Another Firing Among Google's AI Brain Trust, and More Discord (nytimes.com) 68
Less than two years after Google dismissed two researchers who criticized the biases built into artificial intelligence systems, the company has fired a researcher who questioned a paper it published on the abilities of a specialized type of artificial intelligence used in making computer chips. From a report: The researcher, Satrajit Chatterjee, led a team of scientists in challenging the celebrated research paper, which appeared last year in the scientific journal Nature and said computers were able to design certain parts of a computer chip faster and better than human beings. Dr. Chatterjee, 43, was fired in March, shortly after Google told his team that it would not publish a paper that rebutted some of the claims made in Nature, said four people familiar with the situation who were not permitted to speak openly on the matter. Google confirmed in a written statement that Dr. Chatterjee had been "terminated with cause." Google declined to elaborate about Dr. Chatterjee's dismissal, but it offered a full-throated defense of the research he criticized and of its unwillingness to publish his assessment.
terminated with cause (Score:2)
I guess he won't get his unemployment check.
Re: (Score:3)
Yeah, you don't get terminated with cause unless they have something that could be defended in court as illegal behavior or a gross violation of contract. This isn't about AI.
"This isn't about AI." (Score:1)
Unless it is, and they're banking on him not challenging. It if wasn't there wouldn't be reason to withhold the rebuttals.
Sue for wrongful termination and file the withheld documents as evidence. Even if he loses, the papers will be public record.
Re: (Score:2)
Peanuts with regard to time spent on playing any sort of tete-a-tete with Google lawyers, especially if they're TRYING to waste your time.
Re: (Score:2)
The original paper's authors were also from Google.
It sounds like he was calling out co-workers on potentially misleading to fraudulent results and research, and only insiders would likely have enough detailed knowledge to do that.
Re: (Score:1)
Everything cannot possibly be either an application or an application hosted by an application or an app on that.... That is sure to cause conflict. Sometimes it's not "the policy of..." place your platform name here. Sometimes you just know "they" got it to work, then build on that.
Re: (Score:2)
Smart people can be jackasses too. Over my long career, I have worked with really smart folks, who were terminated, and my work defiantly got much better because of it, even if their replacement wasn't as smart as they were.
They often got fired because they crossed the line, and they thought they were untouchable, because they were so smart, and they were like when they were fired well the company is just going to fail without me.
Then we usually pickup his work, which usually isn't a major burden, because h
Re: (Score:2)
Google didn't cut this guy loose until they were absolutely certain they had the goods to make it work their way. They don't arbitrarily lash out at anybody or anything which opposes them; they are quite careful about it, and this implies that they have reasons they believe will stand up in court. Now, this doesn
Re:terminated with cause (Score:4, Interesting)
That's not actually true. Sometimes companies will claim that you are terminated with cause and then just not defend it at unemployment office.
In California courts, it doesn't matter whether it was with cause or not, as long as the cause wasn't racist, sexist, etc.
Re: (Score:2)
Yeah, Californian here. 100% "right to work" state, meaning you can quit at any time (because nobody can stop you from getting a different job) but also, paradoxically perhaps, you can be fired at any time for any reason or for no reason. The only thing that can't happen is that you can't be fired because of your race, gender, religion, and and other legally defined protected classes—not even if the excuse is something like "you don't get along with everyone else here." And there are other causes of "
Re: (Score:2)
Worth mentioning that "for cause" or "not for cause" is meaningful in terms of whether you get unemployment benefits or not. Other than that it doesn't matter.
Re: (Score:2)
It also matters if a person has a contract that stipulates they can only be fired for cause. Most of us non-union wage slaves in the US do not have contracts like this, but it doesn't mean there can't be some people who do.
Re: (Score:2)
Good point. Something worth taking into consideration when deciding to join a union or not.
Re:terminated with cause (Score:4, Insightful)
Re:terminated with cause (Score:5, Insightful)
You have to understand, collective entities such as corporations and even governments are not bound by morals and ethics, only laws. I feel (as strongly as you, I suspect) that Google is behaving quite amorally, but Google is not a person. The only applicable yardstick is legality, and until Google loses in court, their behavior, while unethical, was justified. For Alphabet Inc., profit is its own reward. That's how free enterprise works.
I'm quite certain Alphabet Inc. has more than ensured that they will have evidence to justify their termination of Satrajit Chatterjee's employment. I'm equally certain they have the legal justification to see to it that the unfortunate research conducted by Mr. Chatterjee will never see the light of day, It sucks - it's sure wrong, but it's quite legal.
Re: (Score:2)
My assumption is that, as a PhD, he had a requirement to publish a certain amount of work. When he didn't do that, he was fired.
