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

Google told its scientists to ‘strike a positive tone’ in AI research

reuters.com

Post reply on HN
Print view View original

211–220 of 307 posts

Re: Google told its scientists to ‘strike a positive tone’ in AI research

#211

Post by abdullahkhalids »

jchallis wrote: post #146

Every specialty has taboos and learning what they are and how to avoid them is part of being successful with the job. As a physics graduate student I made an impassioned case to a high energy researcher that the money spent on high energy research was low yield and should be pushed to the biological sciences ... suffice it to say he did not take my points on their merits. Upton Sinclair’s famous “ It is difficult to…

I am a physicist, and I would say that, perhaps, you don't understand why we "waste" money on high energy physics. It's very similar to the reason why rich people buy useless status symbols, when they could instead buy useful things. It's essentially telling everyone, "I am so rich I don't care about money anymore and can afford to waste some of it on this useless thing".

In science, we as a society throw money at something like high energy physics, because we are telling ourselves, "We love knowledge so much, that we will throw it at high energy physics, which has little economic benefit to us" [1].

In other words, you can't prove that you love knowledge for its own sake, if you only invest in fields which have economic benefits. You have to invest in fields with no benefit, to prove to ourselves that we still care. That has deep and lasting philosophical impact on society.

[1] If you read grand funding agencies internal documents, you will learn that the primary economic benefit of fields like high energy physics is training people on hard problems, who will then go and apply those skills and raise the bar in other fields.

Re: Google told its scientists to ‘strike a positive tone’ in AI research

#212

Post by cmiles74 »

googlethrow3134 wrote: post #208
cmiles74 wrote: post #176

Earlier quoted context omitted.

This seems very different to me than hiring researchers doing work in fields that directly impact your business and telling them their work will not be censored. And then censoring them from a position of power, since you now provide their paycheck. Keep mind that this is a new, recently instituted process. During the interview process when these researchers were hired, this policy did not exist. Perhaps they were na…

The policy existed, management just assumed that they would implicitly understand that they shouldn’t work on research that will knee-cap Google. Of course, they also hired unhinged woke employees so it’s no surprised they’re reaping the seeds they sowed

The Reuters article says exactly the opposite. If you have documentation that this policy has been in place but ignored, I think that would have a big impact on this unfolding story, it definitely strike me as newsworthy.

The Reuters article clearly refers to this as a new change in policy and procedure.

"Google’s new review procedure asks that researchers consult with legal, policy and public relations teams before pursuing topics such as face and sentiment analysis and categorizations of race, gender or political affiliation, according to internal webpages explaining the policy."

"Studying Google services for biases is among the “sensitive topics” under the company’s new policy, according to an internal webpage. Among dozens of other “sensitive topics” listed were the oil industry, China, Iran, Israel, COVID-19, home security, insurance, location data, religion, self-driving vehicles, telecoms and systems that recommend or personalize web content."

Re: Google told its scientists to ‘strike a positive tone’ in AI research

#213

Post by joshuamorton »

mlthoughts2018 wrote: post #187
daniel957 wrote: post #81

Earlier quoted context omitted.

Why did you find the mention of Timnit Gebru ludicrous? The issues are closely related. In fact, I was _expecting_ the article to mention it. Google tells Researcher 1 to re-word things in X way for Y reason. Google tells Timnit Gebru to re-word things in X way for Y reason. The X and Y are the same. Based on the rest of your response, I believe you are having trouble reading the article in an objective manner. You s…

You appear to be mistaken on nearly all the facts about the Gebru situation. Given this it seems clear we would draw very different conclusions. Anyone claiming Jeff Dean lied must be so disconnected from the actual facts and so prejudiced to blame Google and view Gebru’s actions as acceptable that their opinion amounts to nothing but partisan gainsaying. To be clear, it’s ok to disagree with Jeff Dean or to dispute…

It really isn't. He lied. Normalize calling out corporate executives when they lie.

I wish he hadn't. I'm disappointed that he did. But I don't know how else to describe sharing statements one knows (or should know) to be false.

