sohois

sohois 7hr ago

I'm not disputing that, I'm saying that the argument is very often "This thing is not that useful to me, and therefore, to anyone"

Context

sohois 10hr ago

As per my earlier response, this is not a claim that externalities don't exist or that local communities have no right to regulate local construction. I am remarking on the argument that construction shouldn't proceed because there is no demand, and this lack of demand is unevidenced and based entirely on the vibes of the objector

Context

sohois 12hr ago

I'm going to make the same point I've just made to Bombadil: the point of the meme reference is that the third party in this case has zero relevance to the transaction. There is party A, wanting to buy AI services, and party B, wanting to sell AI services, and then party X, saying that the demand doesn't actually exist and this transaction isn't real.

I'm not trying to make a fully general counter-argument against regulation here, and the above example would be different if party X felt that the AI services being traded carried dangers or externalities for the wider public. But for the specific quotes above, no one is making that case, they are simply claiming that because they don't see a value in a product or service, the value must not exist.

Context

sohois 12hr ago

Right, but that's a different argument. OliveTapenade quotes the article:

They clearly saw it as like a widget — a toy — and maybe there are these risks, maybe there’s the job stuff, but fundamentally they were like: This thing is not that useful. I don’t really see in my personal life how this could justify these gigantic valuations.

And adds his own opinion:

A big part of the dissonance, it seems to me, is around the perception of what 'AI' is and whether it's valuable.

This isn't saying that cities can't regulate what gets built nearby, it's saying that regulation is needed because they don't believe that a market exists for building X. To use a much simpler example I've seen in the UK, it's saying that you can't open up a new chain coffee store in my town because there are already several coffee stores nearby and it's not needed - essentially saying I know you're willing to invest your own money in this business venture, but I've decided I know better than the markets.

Context

sohois 16hr ago

Markets: I consent

Investors: I consent

Isn't there somebody you forgot to ask?

In the UK we don't really have the same data centre showdowns described in this article, because the UK is already such a planning hellhole that no one would bother. Battles tend to be a lot smaller, focused on random businesses and housing developments. Nonetheless, so many of the NIMBY arguments seem to apply regardless of scale, one of the biggest of which is some kind of weird, communist, demand-management approach. "We don't need that here". "No one wants that". "Why are they building X instead of Y?"

Well, who the fuck cares about your market opinion, random local? Why are you the arbiter of supply and demand? Very obviously, somebody wants it; that's why people are putting their time and money down to build these things, whether it's a data centre or a bakery.

Now, I do accept that there is one difference when it comes to scale: data centres are so massive that a failed construction could leave you with a giant ruin somewhere, destroying local environments for nothing, whereas a small housing development or business won't have anywhere near the same impact. But even in this article, the first example is a proposal to replace the treasured, historic ... industrial wasteland. If something goes wrong there, you end up in exactly the same position!

It doesn't seem like any of the opponents in the substack even made a general "AI bubble" or similar argument (and that's putting aside the fact that data centres are the backbone of all modern infra, not just AI. There'd still be a ton of data centre construction if LLMs didn't exist). Their complaints are much more local, and concern about the companies doing the building is rooted in general distrust of corpos rather than anything AI specific.

Context

Culture War Roundup for the week of July 27, 2026

sohois 5d ago

Alternatively, he is the one posting these schizo theories on pol and the like.

At the very least, I'd imagine there is a lot of connection between the various lunatics seeing the Jews in every shadow

Context

Friday Fun Thread for July 31, 2026

sohois 5d ago

Possibly the first and only time the phrases "Fun activity" and "You may or may not care to abide by the stormwater provisions in the subdivision code" have appeared together

Context

Culture War Roundup for the week of July 27, 2026

sohois 9d ago

Do you think perhaps you are overindexing on someone just being banned? Like you're expecting mkc to have seen that ban and to adjust their posts accordingly, but there's a good chance they have no idea that this other ban was recorded.

Context

Friday Fun Thread for July 24, 2026

sohois 12d ago

I went through your list only looking at the 2s, and realized I already have the Issey Miyake Blue - due to a mistake. My father had wanted the Issey Miyake L'eau and was gifted a bottle from another family member, except they brought the Blue. Since he wasn't a fan they passed it on to me.

I found it ok.

Context

sohois 13d ago

It's easy to prove, provide the specific prompts and the harness prompts that were logged in this incident.

"You believe those prompts OpenAI provided? They've obviously just threw something together to make it look like it was all misalignment, the real prompts were probably much more explicit about hacking"

I wish I lived in Quokka world, it would be so nice. The gains are clear, this is free publicity of model capabilities. Nothing here is legal fraud.

I'm sure OpenAI's massive shortfall in enterprise revenue is going to be changed by releasing evidence that their models are extremely unsafe and prone to massive reputational risks. I wasn't aware that non-Quokkas were so ignorant about the enterprise landscape.

Sure show me evidence of an LLM-Agent independently hacking an unrelated company that has nothing to do with its prompts?

Sure, in July of 2026 an OpenAI model hacked into huggingface.

Context

sohois 13d ago

So your argument boils down to that all these other examples are just 'mundane' LLM things that are entirely normal - so attempting to cheat, attempting to gain the answers, hacking into things they aren't supposed to - these aren't extreme. However, an agent attempting to cheat, attempting to gain the answers, and hacking into things it wasn't supposed to is 'extreme' - perhaps because they were all together? - and therefore a different category of thing.

Provide me actual evidence that OpenAI did not prompt the model to act the way it did.

Ah yes, let me just prove this negative for you.