Whether his paper deserved to be published or not, we can't know at this point. My assumption is that the paper was 100% correct, but it might not have been novel enough to merit publishing.
Re: (Score:2)
Unless, of course, somebody with deep pockets gets involved (or a lawyer who can afford this kind of pro bono work) and uses the Freedom of Information act to pry it loose.
Re: (Score:2)
Your statement is so utterly absurd I don't know where to begin.
I think you're high. Most of your comment has nothing to do with what I said. You're hallucinating or something. I didn't "side with google" or anybody.
Re: (Score:2)
I work in a Right to Work State (Texas) and had no trouble getting on unemployment after being terminated.
But since this is a Google employee, I would think the odds are that he is from California.
Re: (Score:2)
It's possible this person negotiated a contract that stipulated they could only be fired for cause. Usually you only see that kind of thing in association with union contracts, but it's not out of the question that an individual who has skills that are in high enough demand they could get it added.
Re:terminated with cause (Score:4, Informative)
If the guy is in a "right to work" state, he can be fired at any given moment for any given reason and you can't do crap about it
First, you mean "at-will," "right to work" means you do not need to join / cannot be forced to join a union for employment.
Second, no, it doesn't mean that at all - that would be ripe for all sorts of abuse, and violation of federal law. They have much more leeway in firing someone, but it still cannot be for illegal reasons (violating federal labor laws) - and yes, you still have recoruse through the NRLB, etc.
This bullshit being spread is why people are so unaware of their actual rights. FFS... ~_~
Re: (Score:2)
Re: (Score:2)
"Right to work" has nothing to do with unemployment. Yes, a company can fire you for any reason but they have to pay unemployment unless the termination is "for cause".
Re: (Score:2)
Yeah sorry, as so many people have pointed out, some places in the USA have "right to work", meaning you can be fired at any time for any reason. To those of us elsewhere in the world, this boggles the mind. "For Cause" here in Canada basically means you have to have fucked up really, really badly - to the point where I was once told by an executive of a large public company that "nobody gets fired for cause".
Here in Canada, if they want to get rid of you, it has to be:
(1) Part of a larger, defined and di
Re: terminated with cause (Score:2)
"Right to work" means the law in the state gives workers the right to get and hold a job without having to join or pay dues to a union, and the union can't interfere with a worker's ability to receive promotions, hours to work, etc.
Not everybody feels comfortable being forced to give money to an organization who supports causes or political candidates they despise.
You probably are thinking of the doctrine of "at will employment" which means your employment is at the will of both sides and there is no contra
hive mind (Score:1)
Re: (Score:1)
> Just ask Trump.
Should I ask him over Twitter?
Re: (Score:2)
Yeap, according to Mr. Musk
Re: (Score:2)
Re: (Score:3)
And what problem could there possibly be for freedom of scientific exchange if major research labs fire people for trying to publish contrary opinions?
If you're attitude is "but of course, they're the boss, why should they be foreced to pay him to work against them", you're pretty much establishing that commercial research groups such as Google Brain [1] can't be trusted and probably shouldn't be allowed to publish in journals such as Nature.
[1] "Google Brain" sounds like the condition that afflicts go
Re: (Score:2)
I think you'd have to go back and look at Edison and Tesla's falling out. Tesla had a better idea, alternating current and Edison though his direct current which was simpler was better and ultimately should be the winner. Westinghouse and Tesla proved Edison wrong. Ultimately this guy may join up with another industry leader and prove Google wrong, you never know.
Re: (Score:2)
Re: (Score:2)
If they pay his wage, then under normal employment agreements they own his work product.
I have two patents to my credit that were based on technologies I developed during my career, do I own them? No. Why? Because I worked for a company that I agreed to give them exclusive rights to what I produced while employed by them.
Re: (Score:1)
If Mr. Chatterjee feels that he was wrongfully terminated (say, because of race, creed, retaliation, sexual orientation, whatever), he should act im
Re: (Score:2)
Then you don't see a problem in the fact that "commercial research groups can't be trusted".
I couldn't care less about the legalities here, my point is that Google Brain isn't doing science, they're doing
Re: (Score:2)
Re: (Score:2)
Nope. Sure don't. That's why scientific papers are reviewed.
Spoken like someone who hasn't ever peer-reviewed a thing...
The ROC curve for scientific peer-review is mediocre on a good day and dog-shit every other day (i.e. I have often been the sole dissenting vote on approving papers that were beyond fundamentally flawed and an absurd ratio of published results are simply not-reproducible). So if your corporation is a major player in field X, and you selectively flood the system with only papers pushing a particular narrative/result while simultaneously selectively
Something rings wrong in my brain... (Score:2)
"...my point is that Google Brain isn't doing science, they're doing advertisements, and they shouldn't be allowed to publish in scientific journals."
