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

71–80 of 307 posts

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

#71

Post by curiousllama »

frankfrankfrank wrote: post #46

I deal with this on a regular basis in the research world, there is what is a really toxic and dangerous bias towards constructing the positive, i.e., making things seem, appear, and even fabricate outcomes that will be affirming of desired outcomes or other group biases. It is not research, let alone science when you are striking a positive tone, it’s fraud.

This is a really important point. Executive-land is all about image and framing, so I can totally see how execs think it's not a big deal to "just make the case" or frame things less negatively.

But while that's ok in a slide deck, it's _not_ ok in a research paper. Researchers aren't (shouldn't be) accountable to business outcomes, but rather the truth of what they argue.

It's tough to see this play out real time.

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

#72

Post by pjc50 »

Shivetya wrote: post #57
drfuchs wrote: post #9

Consider the historical parallel to the tobacco industry’s research departments. And asbestos. And lead. And talc. And etc. The researchers found that the products were harmful, but their results were squelched. “Do no harm,” indeed.

tl;dr - if we don't create it, control it, and make sure everyone knows about it, someone else is going to make all those decisions for you. Its coming regardless These are not equivalent comparisons to me. Tobacco was always harmful, inhaling smoke in general is just bad, and the others could be through improper use but safe otherwise. Let alone all of what you mentioned will not have the impact AI will AI's potenti…

> someone else is going to make all those decisions for you

Well, yes, and in this case that someone appears to be Google, which is why people are complaining about them making AI decisions.

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

#73

Post by simion314 »

chaboud wrote: post #33
_cs2017_ wrote: post #14

IMHO, a scientist capable of striking any tone in their work, positive or negative, is doing marketing / PR work rather than science. There's nothing wrong with that. Just like any other member of society, a scientist may want to contribute to a (for-profit or non-profit) organization of their choice, be it for material or spiritual rewards. However, I think it's important not to get science and marketing confused. S…

When I was young and (more) idealistic, my department chair pulled me aside after I’d been dogmatic about something and said “everything is political... The sooner you realize that, the better off you’ll be.” It took 10 years for that to really sink in, but I knew that every paper has an angle well before that. There’s a position of inevitability about AI around Google, and, to an extent, I agree. Unlike tobacco, alc…

>Some things are going to be hard to explain. Some skin tones, hair styles, accents, pitches, timbres, writing styles, etc. may have more/less separable entropy, and that’s going to inevitably make models seem racist/sexist for some applications.

We all know that if you have garbage as an input the output can be anything. So a professional data scientist should do his job and analyze the data first , identify the problems and if it can't be clean go in the world and collect better data.

If you are a high paid professional you should understand that say for face recognition you should get a correct sample size, correct distributed (sorry I am not a professional so i don't know how they should do their job but I know that there are statistical tool that will measure things like this) , instead it seems developers use as input for their experiments some garbage data that they can get their hands on.

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

#74

Post by unishark »

hajderr wrote: post #24

Idea: AI bot for detecting positive tones in Google's papers.

How does it make you money?

My money-making idea from this is AI to detect sentences or fragments from your emails that could be taken out of context and made into embarrassing headlines if the email gets leaked.

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

#75

Post by ClumsyPilot »

mlthoughts2018 wrote: post #34

The “sensitive topics” extra review seems completely fair and obviously necessary. Anyone acting like that amounts to censorship in a private company is totally off their rocker. This article seems really weak, and of course there’s the obligatory attempt to tie it in with Timnit Gebru’s resignation and her very inappropriate behavior, even though for all of Google’s flaws, Google clearly and obviously did the right…

"Gebru’s weak paper"

Could you please back up your claim as to the paper's quality - do yoy have credentials and background that enable you to cast this judgement, or is this pedestrian?

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

#76

Post by ClumsyPilot »

mhh__ wrote: post #68
drfuchs wrote: post #9

Consider the historical parallel to the tobacco industry’s research departments. And asbestos. And lead. And talc. And etc. The researchers found that the products were harmful, but their results were squelched. “Do no harm,” indeed.

It seems to be an unfortunate reality of doing research for anything other than the sake of research. My physics teacher worked for some kind of government air pollution modelling group in the past, and effectively said that they would often be told "Here's the result, find us data to make it work".

It seems like corruption that does billions in damages and costs lives

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

#77

Post by pjc50 »

piokoch wrote: post #53
oblio wrote: post #21

Earlier quoted context omitted.

> The difference is that AI research teams are full of philosophy graduates who think if statements are racist. I'm not sure I understand. Are they worse or better?

That's an actually good question. AI is supposed to be fed with real-live data, after all it has to be something usable in the real-world. Now, you feed AI with crime data, looking at statistics such AI will suspect people having black color of the skin to be criminals more often than people with white skin. Is that ok? Is it ok to say that if some person happens to have black skin there is a bigger chance that it is…

> For ages Christianity was a source of ethics in Europe (less in US, that's why you could have slavery in modern times there)

This is completely incoherent and ahistorical; the US was in many ways more Christian, or at least more intensely Christian and making it a key part of its politics, right from the Pilgrim Fathers to televangelists.

People gave pro-slavery speeches from pulpits in both the US and UK. There were some Christians against it, most notably the Quakers, but slavery and the alleged moral justifications for it were in no way incompatible with Christianity.

Even today, US churches tend to be ("voluntarily") segregated.

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

#78

Post by kovac »

_Nat_ wrote: post #54
nisuni wrote: post #13

And that’s why I was happier when basic research was mostly being carried out in public institutions.

Public institutions tend to observe the same policies, even when there's no direct commercial interest. For example, if you join a Math department and start publishing papers that cast the department's research interests in a negative light, that's often a good way to get fired. I mean, seriously, go ask just about any professor of any topic at any university if their field is undervalued or overvalued. They'll almos…

General proposition may be true, but how can a math publication place the department interests in a negative light? Like "Doesn't matter how hard we try, there are always undecidable facts in our own system"? :p

I still have faith in the research done by hard science research teams.

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

#79

Post by saalweachter »

xiphias2 wrote: post #59
necheffa wrote: post #51

Earlier quoted context omitted.

I mean, when you have to go, you have to go. If you have to travel to an entirely different floor to find an open shitter, that makes for a poor experience.

Going 1 floor is fine...going through 7 floors, visiting every bathroom sucks. Anyways COVID solved the issue I guess.

I keep a list of things that would make me quit, and "shitting my pants in a stairwell looking for an open bathroom stall" is on that list.

I went through the floor map one day counting desks & toilets, verifying we were above OSHA minimums.

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

#80

Post by _Nat_ »

It's important to draw a distinction between positive tonality and censoring negative results.

For example, the recent COVID-19 vaccines: they're commonly reported to be 95% effective and have some risk of side-effects. This can be reported factually, or spun in a positive or negative light. For example, a sensationalist might try to stir up interest by spinning a narrative that focuses on the vaccines being unreliable and dangerous, as proven by science that Big Pharma doesn't want you to know about!

And as we've seen, sensationalists can get cult followings. However bad their ploy may be for society overall, the sensationalist themself may stand to profit.

So as reprehensible as censoring legitimate scientific research would be, a directive to "strike a positive tone" sounds like it could be a very different, far more sensible thing.

Post reply on HN