfairly.

In the parlance of his climate change example, he's al Gore claiming that the world sea levels will rise by 20 feet soon, thereby giving ammo to every climate change denier and turning off anyone who needs convincing.

Reply on news.ycombinator.com