Yann LeCun isn’t buying the AI doomer narrative | Hacker News Reader

Yann LeCun isn’t buying the AI doomer narrative

fastcompany.com

30 pointsmfiguiere46 comments

I was hoping for more substance than "guy who got rich off thing thinks thing is great" but there really isn't much else here.

Yann is the sole anti-doomer from the Bengio and Hinton trio. Also, the only one whose "salary depends upon his not understanding it," hmm...

> Also, the only one whose "salary depends upon his not understanding it,"

All of them make, and continue to make, money from talking about AI, the doomers are probably getting more attention because the public and media love a good thriller story.

Would it not be the opposite though, that instilling fear to cause regulation will limit competition and new entrants?

To me, the doomers of the trio seem to be the ones

> whose "salary depends upon his not understanding it," hmm...

The irony that your website is named "hofstadter", yet:

"And my whole intellectual edifice, my system of beliefs... It's a very traumatic experience when some of your most core beliefs about the world start collapsing. And especially when you think that human beings are soon going to be eclipsed. It felt as if not only are my belief systems collapsing, but it feels as if the entire human race is going to be eclipsed and left in the dust soon." -Hofstadter

https://www.lesswrong.com/posts/kAmgdEjq2eYQkB5PP/douglas-ho...

I suppose after everything you've read, he's just trying to make money selling more books or something...

You know.. you can believe in or be inspired by someone without buying into everything they say...

fyi, your comment is super low quality and on the line of antagonism, ironic given your HN handle would make one think you would not impose unnecessary suffering (by deriding something someone said)

---

edit, your comment history is quite telling, I would change my HN handle if I were you

> low quality

> antagonism

> change your handle

Pot meet kettle

Do you personally find yourself suffering because of this comment chain?

I find myself in the school of prioritizing serious matters like homelessness, loneliness, destitution, slavery, truth, and averting death, when weighing against someone falsely claiming any amount of real suffering from an internet comment.

Btw you should seriously consider the opinions of Hofstadter, Bengio, Hinton, many other preeminent scientists, and spend time figuring out why they think this way. There’s scenarios where a lot of unnecessary suffering comes if people don’t think it’s a catastrophic danger and try averting it.

ok doomer...

review your comment history and see how your takes have fared

With net positive upvotes? Are you really making an appeal to popularity, anyway? You realize this place mostly laughed at OpenAI about AGI for years until ChatGPT came out. They were popularly wrong. Time will tell, you're just here: https://twitter.com/kristjanmoore/status/1663860424100413440

You're really good at whataboutism, I'll give you that

https://youtu.be/c1Fv1uKTd-w?si=FnDPTkb1maXDiRLx

I think the AI doomerism is ridiculous. I guarantee there are some sleep-deprived Russians, either sitting at the bottom of the ocean in a nuclear submarine, or in front of a control panel with a button that can more or less end the world. And people are making all this noise about a glorified chatbot.

We already have the capability to end the world and it’s hanging by a thread. People should take the real threats seriously, there are enough of them to not need to fantasize about hypothetical ones.

I don't think people are afraid of skynet level AI taking over the world. Current AI is more likely to cause people lose their jobs to vastly inferior but cheaper AI bots. Executives are overestimating its value over real humans, and with right marketing you can easily convince them to fire people in favour of bots. Its a threat of automation where only people who profit are top 0.01%. While nukes might end the world, I don't think we should ignore any other threat just because nukes exist. We shouldn't mindlessly add more tech in our lives where none is required.

I have been thinking a lot about it recently and I do think they are significant.

If you look for example at chagpt4, it achieves a certain sort of intelligence. It gives you many sensible responses.

The thing Is that this is basically what we humans do. Give so-so explanations about all sorts of stuff that are reasonable-sounding.

Im confident that they well become evidently better than all of us. More complete arguments, with better data backing that up that the regular folk. In some sense I feel that they almost above that threshold.

The basic opinions of people are neither sophisticated nor deep. And llms are always improving to become exactly that.

Sell Yann on the climate doom, see if they/them buys that.

I think Yann LeCun is living in a fantasy world.

He seems unaware that we already saw a massive failure on his watch where a US election (2016) was tipped by algorithmic intervention far beyond anything that had come before, using his company’s personalization platform.

Yes elections are influenced all the time in different ways, but this was next level.

And it’s already eclipsed by the next stuff, which is not stopping, and he doesn’t even acknowledge it at all in anything I’ve seen.

