Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Google fires another AI researcher who reportedly challenged findings

engadget.com

Post reply on HN
Print view View original

181–190 of 238 posts

Re: Google fires another AI researcher who reportedly challenged findings

#181

Post by heavyset_go »

brian_cloutier wrote: post #177
heavyset_go wrote: post #175

Earlier quoted context omitted.

Anthropomorphism isn't just about directly comparing things to people, it's also about assigning qualities that are related to people, like motivations, goals, opinions, concepts like good and bad, etc, even if the specific qualities would be alien to most people. In that sense, the AI safety crowd has been anthropomorphizing hypothetical AI for close to 20 years now.

Do you have an example of someone in the community making this mistake? My understanding is that if we knew that all future AGIs would have human-like motivations, goals, opinions, and concepts of good and bad then that crowd would be much less concerned. Smarter humans are not the concern. Intelligences which are _not_ human-like are the explicit concern.

Roko's basilisk comes to mind.

Re: Google fires another AI researcher who reportedly challenged findings

#182

Post by sjg007 »

mcguire wrote: post #169

" Google fired machine learning scientist Satrajit Chatterjee in March, soon after it refused to publish a paper Chatterjee and others wrote challenging earlier findings that computers could design some chip components more effectively than humans. The scientist was reportedly allowed to collaborate on a paper disputing those claims after he and fellow authors expressed reservations, but was dismissed after a resolut…

It is a strange hill to decide to die on. Even if you are the expert in the area, AI is weird and changing the world.

Most Nature papers have something wrong with them. I don't see any reason that reinforcement learning cannot learn the best layouts or what not. I don't know the space though enough to say more than that.

The NYTimes article alleges that he waged a misinformation campaign against two researchers. So there's that in the mix as well.

Re: Google fires another AI researcher who reportedly challenged findings

#183

Post by godelski »

tedivm wrote: post #4

What really kills me is that Google claimed previous papers weren't up to their standard, but then ended up citing them in later papers that they've had press events around. It's really unfortunate that so much of our science is being done at large companies rather than universities, since the companies have a bias towards their bottom line which keeps them from publishing things that might damage that.

Papers may not meet the standards for publication but still be useful. For example, I have a paper with 50 citations that is having difficulties getting published (either "not novel enough" or "not motivating"). There's also a randomness to review, as NeuralIPS has continually demonstrated.

I should also mention that you can also pad your citations this way and build your h-index. It is metric hacking, but if it isn't too egregious this is normal. Classic example of Goodhart's law.

Re: Google fires another AI researcher who reportedly challenged findings

#184

Post by brian_cloutier »

SpicyLemonZest wrote: post #144
brian_cloutier wrote: post #115

Earlier quoted context omitted.

I wonder if you've had an in-person conversation with someone who would identify as being a part of the AGI safety community. I've talked with several people and their arguments are a lot more sophisticated than an intuition that ML models would have human motivations. This post isn't written by somebody inside the community but he presumably has access to them and has had conversations with them which shape his beli…

The core thing I (and I suspect the original commenter) struggle to get past is the invetiable twist that in this post happens halfway through this post's part II. > If it’s a very smart mesa-optimizer, it might think “If I throw the strawberry at the streetlight, I will be caught and trained to have different goals." It seems to me that this is a category error, like having the very smart mesa-optimizer start thinki…

I don't think I understand your objection because it doesn't seem like a category error to talk about optimizers having goals.

I think you would agree that thermostats have goals? They try to minimize the error between the desired and the actual temperature. And you would also agree that gradient descent has a goal? It tweaks parameters in the search for models which minimize error in the training set. The system performing that gradient descent was designed by humans and exhibits goal-like behavior.

But you think it's a step too far to believe that gradient descent could create a model which also exhibits goal-like behavior?

What is the difference in category that you see between those two steps?

I agree that humans are not goal-directed in the same way the community is worried that AGI might be. This makes it surprising that seeing AGI as potentially goal-directed is seen as anthropomorphism, humans often question their goals in exactly the way there is concern that AGI will not!

Re: Google fires another AI researcher who reportedly challenged findings

#185

Post by RC_ITR »

pharmakom wrote: post #114
RC_ITR wrote: post #101

Earlier quoted context omitted.

Saying a paraphrase of "I didn't quit, I gave my employer an ultimatum that they then rejected" [0] is a very toxic way to interact with people. 1) Ultimatums in general are not a healthy way to interact with others because they intentionally try to skew power dynamics towards the giver (Do what I say or else...) 2) To then act (upon failure of the ultimatum) as though her employer acted inappropriately implies that…

If you cannot give an ultimatum then you can always be pushed around.

Unless you stop making comments like that, I will never upvote a comment you write ever again.

^that is certainly more 'toxic' than:

I disagree with your point. I think disagreements and negotiation with an employer are fine (to your point, one can even use the threat of quitting as part of said negotiation), but ultimatums just serve as an opportunity for the giver to imply that they have more power than the receiver, which is 'toxic' behavior

Re: Google fires another AI researcher who reportedly challenged findings

#186

Post by abduhl »

bluefirebrand wrote: post #165
Manuel_D wrote: post #137

Earlier quoted context omitted.

