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

301–307 of 307 posts

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

#301

Post by LoSboccacc »

Closi wrote: post #296
LoSboccacc wrote: post #63

Earlier quoted context omitted.

journalist should at the very least receive, see and validate documents veridicity. here it's "internal correspondence read to Reuters." from a highly politicized topic with a lot of disinformation being thrown around from both side, it's hardly sufficient. but apparently you all happy enough to live in a post factual society where journalist have no longer to validate sources' claims, so be it

Lol, this is how journalism has been done for decades - There is nothing 'post factual' about it.

yeah no, imagine watergate without the address book, the tapes and the bank records.

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

#302

Post by throw_m239339 »

asddubs wrote: post #100
throw_m239339 wrote: post #8

Google is a for profit corporation, not a college campus, news at 11.

therefore we're supposed to ignore their transgressions, since they're to be expected?

Getting a job at Google and expecting academic freedom is certainly a pipe dream.

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

#303

Post by asddubs »

throw_m239339 wrote: post #302
asddubs wrote: post #100

Earlier quoted context omitted.

therefore we're supposed to ignore their transgressions, since they're to be expected?

Getting a job at Google and expecting academic freedom is certainly a pipe dream.

Yes, I agree it's not surprising that this happened. Doesn't mean we should stop paying attention or caring.

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

#304

Post by chaboud »

blackbrokkoli wrote: post #121
chaboud wrote: post #33

Earlier quoted context omitted.

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…

This is such a tired argument. Sure, politics is in everything down to the design of the armchair you are arguing from. That however doesn't fucking mean that all papers shall be considered equal? Do you think you would be able to differentiate the value of a census statistics from a 2015 paper written by the state of Nevada from a 1930 census statistic paper from Nazi Germany? Excellent! Because very obviously there…

I’m saying that papers should be read with a default position of skepticism and consideration of the motivations of the authors (and they all have an angle). When I read an ML paper with authors from Google, I assume they have an angle. When I read an ML Ethics paper with authors from Google, that already went double. Now it’s triple.

So look at my argument differently. Every single scientific paper you read should be read as non-objective as a starting position. If it can overcome starting from that position with data, methodology, and reasoning, it’s a solid scientific paper.

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

#305

Post by chaboud »

benlivengood wrote: post #221
izacus wrote: post #69

Earlier quoted context omitted.

The thing is - I don't think the bias in AI is fixable. The way we fix it is to regulate (not self-regulate, Google, Apple, Facebook and other AI companies aren't trustworthy to do it right) where AI is allowed to be used and how much of a negative impact of a persons livelihood can it impact. In other words, we need to use all the power we have to avoid "The computer said NO, you're homeless/jailed/poor forever now"…

Unfortunately a legal solution is weaker because it has to enumerate the precise conditions that are unacceptable. Do you ban the logistic function? Tanh? Statistics? Actuarial tables have been a part of decision-making for a very long time and it's not possible to form a comprehensive legal statement of "all decisions must be made in good faith with respect to inherent biases and to the most rigorous epistemological…

Funny enough, Aggregability is NP-Hard, so comprehensive consideration of the larger set is not actually practical/possible.

Here's the proof: http://www.cs.utep.edu/vladik/2006/tr06-29.pdf

Perhaps there are mechanisms around performance under external test, auditing, or safe harbor due to documented good-faith, but it is not possible to guarantee correctness for the full set from a subset.

Similarly, some approaches will plainly fail to perform well with subgroups of people due to sensing technologies. That may appear as model bias, but it's really a matter of physics and available entropy.

For the time being, mechanisms that use machine inference as an aid rather than a guide (leaving the judgment to the human in the loop) are probably the best we can hope for.

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

#306

Post by chaboud »

kovac wrote: post #97
chaboud wrote: post #33

Earlier quoted context omitted.

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…

> every paper has an angle Not really sure if this is true in general. May be true of commercial entities, but I really hope this isn't true in general. > Furthermore, ceding the technological lead and removing the business incentive likely has defense (read: offense) implications on the global stage. Does it though? Uncovering faults and cracks of a system does exactly the opposite. Discouraging experts from looking…

I don't think that uncovering faults and cracks is a disincentive. Prohibiting use of machine learning techniques and technologies definitely reduces the addressible market, and, thus, profit incentive.

There are plenty of reasons to do research and make strides, but "big pot of gold at the end of the rainbow" is one we have to acknowledge the existence of.

As far as paper angle goes, sometimes the angle is "what we spent the last year doing is noteworthy enough to warrant publication".

Read every paper skeptically. Make it prove to you that it has the data, rigor, and reproducibility to support its claims.

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

#307

Post by silexia »

ludamad wrote: post #133
tomcam wrote: post #98

Earlier quoted context omitted.

Grant-land is very very often about image and framing. How often are grant administrators interested in pure research and let the chips fall where they may vs getting results that yield happy articles, then more grants?

Agreed, the bias towards culling negative results is not unique to Google, although the mechanism may be

There is a witch hunt against Google that is politically motivated happening now.

If Google is funding its own research, it should be able to look at whatever it wants. Why do we hold Google to a standard that most other companies are far worse at following?

The appropriate place for independent critical research is outside of the company itself.

Google is actually a pretty well intentioned entity to put out as much self critical research as it does.

Post reply on HN