Then there is the delusion that there is a cohesive “we” who are deciding what happens with AI, which he subscribes to. There is no such thing. Humanity is not at all unified, including on whether human extinction is desirable. He assumes we are. Dangerous assumption.

I dunno, the most parsimonious explanation for Clinton’s loss is low black voter turnout, which is more a simple failure of her campaign than a nefarious new technology propelling us to extinction.

It was a very close election. There are any number of things, that, had they been different, Clinton would have won. Trying to ascribe a single cause is fallacious.

It's much complicated than that.

https://fivethirtyeight.com/features/the-real-story-of-2016/

That’s the first in a series of articles. The next article in the series states:

> the regression … can explain the Electoral College drop-off Clinton experienced relative to Obama based on some simple demographic variables and the 2012 vote alone.

GP didn't state that it was the sole reason for tipping the election.

Clinton dropping the ball, Comey etc. are events I can live with even if I didn't like the result. But a foreign adversary exploiting network effects to manipulate online discourse and propagate misinformation on a large scale – the influence that had will always be an unsettling "what if" in my mind.

There has been foreign promoted disinformation since at least the Cold War. Newspapers and Hollywood have propagated conspiracy theories about the JFK assassination that can be traced to soviet disinformation.

There is so much noise online that I really doubt this made any difference

The difference is that thanks to Google and Facebook you can now do it on an industrial scale. It's like comparing an iPhone to two tin cans connected with a string.

The Hungarian government is also using ads to brainwash people very effectively here. You can't watch a YouTube video without seeing some kind of propaganda.

Not to mention the real power of targeting: there's even less need for a coherent narrative, you can tailor it to each niche group (or even potentially each individual, at this point). And other's don't even know what your message is. If a lie could previously travel around the world before the truth had got its boots on, a lie can now travel around the world before the truth even knows there's a race.

I guess, maybe I misinterpreted what “tipped by algorithmic interference” means. Regardless, I just don’t think the world’s going to end if somebody’s uncle in Florida reads dozens of false headlines. A small number of people consume a lot of misinformation. I agree with Yann on LLMs, and I don’t think his work on Facebook invalidates that.

LeCun comes from a country (France) where people vote, in person, using paper with the name of your candidate, in a locked transparent urn, after showing their ID documents. And people take politics seriously enough not to be influenced by ads on Facebook.

Maybe he's just not swayed by arguments about the unreliable US election system?

   Yann LeCun is living in a fantasy world

It's hilarious he'd be accused of this for not believing computers will somehow come alive and get us.

If that’s his interpretation of peoples concern, then he is living in a fantasy world.

There are partisan nutters on on Twitter peddling AI generated images of politicians engaged in things they’ve never done.

It’s not Skynet going rogue that’s the concern, it’s Peoplenet triggering nut jobs (regardless of ideology they proclaim to serve).

Damn, you've pointed out the problem.. the issue isn't that this feared AI is not as smart as a 10 year old (to use TFA's example), but that there are millions of adults with less reasoning capabilities than 10 year olds.

> believing computers will somehow come alive

Of course practically everytime the anti-doomer argument comes out, it fails to even understand the basic premise. Instead preferring to laugh at hyperbole, the core of a serious argument, everyone knows...

Can you elaborate a bit more on the thing that everyone knows?

It was /s . "Everyone knows" that using a hyperbole and laughing about it, isn't actually a valid argument.

It’s hard to tell sarcasm or hyperbole when it comes to this topic. People will straight-faced post stuff like “I demonstrate my expertise in this field by being extra nice to my Hatsune Miku girlfriend to show fealty to Roko’s Basilisk” and get mad if that doesn’t serve as the bedrock for serious discussion

That’s not the only failure mode though so it’s possible you’re missing something.

Also I said “I think” which means I’m describing my own opinion, not stating with certainty that I know something to be the case about him.

I think people overestimate the likeability of Hillary massively

Do they? She won the popular vote by almost 3 million in the 2016 Presidential election.

When one state had 4M+ in favor of hillz, does this mean 3M in popular is a moot arguing point, since by these maths, she lost by more than 1M in the rest of the country...

> tipped by algorithmic intervention far beyond anything that had come before

Hasn’t the data shown that people don’t really change their minds based on what they read on social media?

I think you are underestimating all the other factors that probably had stat-sig influence on the elections.

Low voter turnout from some segments, broad voter disenfranchisement with both parties, media relentlessly pushing Trump to front and center because it drove ratings, Bernie Sanders’s voters sitting out as protest, etc. etc.

