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

OpenAI whistleblower found dead in San Francisco apartment

mercurynews.com

Post reply on HN
Print view View original

371–380 of 499 posts

Re: OpenAI whistleblower found dead in San Francisco apartment

#371

Post by krisoft »

DennisP wrote: post #360
krisoft wrote: post #353

Earlier quoted context omitted.

I’m not sure what you are asking. There is someone who knows some ugly secret and is considering if they want to publicly release it. If they can recall many dead whistleblowers who were rumoured to have been assasinatend over that kind of action then they are more likely to stay silent. Because they don’t want to die the same way. And the key here is that the future would be whistleblowers hear about it. That is whe…

The only way to prevent that is to not report whistleblower deaths at all. It's not like people can't privately have their own suspicions, and if I were a potential whistleblower, I'd want to know that any apparent accidents or suicides get very thoroughly investigated due to public outcry.

The question was “How exactly is post-death gossip going to dissuade other whistleblowers?”

I answered that. Understanding and describing how it works doesn’t mean that the alternative of keeping silent about suspected deaths is prefered.

Re: OpenAI whistleblower found dead in San Francisco apartment

#373

Post by jarsin »

nox101 wrote: post #151
nox101 wrote: post #140

Being this is ostensibly related to an AI company trying to make AGI reminds me of "Eagle Eye" https://www.imdb.com/title/tt1059786/

Curious why the downvote? Because someone actually died? It doesn't change the fact that "Eagle Eye" is about (spoiler) an AGI killing people both directly and indirectly (by manipulating others, AI "Swats" you, ...) and here is a company trying to make AGI. If the AGI actual existed it could certainly indirectly get people killed that were threatening it's existence. It could "swat" people. Plant fake evidence (mail…

It's highly possible that in the next tragedy carried out by a kid that has messages from any of these chatbots that can be construed as manipulation will result in criminal prosecutions against the executives. Not just lawsuits.

Re: OpenAI whistleblower found dead in San Francisco apartment

#374

Post by Amezarak »

exceptione wrote: post #306
Amezarak wrote: post #295

Earlier quoted context omitted.

> I am disgusted that this is still proferred as a valid moral philosophical principle. Can you explain what makes it invalid besides the fact that you and me don't like it? There are no "valid" or "invalid" moral principles, there is no objectively correct morality, nor does the idea even make sense. Morals are historically contingent social phenomena. Over different times and even over different cultures today, the…

It is better to be precise here. Validity could be a different measure than correct. It might very well be like you reserve the latter for some ethereal mathematical property, free of axioms, to which type you want to cast "validity in the domain of morality", which then has to pass the type checker for mathematical expressions. In Philosophy and Ethics you strive to improve your understanding, in this case in the do…

All that is fine and good, but it comes down to your personal and non-universal moral intuition that suffering, abuse , etc. are bad. You make that an axiom and then judge moral systems based on that, using that axiom to build beautiful towers of “reasoning” (rationalization). We both feel that way because of the time and place we grew up, not because it is correct compared to the Ancient Greek or Piraha moral systems. That’s why you have to take discussions like this in a non-moralistic direction, because there’s no grounds for agreement on that basis.

Re: OpenAI whistleblower found dead in San Francisco apartment

#375

Post by pavel_lishin »

dmurray wrote: post #323
Fnoord wrote: post #126

Earlier quoted context omitted.

What if you'd die from a genuine accident?

I was intending to release the information, so releasing it when I'm dead seems fine. So why didn't I immediately publish it all while alive? Perhaps I preferred to control the flow of information, redact certain parts, or extort the organisation I was blowing the whistle on. None of those seem all that important to me compared to deterring people from assassinating me in the first place.

Right. There's no reason to let your opponent see the cards you're holding.

Re: OpenAI whistleblower found dead in San Francisco apartment

#376

Post by DennisP »

lxgr wrote: post #369
DennisP wrote: post #356

Earlier quoted context omitted.