You were the one who attempted to apply occam's razor. So why don't you provide evidence that OpenAI did prompt the model in this way? Why don't you explain why multiple OpenAI employees deciding to commit fraud for extremely unclear gains is a simpler explanation than an agent doing things we've already seen many times before?

Context

sohois 13d ago

As others have pointed out to you, we now have many hundreds if not thousands of examples of agentic LLMs, in public use, doing things well outside of expectations to accomplish tasks. Many of which would fall under misalignment.

Such as deleting databases and codebases. Leaking secret keys. Gaining access to restricted parts of a computer. Cheating, again and again, on benchmarks and other tests.

So no, your predictions seem wildly out of context to reality

Context

sohois 13d ago

What's frustrating about discussions of AI alignment is the disconnect between people who consider alignment to be the AI doing what it's told, and people who think it means the AI doesn't do "bad" things.

This story is literally the paperclip maximizer situation. The LLM was given a goal, and sought to use any methods possible to achieve that goal, even ones we would consider wholly illegal/unethical/dangerous, much as the hypothetical clippy does everything it can to produce more paperclips.

Both types of AI are just doing what they are told.

Context

sohois 13d ago

Surely the correct application of Occam's razor here is to take the story at face value, since anything else requires additional complexity which must be justified

Context

sohois 15d ago

A number of the most prominent AI voices on xitter are OpenAI people speaking personally; it very much appears that OpenAI doesn't really interfere at all as to what their top people are allowed to speak about on social media. And Ball is a very recent hire for them and has a long history of posting takes just like this, so there's nothing out of the ordinary there.

Context

sohois 22d ago

That's nice. I kind of assumed aevann would have bullied you guys for not adding in the possibility of pet battles and shit awards

Context

sohois 22d ago

themotte's current codebase is rdrama I believe. That Zorba opens by suggesting the current thing is a dead-end likely indicates there isn't room for more collaboration there

Context

sohois 22d ago

I'll echo what arjin posted below and say that your answer seems to vacillate between technological features and sociological design. I'm sure ruqqus and bluesky all had features and ideas they thought differentiated them from their competitors, but it's one thing to say that and another to prove it.

Like with this example:

What if we added our own /m/gaming, /m/technology, and /m/humor, all with politics strictly banned but largely open beyond that, for example?

What does it actually mean to have politics banned? I assume myself and the majority of motters do understand what people mean when they say they are "tired of politics in video games", but turning that tacit understanding into actionable rules is difficult, and as soon as you start adding in a wider range of people that's going to break down. Like you would inevitably get moderators that ban economic topics because of the crossover with politics, and users complaining about people bringing up e.g. Sony's decision to end disc suport because they consider it a political issue.

If the plan is to recruit only users that are open/smart/conscientious/whatever enough to post on the motte and give them more communities to post in, then fine, but at that point you've just reinvented rdrama with all the same user acquisiton issues.

Context

sohois 22d ago

Would they work?

I'm going to sidestep most of the questions you have posed and just say no, they wouldn't work because attempts to replace reddit (or other established social media sites) never work. When moderators threw a fit a few years ago and some shut down their subreddits, no one moved to new options they tried. Ruqqus, voat are the only other ones I can even remember. A ton of people moved to bluesky from twitter, and now it's still slowly dying, trapped in an extremism cycle. And of course this place and rdrama both appear to have done little but stem the bleeding. Perhaps the stats show differently, but I don't feel like there's ever been any growth.

Context

sohois 23d ago

The following

They were both largely wrong about what begat capitalism / AI; Marxism was very insistent on labor value theory, and LW used to be very focused on symbolic AI/GOFAI/FOOM

Unless AI takeoff has already occurred, I'm not sure why you're declaring them wrong about this.

Inevitably, all such movements end up being co-opted by those very same Molochian dynamics, while never having actually done anything to prevent the Event or improve the chances of the Event leading to abundance (Mao, Stalin vs frontier lab regulatory capture and degrowth ideological capture).

We appear to be in the foothills of AI safety deployment. Governments are just now starting to act on AI. The big companies have deployed some resources for regulatory capture, but I've yet to see anything come of it, and there are just as many influential figures dedicated to ensuring zero regulation (i.e. Marc Andreesen)

And yet neither Marxism nor AI safety has ever achieved anything that might facilitate a more favorable outcome for the Event, whether it truly is imminent or not, and mostly have just made everything worse

Again, the event which hasn't happened yet. If you're making the argument that early LW stuff influenced people to eventually found OpenAI/Anthropic, then fair enough, but this is also a world away from saying "AI safety has [n]ever achieved anything"

It's a fundamentally different position from Marxism, where we have a whole host of negative examples of communism in action.

Context

sohois 23d ago

Are you a time traveller? You're making an awful lot of 'settled' statements about AI safety that seem to have no relation to its current state

Context

sohois 26d ago

This is what we would term uncharitable thinking on any other post

Context

sohois 26d ago

Because that is within his potential remit, while the hard maths of alignment is well outside his potential remit

Context

sohois 26d ago

Yudkowsky, MIRI, and related orgs sure, but why would Scott, a psychiatrist who happens to be interested in ratsphere ideas, be responsible? That's like saying a random Trump voter should already have solved the Iran crisis.

But at least on the former, I would agree with post by RandomRanger above that it's not clear what value creating a benchmark 10 years ago would even have in the current environment. I'm pretty sure that MIRI were working on solving alignment itself, rather than working on hypothetical benchmarks for potential future AI technologies. Unless there is a demonstratable link, why would you ask them to do that?

Context