My first inclination is to begin searching for examples of commercial enterprises and private companies that published productively to scientific journals. Then I realized that the narrowing process was more effort than I wish to extend. So I shall go only as far as saying that blanket statements like that are likely neither defendable, nor useful.
It's up to t
Re: (Score:1)
Re: (Score:2)
Re: (Score:2)
So you're the one with the dangerous fruit [cnn.com].
Re: (Score:2)
Re: (Score:3)
If they're having papers published in a scientific journal, they are reasonably held to a different standard. If they refuse, I would suggest that peer-reviewers consider carefully the likelihood that a researcher working for Google may be under coercion and so their paper requires more stringent scrutiny before publishing. Or perhaps it simply shouldn't be published at all unless and until the result is reproduced by someone with no association with Google. To do otherwise risks the journal coming to be re
Re: (Score:3)
That's exactly it. Scientific journals are for people doing science, and science is based on a free and open exchange, and Google is making it look a lot like they're using Nature to publish advertisements, which you you'd think Nature would have problems with.
Re: (Score:3)
I'm confused: ask him how? Twitter?
Re: (Score:2)
Terminated with cause gets rid of the illusion that he was silenced for his opinions. I'm thinking this guy had a contract that had specific Ts&Cs that prohibited certain behavior. Or he could have been like a boss I had a long time ago, getting caught with the receptionist doing the horizontal slide in an execs office.
Re: (Score:2)
Re: (Score:2)
Love how the morons are jumping to Google's defense. Like a corporation needs defending.
Re: (Score:2)
The termination reasons are between Alphabet and the employee. Unless they want to make that public we'll probably never know the reason, until then this is all speculation. I don't agree with blocking free speech but for some reason I kept hearing "It's a private company, they can do what they want" from certain factions over the past few years. No presumably free speech should be allowed even when a company has Ts&Cs owning your IP?
Shit, we still have people yelling about their feelings years after "G
Re: (Score:1)
Re: (Score:2)
Re: (Score:2)
He got terminated because he's lying piece of shit.
Re: (Score:2)
Do you know this as a fact?
I've seen many companies terminate employees. Not all of them are a 'lying piece of shit'. I typically find executives being that more often. And if they force their employees to agree to arbitration (vs. being able to sue), then they are basically untouchable unless they fire you over something that is protected (i.e. getting fired because of your religion or you do what they tell you and then they fire you because you did exactly that, etc.). As we all know, arbitration is not a
Nature (Score:2)
Articles that are about AI that appear in Nature have a habit of turning out to be wrong or irrelevant within a few years. Nature's editors have expertise in biology, not computer science.
Re: (Score:2)
Google has a habit of exaggerating the power of the AIs they develop in the press. (That is, the AI they develop is good, but not as good as they make it sound).
Re: (Score:2)
And Nature's editors buys it easily...
Re: (Score:2)
They just want something that attracts attention for readers.
Re: (Score:1)
Scientific papers are not reviewed by the editors, but by your peers. In the case of the Nature main journal, usually three peers. And I would be quite surprised if any of these reviewers weren't from computer sciences, and more than that, if they had not published recently several well-cited papers in exactly the same area of the submitted paper.
When you submit a paper to a respectable journal, usually you get your worst enemy, the most skeptical person, as a reviewer. And the fact that this happens, is ve
Seems Google is telling us (Score:5, Insightful)
How toxic the environment is at their AI division without actually telling us.
In other news (Score:1)
"terminated with cause" (Score:3)
... Dr. Chatterjee had been "terminated with cause." Google declined to elaborate about Dr. Chatterjee's dismissal, but it offered a full-throated defense of the research he criticized and of its unwillingness to publish his assessment.
Translation: oUr AI iS dA B3sT!!! DiS gUy dOn'T KnOw cRaP!! oH! aNd hE tOtALLy diD sUmThIn' iLLeGal, aNd dAt'S Y wE fiReD hiM!
Yeah... when a corporation has to work this hard at justifying their position, it seems to me that they know full well they're full of crap. I would hazard a guess that the good doctor's assessment will eventually come to light, especially after their use of hard sale tactics to suppress his opinions. Invoke Streisand Effect, much?
"... not permitted to speak openly" (Score:2)
Our world favors the powerful. Journalists try to discover when justice is at risk, but contracts and even court orders often require witnesses to be silent. How then can the public discover injustice? Journalists get statements off the record, anonymously. OK, that helps, but without attribution from known witnesses, the story is weak. Maybe the author made up some 'facts'?
The courts should consider outlawing secrecy in contracts and jurisprudence, with allowances for competitive trade secrets. Controversi