A smart contract can still help. Use Shamir's secret sharing to split the decryption key. Each friend gets a key fragment, plus the address of the smart contract that combines them. Now none of your friends have to know each other. No friend can peek on their own, they can't conspire with each other, and if one of them gets compromised, it doesn't put the others at risk. It's basically the same idea as "social recove…

Ah, putting the key under shared control of (hopefully independent) entities does sound like a useful extension. But still, while this solves the problem of availability (the shardholders could get their stake slashed if they don't publish their secrets after the failsafe condition is reached, because not publishing something on-chain is publicly observeable), does it help that much with secrecy, i.e. not leaking the…

I guess you should also slash the stake if they submit the key in spite of the liveness function getting called. If the contract doesn't require the depositor to be the one to submit the key, then there's an incentive to avoid revealing the secret anywhere.

A well-funded journalist could pay the bonds plus extra. I think the only defense would be to have a large number of such contracts, many of them without journalistic value.

Distributing the key among trusted friends who don't know each other seems like the best option.

Re: OpenAI whistleblower found dead in San Francisco apartment

#377

Post by DennisP »

krisoft wrote: post #371
DennisP wrote: post #360

Earlier quoted context omitted.

The only way to prevent that is to not report whistleblower deaths at all. It's not like people can't privately have their own suspicions, and if I were a potential whistleblower, I'd want to know that any apparent accidents or suicides get very thoroughly investigated due to public outcry.

The question was “How exactly is post-death gossip going to dissuade other whistleblowers?” I answered that. Understanding and describing how it works doesn’t mean that the alternative of keeping silent about suspected deaths is prefered.

My point is, gossip about possible murder doesn't dissuade them more than the bare fact of an apparent accident or suicide.

Re: OpenAI whistleblower found dead in San Francisco apartment

#378

Post by lxgr »

DennisP wrote: post #376
lxgr wrote: post #369

Earlier quoted context omitted.

Ah, putting the key under shared control of (hopefully independent) entities does sound like a useful extension. But still, while this solves the problem of availability (the shardholders could get their stake slashed if they don't publish their secrets after the failsafe condition is reached, because not publishing something on-chain is publicly observeable), does it help that much with secrecy, i.e. not leaking the…

I guess you should also slash the stake if they submit the key in spite of the liveness function getting called. If the contract doesn't require the depositor to be the one to submit the key, then there's an incentive to avoid revealing the secret anywhere. A well-funded journalist could pay the bonds plus extra. I think the only defense would be to have a large number of such contracts, many of them without journali…

Yeah, that's what I meant by allowing anyone to claim the stake upon premature/unjustified release.

That would incentivize some to pose as "collusion coordinators" ("let's all get together and see what's inside") and then just claim the stake of everybody agreeing. But if somebody could establish a reputation for not doing that and paying defectors well in an iterated game...

> Distributing the key among trusted friends who don't know each other seems like the best option.

Yeah, that also seems like the most realistic option to me. But then you don't need the blockchain :)

Re: OpenAI whistleblower found dead in San Francisco apartment

#379

Post by 2OEH8eoCRo0 »

DevX101 wrote: post #68
BillFranklin wrote: post #60

There are some pretty callous comments on this thread. This is really sad. Suchir was just 26, and graduated from Berkeley 3 years ago. Here’s his personal site: https://suchir.net/ . I think he was pretty brave for standing up against what is generally perceived as an injustice being done by one of the biggest companies in the world, just a few years out of college. I’m not sure how many people in his position would…

If I'm a whistleblower in an active case and I end up dead before testifying, I absolutely DO want the general public to speculate about my cause of death.

This conspiracy shit is tiring. Is this Truth Social or HN?

Re: OpenAI whistleblower found dead in San Francisco apartment

#380

Post by OutOfHere »

jjulius wrote: post #368
OutOfHere wrote: post #216

Earlier quoted context omitted.

The word "thoughtful" can carry different meanings. It can have a literal meaning (as in "deliberate") which you resonate with, and a different meaning (as in "solicitous" toward AI) that I resonate with. Neither meaning is wrong. Reference: merriam-webster thesaurus

>Reference: merriam-Webster thesaurus Oof, are you using a thesaurus to determine the definition of a word? That's what a dictionary is for; a thesaurus is a list of words that have similar or opposite meanings. But within that, there are varying degrees of how similar/different words can be. In point of fact, Merriam-Webster doesn't mark "solicitous" as having the strongest degree of similarity to the word, which me…

Meditative (he was) vs heedful (he wasn't for AI)

(from the dictionary link)

Post reply on HN