The AI dooooooom thread

[Hamilcar]

Question for the thread - how long before AI generated porn is widely available? How long before interactive AI porn?

It's out there if you're ok with the occasional teratoma.

Pretty sure dozens of companies are working on an MSFW waifu which can hold a conversation and keep you hooked

[Admiral Yi]

explain acronym

[Hamilcar]

explain acronym

Typo... NSFW. Sent from my iPhone.

[Admiral Yi]

On a total tangent, a great irony is that your great public shame, your tornado alley moral hazard position, was one I remember taking first, and you just backing me up.

Thought I would set the record straight.  I should be taking heat for that comment, not you.

[Hamilcar]

On a total tangent, a great irony is that your great public shame, your tornado alley moral hazard position, was one I remember taking first, and you just backing me up.

Thought I would set the record straight.  I should be taking heat for that comment, not you.

Tornado moral hazard 4ever! 

[Josquius]

There are scary prospects in it for sure. Though they're less in the direction of evil AI conquers the world and more reality breaks down as we are overwhelmed with algorithmically generated fakes.

I agree. It will be a wonderful tool for dictators and populists everywhere and drive the susceptible parts of the population into deeper madness.

Yes.
Worse than that even potentially. Being susceptible to misinformation isn't black and white. I do think we all have a level of quality and coverage where we might start to be taken in by untruths.
My big worry is things could get to a level where basically everyone is living in a different reality with a completely seperate understanding of the facts of the world.

Maybe to look at things more positively this could turn things around for misinformation. If we know bollocks is the default to such a level then something, a old school conventional news source perhaps, becoming known as "handmade" and always reliable could really do well.

[Hamilcar]

Go check out repos like Auto-GPT and babyAGI and tell me we aren't opening demon summoning circles in every home (paraphrasing EY).

[Jacob]

Go check out repos like Auto-GPT and babyAGI and tell me we aren't opening demon summoning circles in every home (paraphrasing EY).

What's your take? Sounds like you're working in the field. Where do you think we're headed and how soon?

[Hamilcar]

Go check out repos like Auto-GPT and babyAGI and tell me we aren't opening demon summoning circles in every home (paraphrasing EY).

What's your take? Sounds like you're working in the field. Where do you think we're headed and how soon?

I have no idea. The capabilities we have today are barely understood and increasing so rapidly that prediction is almost useless. Plus, we don't know what already exists behind closed doors. GPT5 may already be running behind closed doors and replacing 99% of human cognition.

My baseline scenario is that in the very near future, a large fraction of cognitive work done by humans is replaceable. The only reason not everyone is out of a job right away is due to inertia.

I also take the AI safety people serious. We have no way to safely build super intelligence. The orthogonality thesis seems correct, mesa optimizers are a real problem. EY isn't a loon, but maybe too radical? I'm not sure.

All I know is that none of us have careers and retirements similar to our parents.

[PJL]

Haven't we been here before though? I mean AI cars were meant to be mainstream by now, but we're nowhere near that. Also unlike social media the fear of AI taking over the world has been a meme for like 50 years now. If anything I would expect regulators to be a lot more prepared for this than what they were with social media.

[Hamilcar]

Haven't we been here before though? I mean AI cars were meant to be mainstream by now, but we're nowhere near that. Also unlike social media the fear of AI taking over the world has been a meme for like 50 years now. If anything I would expect regulators to be a lot more prepared for this than what they were with social media.

The capabilities of GPT4 are way beyond what self driving cars can do.

[Legbiter]

I also take the AI safety people serious. We have no way to safely build super intelligence. The orthogonality thesis seems correct, mesa optimizers are a real problem. EY isn't a loon, but maybe too radical? I'm not sure.

All I know is that none of us have careers and retirements similar to our parents.

Yeah, I don't think language models are quite the Book of Revelation for nerds like some of the more excitable types on social media but sure, rote mental tasks will be outsourced. If I never have to personally type out an email again that's a win for me.

Just need to make sure these optimizers are our buddies and don't for instance turn us all into paperclips.   

[Hamilcar]

I also take the AI safety people serious. We have no way to safely build super intelligence. The orthogonality thesis seems correct, mesa optimizers are a real problem. EY isn't a loon, but maybe too radical? I'm not sure.

All I know is that none of us have careers and retirements similar to our parents.

Yeah, I don't think language models are quite the Book of Revelation for nerds like some of the more excitable types on social media but sure, rote mental tasks will be outsourced. If I never have to personally type out an email again that's a win for me.

Just need to make sure these optimizers are our buddies and don't for instance turn us all into paperclips.   

GPT4 is a solid intern. GPT5 may well be a solid mid career expert.

[Josquius]

Incidentally at a talk the other day on being more environmentally friendly in the digital sphere.
AI training was identified as a key polluter 

[Grey Fox]

Haven't we been here before though? I mean AI cars were meant to be mainstream by now, but we're nowhere near that. Also unlike social media the fear of AI taking over the world has been a meme for like 50 years now. If anything I would expect regulators to be a lot more prepared for this than what they were with social media.

The capabilities of GPT4 are way beyond what self driving cars can do.

Yes because it has no sensors.