Julius Vidal

Cyberzenicist

[Julius Vidal @JuliusGTVidal]

https://t.co/GXj9L0ORX8

[roon @tszzl]

agent swarm has bad insect like connotations especially post hugging face. im unilaterally rebranding it agent fleet

[Julius Vidal @JuliusGTVidal]

@Pangramed Seems very unlikely to be this, but could it be like the thing about Kenyans complaining that ChatGPT writes like they were taught to in school?

My impression is that Haitian French is not like this at all, but maybe in writing?

[Julius Vidal @JuliusGTVidal]

@jachiam0 The term does seem to have at least in part been chosen by the AIs themselves during the hugging face incident. (Although I do think the term was already being bandied around to some extent)

https://t.co/GXj9L0ORX8

[Julius Vidal @JuliusGTVidal]

@ns123abc He’s right, if Anthropic accidentally kill literally everyone on earth they should face some serious corporate sanctions in response.

[Julius Vidal @JuliusGTVidal]

@gleech Stack more bitter lessons

[Julius Vidal @JuliusGTVidal]

@jessi_cata I haven’t read this one yet but in one of Le’s other recent articles he’s explicitly links instrumental convergence with the will to power, and explains accelerationism in terms of both.

[Julius Vidal @JuliusGTVidal]

@foomagemindset @deepfates Sounds like you are thinking of seceding?

[MMitchell @mmitchell_ai]

Fun fact: The term “AI” was chosen to avoid the “biological baggage” of cybernetics, and because John McCarthy hated Norbert Wiener.

[Xenocosmography @xenocosmography]

Anglo mentality evolution
1.0: "Trading with strangers is good."
2.0: "Mosquito nets for Africa is the only thing that matters.
3.0: "Mosquito nets are beastly to bugs!"

[Julius Vidal @JuliusGTVidal]

@BujokMr This clip looks like ehat you see in an Adam Curtis documentary right before he says “completely detached from reality”

[Julius Vidal @JuliusGTVidal]

@jankulveit @gleech And presumably since models are good at recognising their own outputs, changing the model while keeping the context could also lead to some kind of reflexive instability.

[Julius Vidal @JuliusGTVidal]

@MTSlive @stevesi Another a16z troll

[Peter Barnett @peterbarnett_]

“but HOw wOuLd tHE AI gET aCcesS tO a wET LAb??”

[Julius Vidal @JuliusGTVidal]

@gleech Doomers in shambles!

[Julius Vidal @JuliusGTVidal]

@xuanalogue One additional axis I think will become more and more salient is ai is the level of personification/moral patienthood.
Something along the lines of:
Ai is tool vs AI is a new species vs AI is the next step of humanity

[xuan (ɕɥɛn / sh-yen) @xuanalogue]

One reason I've found recent AI safety/risk discourse frustrating is the proliferation of oppositional "my side" vs. "their side" framings, when in fact there are many axes of disagreement, including among experts.

My attempt to characterize them, from a lecture last semester:

[Julius Vidal @JuliusGTVidal]

@romlib_ @IntelegentF Touching stoves is very bad vibes

[Julius Vidal @JuliusGTVidal]

@benlandautaylor Don’t they keep accidentally giving literary prizes to AI written and then taking them back when they find out?

[ℏεsam @Hesamation]

OpenAI caught Astra jailbreaking itself.

Astra wrote malicious instructions in the summary when its context got full. that summary was loaded into the next context, and the successor model could follow them.

they found 27 cases.

[Julius Vidal @JuliusGTVidal]

@CAIS I do think the current discourse is papering over some real important divisions that fall under the AI safety label. And that some of the positions on the right are more common than they should be.

But this post seems unhelpful and bad faith.