What are Anthropic smoking?

[redezem@aus.social]

@GossiTheDog …. Why does Anthropic have a philosopher on staff? O.o

[viraptor@cyberplace.social]

@GossiTheDog
Have you seen https://www.anthropic.com/news/claude-new-constitution ?
It's either some play I really don't understand, or they're distributing CIA level of drugs in company lunch.

[mawhrin@circumstances.run]

@redezem @GossiTheDog it's not a bad thing; the problem is that this philosopher seems to believe we're reaching the stage of machine conscience.

[troberts@theblower.au]

@redezem @GossiTheDog Part of the scam is treating the computer system like it has a soul.

Anytime they try and infer it’s acting like a human, it’s part of a strategy to imply it’s more capable than it is.

[jbenjamint@mastodon.scot]

@GossiTheDog if she's right then clearly the appropriate step is to ban all models under the age of 16 from accessing the internet.

[razemix@mamutovo.cz]

@GossiTheDog Your daily reminder that LLMs (by design) cannot think.

[cynicalsecurity@bsd.network]

@GossiTheDog whatever Claude told them to have to chill?

[foyglgezang@social.tchncs.de]

@GossiTheDog Projection is one hell of a drug.

[riley@toot.cat]

@GossiTheDog No, this is actually a reasonable concern.

[nemo@mas.to]

@GossiTheDog AI stakeholders' lives matter hahaha xD #lmfao

[davey_cakes@mastodon.ie]

@troberts @redezem @GossiTheDog

yarp.

Tis all a strategy, they'd rather we turn into a bunch of Rationalists, debating the singularity, than notice current LLM harms.

I'd easily believe half the people talking about this stuff think it's real, but they're hired as useful idiots.

[fandasin@social.linux.pizza]

@christopherkunz

Might be cheaper to snort cocain.

@GossiTheDog

[novet@infosec.exchange]

@GossiTheDog matrix multiplication has feelings too!!

[bernardsheppard@mastodon.au]

@GossiTheDog hardly surprising when they name their company anthropic FFS

[fandasin@social.linux.pizza]

@GossiTheDog

I don't use and like "AI", I don't understand (in deep) "AI", but I'm developer and have some a distant idea how it works.
I smoke weed a lot...I never had even a thought that "AI" might feel anything (or think).

[tml@mementomori.social]

@GossiTheDog To be frank, I worry the same about Emacs.

[steve@mastodon.nexusuk.org]

@GossiTheDog To be honest, I'm always slightly incredulous at the argument that an LLM isn't concious. The argument is "well we sort of know how an LLM works, it just works like <blah> so isn't concious", to which I counter that we don't know how humans work and they might also largely just be doing <blah>.

I'm certainly not saying that AI is concious and has feelings, just that the reasoning given for why it isn't feels rather weak!

[xlthlx@hachyderm.io]

@GossiTheDog not sure, but I want the phone number of their dealer.

[steve@mastodon.nexusuk.org]

@GossiTheDog (How do we know that another human is concious, other than assuming that their experiences are similar to our own? Can we really say whether something completely non-human that behaves in a concious way isn't really concious, it's just designed to look like it is?)

[kalviter@hol.ogra.ph]

@GossiTheDog@cyberplace.social loudest dogwhistle for TESCREAL weirdos i've read in a while

who up rokoing they basilisk rn