Re: Google told its scientists to ‘strike a positive tone’ in AI research

#215

Post by vl »

cmiles74 wrote: post #166
bite_tongue wrote: post #107

Seems reasonable for a company to not fund research that directly goes against their bottom line.

If Google did not want to fund research that conflicted with their bottom line, they could have avoided hiring researchers interested in that kind of work. Instead they hired these people and told them they would be free of censorship. Now that these people work there and depend on Google for their paycheck, Google is instituting this new policy. During a global pandemic. Sounds like censorship to me: hire the people…

> and told them they would be free of censorship

Of course they never really said it for simple legal reasons, when you sign the employment contract you actually agree to a lot of censorship.

Re: Google told its scientists to ‘strike a positive tone’ in AI research

#216

Post by eternalban »

abdullahkhalids wrote: post #211
jchallis wrote: post #146

Every specialty has taboos and learning what they are and how to avoid them is part of being successful with the job. As a physics graduate student I made an impassioned case to a high energy researcher that the money spent on high energy research was low yield and should be pushed to the biological sciences ... suffice it to say he did not take my points on their merits. Upton Sinclair’s famous “ It is difficult to…

I am a physicist, and I would say that, perhaps, you don't understand why we "waste" money on high energy physics. It's very similar to the reason why rich people buy useless status symbols, when they could instead buy useful things. It's essentially telling everyone, "I am so rich I don't care about money anymore and can afford to waste some of it on this useless thing". In science, we as a society throw money at so…

Your [1] should add that “additionally, there is a huge funding funnel that select companies can tap to without having to produce any results at all!” So funneling money to friends and special (dark) projects is also on the table, as a simple matter of what is “possible”.

Re: Google told its scientists to ‘strike a positive tone’ in AI research

#217

Post by gonehome »

lacker wrote: post #205

This whole saga is such a PR “own goal” by Google. The vast majority of corporate research labs just don’t fund research doubting their own ethics in the first place. Company leadership should choose an ethical stance and defend it. It’s an abdication of responsibility to say, well the ethics are an open question, we’re going to fund some research into it. And now Google is being punished for their abdication of resp…

A lot of the stuff really is an open question though.

For example if you’re training models on historical data to determine credit worthiness, you can end up with models that penalize people based on race, or zip code (as a proxy for race), or names (same thing again).

How to correct for this is not obvious, removing the fields from the data usually just has it appear in some other way that’s correlated to race.

My understanding of current approach is to leave the fields in and then check to see how out of whack things are and there are a few approaches to try and correct for this (also an open question).

The reason modeling risk on race is unfair is because while the color of your skin may be strongly correlated with being a bad (or good) credit risk due to a whole host of complex historical reasons that lead to bias in the data, it’s not because of someone’s skin color that they should be considered a bad (or good) credit risk. You want the models to be judging based on more causal details. Not data that’s biased due to historical favoritism or exclusion.

Doing this poorly can further cement the pattern in the data and make it worse over time in addition to just being less accurate.

That said, you want people working on this looking for pragmatic solutions and not just promoting their own agenda while spinning things to seem worse than they are with emotionally loaded language targeting the company. Making demands and then going public on Twitter about your failed ultimatum doesn’t inspire much confidence.

Re: Google told its scientists to ‘strike a positive tone’ in AI research

#218

Post by kentlyons »

tempodox wrote: post #132
bite_tongue wrote: post #107

Seems reasonable for a company to not fund research that directly goes against their bottom line.

When the result is predefined it's not research anymore, it's advertisement. Still calling it “research” is fraud.

For almost any paper I read, before I pick it up, I know there were statistically significant findings. Those results were not predefined. Instead they are part of the academic review process that filters out null results. That same process (depending on the field) also rewards novelty or all sorts of other things. It's very much a human process.

Re: Google told its scientists to ‘strike a positive tone’ in AI research

#219

Post by hfkldjsjfkdj »

The whole topic around bias in AI is overblown. If anything is to come out to the real world and fails to be inclusive then Google will be swept off their feet by the competition and as long as that is the case then I don't see what I need to be so cautious about.

And problems such as political bias run far deeper than some recommendation system. I feel as though these things are just trendy to talk about.

Post reply on HN