robryk @robryk@qoto.org

@rysiek

ISTM that the crux of your argument is that "toxic" has the same meaning in both cases. Is it? If so, why do you think it does? (E.g. I would expect that for advertising purposes Twitter does not care about toxicity via DMs.)

@rysiek

I don't think they need to understand what toxic is for the purposes they have. Given (c), they are working in a nonadversarial situation, because those accounts have no incentive to try to confuse this process. Thus, they can be using some very poor proxy for "toxic" as their definition of the thing they want to detect.

IOW anti-abuse measures are much harder than abuse-detection measures, even as far as making it harder to define the problem.

@rysiek

Putting aside whether I agree with the conclusion, I wanted to point out that the evidence you provided for this claim is not evidence for this claim: being able to have high-false-positive-rate detector of toxicity (for some definition thereof, which matters less here due to high FPR rate) isn't much (any?) evidence for being able to have a lower-FPR detector of toxicity.

(Not putting it aside, I think that it's immaterial whether they _can_ do it: they chose to put themselves in a position where they need to do it, so inability is not a good excuse for anything.)

@foone Note that adjusting visibility to unlisted does not affect people who follow you (it only affects local and federated timelines).

@agnieszka @kuba @kukrak @arek @icd

Co to jest t9?

@rysiek

You can accept a much higher false positive rate for refusing to advertise compared to refusing service to an account, though. Or what intervention were you envisioning they could apply?

@agnieszka @kuba @kukrak @arek @icd

Strzelam, że chodzi Ci o cel typu "dochodzenie roszczeń bez ryzyka utraty koniecznych do tego danych przez pożar w siedzibie, bez kosztów większych niż <coś>, etc.", nieprawdaż (wpp. np. ewidentnie trzebaby trzymać te dane w odizolowanym od Internetu systemie)? Czy istnieje już jakieś coś-typu-precedens na temat tych dodatkowych własności celu w tej sytuacji?

Szczerze mówiąc moje pesymistyczne przewidywania tego jak to wyewoluuje mówią, że będziemy nierównie skrupulatnie traktować różne zagrożenia dla danych użytkowników i będziemy mieli coraz więcej mechanizmów które nie mają żadnego praktycznego skutku (bo brzmią jakby miały taki mieć, albo taki mają ale tylko w innych sytuacjach) :(

@b0rk If your log statements have identifiers that get logged then you might be able to see whether a given log statement was being triggered before the problem manifested. Alas, this is rather a recommendation for tooling one might wish to build ahead of time.

@jhertzli But then how do they verify them? You ~can't verify them all from first principles, so it encourages reliance on authorities. (Maybe I'm overly pessimistic and these plausibly-sounding claims will still be self-inconsistent in noticeable ways.)

Alas, I think a larger problem is that verifying them, however you do that, takes time. It's now easy to generate such harder-to-recognize spam.

@TheAndSys Which game(s) do you mean?

@retr0id Have you worked out whether damaged sectors are laid along some kinds of paths on the surfaces of the platters? (Or are there too few of them for that to be meaningful?)

The faults it has make it a very depressing development for me. It's able to produce very human-sounding prose and is unable not to inject false statements into ~anything it outputs.

First, this is very depressing for people who actually read what they see and remember small tidbits that were mentioned. This makes it way more likely that they're garbage.

Secondly and more importantly, this is asymmetric tool of disinformation creation. It can be used to generate plausibly-sounding wrong statements much more easily than it could be used to generate correct statements. There are people/organisations who wish to do former. Thus, this will remove many of the factchecking heuristics that work today. (This is a fundamental inescapable problem for factchecking of news, and still a bad problem for factchecking of statements about empirically available knowledge.)

I expect us to get way more hard-to-filter spam and not get much in return. I expect Sybil problem to start appearing in places where hardness of simulating a human prevented it from appearing.

@retr0id What mechanism would you theorise caused the dust to damage the platters?

@retr0id Huh. I'm really surprised that the thing that got damaged are the platters and not the head.

@retr0id Did you do something special to avoid getting dust inside?

@b0rk

objdump?

@kevlin The recent text-generation examples do not seem detail-focused at all.