The AI Doomer narrative is the one Yudkowsky is pushing about some skynet type scenario, not about automated internet disinformation, The latter is taken VERY seriously at Meta, with quarterly reports of ongoing disinformation campaigns and tactics used on the platform.

> The latter is taken VERY seriously at Meta, with quarterly reports of ongoing disinformation campaigns and tactics used on the platform.

Whatever helps you sleep at night I guess.

I have friends working on that team. Every small success in cracking down is still a success.

Your working for an Ad company, lets not overstate the importance here. Its like the bottom rung of being a techy, right next to finance. Next your going to say that 3M,Johnson & Johnson and Montasanto are "doing all they can". Even small wins right, token gestures of repentance? while execs laungh and continue to profit off the mass poisoning of our waterways & lives simply by influencing laws such that they can only be sued by one person/group at a time. Investor profit over all else, once you cut through the HR, is what it boils down too.

Not sure why people think its respectable to work as a propaganda piece for a corp covering its tracks. So many challanges there to solve, so many "smart" people dedicating their lives to making people want things they don't need. Sounds inspiring to be surrounded those types. Truely.

So innovative and cutting edge it is working for a giant ad company that we all know takes money from whoevers shell company offers it or whatever reason. Plausible deniability is the game at this level, just enough to paint a favourable picture in the court of court of law when the day eventually comes.

> US election (2016) was tipped by algorithmic intervention far beyond anything that had come before, using his company’s personalization platform.

It was actually the complete opposite. In 2016 these recommendation systems were fairly unbiased. If Trump content was upvoted it was recommended. Then after his election tech companies had a never-again moment and started banning and demoting right-wing content even though it was popular, thus tipping the 2020 election.

See e.g. https://www.breitbart.com/tech/2018/09/12/leaked-video-googl...

Edit: I guess this is getting downvoted over the breitbart link, but I have through personal sources received confirmation that the video is legitimate.

Your edit in TL;DR: "source: trust me, bro.".

It's not implausible that the mostly liberal tech workers decide to not boost "right wing" content, but I'm not going to go down rabbit holes to find out, especially Breitbart rabbit holes where some random HNer said at the door "what you find in this rabbit hole will be true!".

Thanks for explaining what your issue is. I'll try to build a better case with that in mind. What the video shows is that top Google leadership was clearly upset over the win. The video is the primary source, but there are also secondary sources that implicitly confirm its authenticity. E.g.

https://edition.cnn.com/videos/cnnmoney/2018/09/13/google-vi...

https://www.theguardian.com/technology/2018/sep/12/breitbart...

That shows that they were emotionally upset, but by itself doesn't give evidence of any concrete action. However it helps to put the other things in context.

For concrete things, firstly, there was algorithmic demotion.

This is a 2017 article that explains their Google's effort to reduce upsetting-offensive content

https://searchengineland.com/google-flag-upsetting-offensive...

> Facebook will change algorithm to demote “borderline content” that almost violates policies (...) The change could massively reduce the reach of incendiary political groups, fake news peddlers https://techcrunch.com/2018/11/15/facebook-borderline-conten...

There were also some outright bannings. Two famous cases:

> In 2016, Twitter permanently banned Yiannopoulos for his participation in a targeted racist harassment campaign against comedian Leslie Jones. In 2017, Yiannopoulos resigned from Breitbart amidst outrage over comments he made seemingly defending pedophilia. That also resulted in the termination of his book deal with Simon & Schuster. Universities canceled multiple speaking engagements, including his 'Free Speech Week,' amidst protest to his ideas on, well, everything. And as recently as last week, Politicon pulled him from the speaking lineup — which was to be his return to the speaking spotlight.

https://mashable.com/article/milo-yiannopoulos-deplatforming...

> YouTube, Apple and Facebook have removed main outlets for conspiracy theorist Alex Jones and his Infowars website, citing repeated violations of policies against hate speech and glorifying violence. Infowars responded by accusing the companies of censorship.

https://www.npr.org/2018/08/06/636030043/youtube-apple-and-f...

Now you may perhaps say that this was all justified. But that doesn't matter, as all I'm trying to do is show that it happened.

Facebook alone didn't tip the 2016 election. What about the New York Times critical treatment of Hillary? What about James Comey? This post is misinformation.

Of course it didn’t happen in a vacuum. And Hillary was a horrible candidate imho.

Yann bless

Reply on news.ycombinator.com