I'm pretty sure if you say to your manager "I quit" and stop showing up to work, regardless of what internet forms you fill out, it's a resignation. The fact that Gebru gave her resignation in the form of an ultimatum rather than whatever website form you're talking about is a pedantic point. "I resign if you don't do X." "We're not doing X, we accept your resignation." This is a resignation the way I understand it.

> "I resign if you don't do X." > "We're not doing X, we accept your resignation." This approach from her managers can be criticized because it immediately closes the door on the possibility of her changing her mind and walking things back. However, it sounds like she was causing a lot of internal problems so perhaps they were already working to get rid of her. Accepting the implied resignation may have just been exp…

You can certainly criticize Google for taking someone at their word when that person issues an ultimatum, but then you’re tacitly admitting that the person offering the ultimatum is approaching the negotiation in bad faith. Google should not have to assume its employees are acting in bad faith when they communicate internally, especially when it comes to the future of that employee’s career.

The real question for your framing is: Should Google have to say, “ok that’s it for real this time, no take-backsies”? I think that google should never have to say this any time someone says “or else I quit” because the threat of quitting tells us that they’re not willing to continue operating within Google. I honestly don’t see the reasoning behind drawing the line anywhere else.

Re: Google fires another AI researcher who reportedly challenged findings

#187

Post by themacguffinman »

bluefirebrand wrote: post #165
Manuel_D wrote: post #137

Earlier quoted context omitted.

I'm pretty sure if you say to your manager "I quit" and stop showing up to work, regardless of what internet forms you fill out, it's a resignation. The fact that Gebru gave her resignation in the form of an ultimatum rather than whatever website form you're talking about is a pedantic point. "I resign if you don't do X." "We're not doing X, we accept your resignation." This is a resignation the way I understand it.

> "I resign if you don't do X." > "We're not doing X, we accept your resignation." This approach from her managers can be criticized because it immediately closes the door on the possibility of her changing her mind and walking things back. However, it sounds like she was causing a lot of internal problems so perhaps they were already working to get rid of her. Accepting the implied resignation may have just been exp…

She never gave Google any reason to believe that she had wanted to walk things back or change her mind. That's the nature of delivering an ultimatum by threatening to resign: you are explicitly offering an immediate end to the negotiation. Don't be surprised if management takes you at your word, they're not mind readers. If she wanted to signal to managers that she might not actually quit over this, then she shouldn't explicitly say that she will quit over this.

Re: Google fires another AI researcher who reportedly challenged findings

#188

Post by faeriechangling »

dehrmann wrote: post #54
hn_throwaway_99 wrote: post #22

Obviously don't know the full context of why Satrajit Chatterjee was fired, but I can say that I followed the Timnit Gebru story quite closely, and there was a gigantic chasm, IMO, between what actually happened (and by "what actually happened" I mean that I'm conservatively just considering Gebru's own statements and tweets) and how things were reported in the wider press. The press will always be incentivized to re…

I'm surprised news outlets still report that Timnit Gebru was fired. It's at best disputed. This case looks much clearer; he was terminated with cause.

It’s not disputed by anything other than googles PR. By well established California law legal resignations include an end date, she did not ever specify an end date when threatening resignation, and thus she never legally resigned. She was thus fired.

It would be bias in googles favor to act as though she were not fired which is what any court would find. Sort of a “discuss the controversy” sort of thing over a very clear cut legal issue.

Re: Google fires another AI researcher who reportedly challenged findings

#189

Post by themacguffinman »

dekhn wrote: post #142
gwern wrote: post #136

Earlier quoted context omitted.

> the cause was being a jerk to coworkers repeatedly (she repeatedly attacked internal researchers on internally public mailing lists in an unprofessional way). No, the cause was her ultimatum email demanding the names of the reviewers among several other things (which she refers to as "simple conditions") or she quit, and then also emailing Googlers telling them to stop working. The first was the email her manager r…

Those were all proximal causes, not ultimate causes. I don't think she was a victim, I'm not an apologist, but the way google fired her was still rather unprofessional and had some severe consequences for the people who did it which are obviously still playing out today. What's amazing is that if they're put her on a pip for textually assaulting her coworkers, she would have been gone in 6-12 months without as much n…

Few are really fooled by a PIP response like this. I'd fire anyone who would make the decision to give such a high profile activist 6-12 more months of authorization to internal company materials and direct contact with colleagues to lash out against the company with the not-so-hidden knowledge that they're on their way out. We already know that another colleague who sympathized with her attempted to exfiltrate company material after Gebru was summarily fired. Many companies will still immediately escort you from the premises if you quit amicably.

Re: Google fires another AI researcher who reportedly challenged findings

#190

Post by faeriechangling »

Manuel_D wrote: post #109
dekhn wrote: post #100

Earlier quoted context omitted.

People argued about this a bunch. here's a simple summary that is consistent with the reality: Google (Jeff's docs) says that Timnit offered her resignation (over the paper withdrawl demand and a refusal to let Timnit confront her internal reviewers) and Megan accepted it. In reality, Megan terminated Timnit for cause, but used Timnit's offer to resign against her. At google, resignations are normally triggered by th…

I mean, if you say "do X or I quit" and the company responds "we're not doing X, bye" is it incorrect to say that the employee quit? That's the situation here, Gebru issued demands and said she'd quit if they weren't met. Google said, "we accept your resignation".

Yes, under California law a legal notice of resignation requires an end date. She did not legally resign, and thus she was fired.
Post reply on HN