TechTakes

He wants an AI that handles data security breaches by itself. He also now writes emails with ChatGP

He is the data security breach.

Instead of improving LLMs, they are working backwards to prove that all other things are actually word prediction tasks. It is so annoying and also quite dumb. No chemisty isn't like coding/legos. The law isn't invalid because it doesn't have gold fringes and you use magical words.

This is just standard promptfondler false equivalence: "when people (including me) speak, they just select the next most likely token, just like an LLM"

The problem is that there could be any number of possible next words, and the available results suggest that the appropriate context isn't covered in the statistical relationships between prior words for anything but the most trivial of tasks i.e. automating the writing and parsing of emails that nobody ever wanted to read in the first place.

None of these fucking goblins have learned that analogies aren’t equivalences!!! They break down!!! Auuuuuuugggggaaaaaaarghhhh!!!!!!

if a user on lemmy.ee want to post on lemmy.world, then lemmy.ee have to pay 10 lemmy.world coin to lemmy.world

Note that you don't need cryptocurrencies for this. I think Jaron Lanier talked about an idea like this ages ago, before people tried to put cryptocurrencies into everything.

1 post 6 comments joined 3 months ago, "i'm naive to crypto" "I want to host an instance that serves as a competitive alternative to Facebook/Threads/X to the users in my country,"

yeah he doesn't even have to charge for interacting with him i'll avoid him without it

that's lemm.ee last time i've checked. he made that mistake 14x

Yes crypto instances, please all implement this and "disallow" everyone else from interacting with you! I promise we'll be sad and not secretly happy and that you'll make lots of money from people wanting to interact with you.

Is he inscrutable/obscurantist on purpose, or is it because he never had a proper humanities education nor an editor?

This reminded me of that prediction I made w.r.t the "AI Doom" criti-hype (and touched on after SB 1047 popped up) back when OpenAI was gunning (heh) for DoD dollars.

Personally, I suspect that this might provide another case of "AI doom" becoming a double-edged sword for the AI industry. What can be dismissed as a simple error on their products' parts gets potentially a lot more problematic to deal with when a vocal minority is primed to find malice where none exists.

I literally just saw a xitter post about how the exploding pagers in Lebanon is actually a microcosm of how a 'smarter' entity (the yahood) can attack a 'dumber' entity, much like how AGI will unleash the diamond bacterium to simultaneously kill all of humanity.

Which again, both entities are humans- they have the same intelligence you twats. Same argument people make all the time w.r.t. Spanish v Aztecs where gunpowder somehow made Cortez and company gigabrains compared to the lowly indigenous people (and totally ignoring the contributions of the real super intelligent entity: the small pox virus).

AI misalignment leads to spinal misalignment.

If only we had paid attention to the roomba hitting us in the leg. It wasn't adorable, it was a murder attempt!

Obligatory note that, speaking as a rationalist-tribe member, to a first approximation nobody in the community is actually interested in the Basilisk and hasn’t been for at least a decade.

Sure, but that doesn't change that the head EA guy wrote an OP-Ed for Time magazine that a nuclear holocaust is preferable to a world that has GPT-5 in it.

nobody in the community is actually interested in the Basilisk

except the ones still getting upset over it, but if we deny their existence as hard as possible they won't be there

nobody in the community is actually interested in the Basilisk

But you should, yall created an idea which some people do take seriously and it is causing them mental harm. In fact, Yud took it so seriously in a way that shows that he either beliefs in potential acausal blackmail himself, or that enough people in the community believe it that the idea would cause harm.

A community he created to help people think better. Which now has a mental minefield somewhere but because they want to look sane to outsiders now people don't talk about it. (And also pretend that now mentally exploded people don't exist). This is bad.

I get that we put them in a no-win situation, either take their own ideas seriously enough to talk about acausal blackmail. And then either help people by disproving the idea, or help people by going 'this part of our totally Rational way of thinking is actually toxic and radioactive and you should keep away from it (A bit like Hegel am I right(*))'. Which makes them look a bit silly for taking it seriously (of which you could say who cares?), or a bit openly culty if they go with the secret knowledge route. Or they could pretend it never happened and never was a big deal and isn't a big deal in an attempt to not look silly. Of course, we know what happened, and that it still is causing harm to a small group of (proto)-Rationalists. This option makes them look insecure, potentially dangerous, and weak to social pressure.

That they do the last one, while have also written a lot about acausal trading, which just shows they don't take their own ideas that seriously. Or if it is an open secret to not talk openly about acausal trade due to acausal blackmail it is just more cult signs. You have to reach level 10 before they teach you about lord Xeno type stuff.

Anyway, I assume this is a bit of a problem for all communal worldbuilding projects, eventually somebody introduces a few ideas which have far reaching consequences for the roleplay but which people rather not have included. It gets worse when the non-larping outside then notices you and the first reaction is to pretend larping isn't that important for your group because the incident was a bit embarrassing. Own the lightning bolt tennis ball, it is fine. (**)

*: I actually don't know enough about philosophy to know if this joke is correct, so apologies if Hegel is not hated.

**: I admit, this joke was all a bit forced.

navigating this space without alignment with larger corporate players

stares into middle distance, hollow laugh

No need for xcancel, Gebru is on actually social media: https://dair-community.social/@timnitGebru/113160285088058319