Hacker Newsnew | past | comments | ask | show | jobs | submit | lookingdesk's commentslogin

Jeff Dean liquidated Google's entire AI ethics team because they wouldn't revise an academic publication to align with the corporate PR spin on AI.

This is overstating it by a lot. Jeff was the AI lead at the time, and there was a big conflict between management and the ethics team

And I actually think Google needs to pay more attention to AI ethics ... but it's a publically traded company and the incentives are all wrong -- i.e. it's going to do whatever it needs to do keep up with the competition, similar to what happened with Google+ (perceived competition from Facebook)


You mean he fired one person who threatened to quit if the changes to the paper weren't to their liking? Or am I misremembering?

you are not misremembering.

As far as I can tell, no one seems to think much of value was lost.

even at the time that was the verdict

ai ethics was/is useless. it felt a lot like the movie industry of the 1930s saying they will police themselves just to keep any bigger regulator away

To say "AI ethics is useless" is itself useless.

Morality is not there to be useful, right or wrong in moral sense are normative categories not utilitarian ones.

But what you possibly may mean is really AI ethics self-regulation by large tech corporation does not work. (If that was your intended statement, I'd agree.)


... And then they didn't!

for good reasons.

It turns out when you usher the dissenters out of the room, you hear a lot less dissent.

It’s not like they were summarily executed. They are much freer now outside of corporate control to speak their minds as they please.

The only real question is if anyone deems them worth listening to.


I mean, I follow them.

Their problem domain is pretty niche and they're pretty technical on it so I don't doubt their traction is mostly academics.


That incident was the first time I ever heard of Jeff Dean and remains the main thing I associate him with.

Isn't it crazy how the media can do that? It really doesn't matter how much good you do in the world if your enemies speak louder. At least in the public's perception.

Not sure if he was told to stop the paper and fire the team, or whether he decided that himself.

But in any case, it's a stain on an overwise exceptionally brilliant career with wonderful software engineering achievements.

I wonder if they ever wondered if they'd do it again?


I'm having trouble following what exactly this means.

So Jeff wanted the team to modify an existing publication to fit the PR spin on AI, the ethics team refused, and Jeff dissolved the team?



reading about this incident years later, i gotta say, dang Gebru was right on a lot of things

https://www.technologyreview.com/2020/12/04/1013294/google-a...


Just another Jeff Dean optimizing out unneeded code story.

The Hollywood development company that bought the rights to Skibidi and are developing it filed a DMCA strike against Garry's Mod. It got resolved, but no one involved is talking.


Additionally, news came out today that the original creator has lost creative rights to Invisible Narratives. It's looking quite grim all around.


Oh no. I just read up on this, it sounds awful.

https://www.thegamer.com/skibidi-toilet-creator-legal-disput...


I checked the TIFF talk page and found comments from:

https://en.wikipedia.org/wiki/User:Scarlsen

Turns out the answer was on Wikipedia already :).


Thanks! If you look at his (logged-in) edits on Wikipedia (https://en.wikipedia.org/wiki/Special:Contributions/Scarlsen ), then apart from the lone comment on the talk page (about the reason for "42") and creating that user page, he has two edits to the TIFF article:

- one of them clarifies the (non-)involvement of Microsoft: https://en.wikipedia.org/w/index.php?title=TIFF&diff=prev&ol...

- and the other is even more interesting: though he is being scrupulous and removing a sentence that has no published citations, in his edit summary he confirms that it is basically true:

> The author of the original TIFF specification wanted TIFF to stand for "The Image File Format", but he was overruled by Aldus' president Paul Brainerd on the grounds that it sounded presumptuous.

(The edit summary says: Removed the "The Image File Format" sentence, since it only has eye-witness support (me, for one), but no published citatations)


Ok so then we could technically edit it back in since he's a primary source, right?


It's been a few years since I edited Wikipedia seriously, but the criterion for inclusion on Wikipedia is/was “verifiability, not truth” (https://en.wikipedia.org/wiki/Wikipedia:Verifiability,_not_t...) – what matters is not whether something is true, but whether it has been published in a reliable source (https://en.wikipedia.org/wiki/Wikipedia:Reliable_sources). Accordingly, Wikipedia tries to be based on secondary sources (rather than primary and tertiary ones). The relevant section (https://en.wikipedia.org/wiki/Wikipedia:No_original_research...) says, among other things:

> Primary sources that have been reputably published may be used in Wikipedia, but only with care

and I imagine a Wikipedia edit summary does not count as a reliable source. (For one thing, despite it being very plausible that the Wikipedia user Scarlsen who signed himself as Stephen E Carlsen is indeed that person—I believe it completely!—it cannot be guaranteed that it wasn't an impostor, for example.)


That would be what Wikipedia calls "original research". A big no-no on wikipedia. At a minimum he would have to tweet or blog about it and link the tweet or blog. And even then that's a primary source, which wikipedia considers less valuable. Ideally he would get someone else to report on his tweet/blog and use that as source. Then the wikipedia gods are happy

https://en.wikipedia.org/wiki/Wikipedia:No_original_research


So can we use this conversation on HN as a secondary source, and edit the deletion back in citing Hacker News?


Technically yes, but I'm fairly sure Wikipedia wants cited sources, not "I'm the guy, I said so" anecdotal sources.

Of course, if he was still alive he could have written a blog post or something like that and use that as a source, much like how it's likely this blog post will be used as a source for things surrounding the format and person.


His lone comment:

>Yes it is true: the second word of a TIFF file, 42, was indeed taken from the answer to the Ultimate Question of Life, the Universe, and Everything, from Hitchhikers_guide_to_the_galaxy. StephenECarlsen 23:38, 12 October 2007 (UTC)


If anyone can contact John Buck this sounds like information he'd be interested in. Also an interesting avenue for future investigative work.


thanks adam


Hey John, I'm just curious how people find these comments about "would be nice if X saw this" on HN. I don't think there's any pinging behavior. Did somebody message you? Did you just happen to read it? Do you have an eldritch curse that summons you when called by name?


Somebody subscribed to my blog with ref to hacker news so i just poked my head in :-)


I use f5bot.com, which monitors HN, Reddit, and Lobsters. (I have no connection to f5bot except being a happy user.)


Not who you're asking for, but generally I think it's just a case of the author also being an HN regular. Although, I suppose you could set up some Google Alerts for mentions of your blog posts.


Once upon a time there was a guy who went by Kibo who would search Usenet feeds for posts mentioning his username and reply to them...


I bet Musk hacked something together (or has a column in TweetDeck if that's still around) that continuously searches for mentions. I wonder if there's a tool like that that covers more of the internet, although the primary users would only be famous people and/or their agents / social media staff.


There are lots of tools that do this if you are willing to pay for it. "Social Media Monitoring" or "Brand Monitoring" would be the keywords to search for


You can try our tool for example: https://KWatch.io

It monitors keywords mentioned on X, Linkedin, Facebook, Youtube, Quora, Reddit, and YC.


Kibo greps


From what I've seen, sometimes people see that their blog post got a lot of referral traffic from Hacker News and they come have a look to see where it came from.


I've seen Ken Shirriff do this. You mention him and suddenly he's there.


hi! :-) As I commented elsewhere, f5bot.com is a nice, free tool to monitor HN


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: