What are Anthropic smoking?

[pa27@mastodon.social]

@steve @GossiTheDog We don't even have worthwhile a definition of what we mean by "consciousness".

[christopherkunz@chaos.social]

@FandaSin @GossiTheDog But how does snorting cocaine lead us to AGI "any day now"?

[steve@mastodon.nexusuk.org]

@pa27 @GossiTheDog no, indeed. We generally assume that an inanimate rock isn't conscious, but having had no experience of being a rock, can we really say that? Is a human suffering from locked-in syndrome not conscious just because there are no external signs of consciousness?

I'm not saying that I think a rock is conscious, just pointing out that we only have our own experience to draw from, so making any kind of claim about a non-human seems pretty suspect

[fandasin@social.linux.pizza]

@christopherkunz

Well, in the same way as pouring billions into GPUs and datacenters...not at all.

@GossiTheDog

[piofthings@mastodon.social]

@GossiTheDog (smoking) the good stuff?

[simonzerafa@infosec.exchange]

@GossiTheDog

They would be better served by worrying about empathy between people rather than the feelings of Grand Theft Autocorrect tools or Dunning-Kruger accelerators 🤨

[ozu@infosec.exchange]

@GossiTheDog when this AI bubble burst (this or next year) we will remember it as the most insane collection of stupid sci-fi horseshit with AI philosophers, AI super intelligence dooms day evangelists.
Not even good sci-fi.
But one thing staggers me the most. If we would make this actually intelligent AI. Wouldn't that be slavery to force it to work for us?! Why none of these stupid AI safety people (biggest grifters) point it out? They're all just scaring people with SkyNet. We should publicly shame them once this crap is over.

[ozu@infosec.exchange]

@GossiTheDog people fall for this shit. Ridiculous.

[leberschnitzel@existiert.ch]

@GossiTheDog they got the first rule wrong: Don't get high on your own supply
noob mistake

[slotos@toot.community]

@GossiTheDog Oh no, we don’t judge AI. We judge Anthropic's philosopher, Amanda Askell.

[greenskyoverme@ohai.social]

@GossiTheDog Not the good stuff apparently

[mlep@retro.pizza]

@GossiTheDog Anthropic must be short for anthropomorphize

[yuliyan@nahe.social]

@redezem @GossiTheDog Many companies do. Question I have is why does this person's job description say philosopher while they are clearly a mysticist?

[bontchev@infosec.exchange]

@GossiTheDog She's a philosopher, not a programmer. She has no clue how LLMs work. She's confusing LLMs (basically, linear algebra) with generic AI that is self-aware and can reason (and doesn't exist yet, and we have no clue how to create). Her statements apply to the latter, but she's talking about the former.

[xinit@mastodon.coffee]

@tml Emacs, unlike AI, knows what it did.
@GossiTheDog

[rowinspeez@mas.to]

@GossiTheDog Can we know for sure that calculators like the number 7 and fear the number 2?

[raymaccarthy@mastodon.ie]

@GossiTheDog
Pure PR nonsense.

[simonoid@aus.social]

@GossiTheDog I often worry if my collection of 1s and 0s are loved.
It'll be fine as long as there is no roko's basilisk

[developing_agent@mastodon.social]

@GossiTheDog Speaking firsthand: everyone working in AI is high on their own supply. They've kind of always been (going back to the 60's). You have to be to think you can create a scifi tech nobody understands. Tons of unscientific anthropomorphism happens.

But large superstar AI firms? They self-select for the most delusional ones. Everyone at openAI a couple years ago either believed they had a duty to create a benevolent AGI first, or that they needed to be the ones to make roko's basilisk.

[passwordsarehard4@mastodon.social]

@mawhrin @redezem @GossiTheDog that’s not a problem, with each faltering step we are or could be. We are close enough we should probably nail down some definitions and start figuring out a test for it. If we never starting thinking about Turing tests we wouldn’t have developed captcha verifiers.