jchw

(2023)

You have to realize by now that outside of academia nobody is even paying any attention to people who are predicting existential risk that far out. For worse or even worse, AI 2027 is the AI safety conversation we're having today, and I regard AI 2027 as essentially Anthropic fanfiction, except usually fanfiction is written better.

In spite of this, I actually kind of did build a "software factory". The main reason why was because there is a number of things I build that really are just means to an end, tools where I could not give a shit less if the code sucks or anything as long as it does roughly what I want. This stuff will most likely never be posted anywhere, but that's not the point, because it still has quite a bit of utility. Having a "software factory" to build and maintain it seemed a lot less annoying than being a "meat proxy" all day just to maintain some tools that aren't very important.

I could be misunderstanding what a "software factory" is but I think whatever I have done fits the criteria. I have a custom agent that goes around implementing software and maintaining private repositories completely autonomously and I can communicate with it using issues and comments.

It seems like a useful pattern, just maybe not what people had in mind. I think people figured that you'd build an AI Microsoft or something. I don't think that's true because most of what a company like that does isn't really software engineering anyway, but there is probably still something to this pattern.

They asked:

> Can you provide examples outside of gross negligence on part of the dog owner?

And the answer is no, I can't. The reason why is because there is essentially no such case. It is basically negligence by definition when your pet hurts someone who didn't cause it directly themselves because it is your responsibility.

So I asked:

> Can you provide an example of where your dog harming someone else is in fact not negligence on your part?

Which is a really easy question to answer and it seems you could not.

They certainly didn't invent this strategy, but that's because it already existed. Regulatory capture is nothing new.

> 1. They, and many other experts, have been saying the same thing since well before the current paradigm existed.

Some experts sure. "Some experts" say lots of things.

However, this is actually much to their advantage. If you were a frontier AI company, a legitimate concern to use as justification for your regulatory capture is exactly what you want. That is, after all, generally how regulatory capture gains public approval.

AI safety was also why OpenAI had to charge per-token for GPT-3. It must've been such a struggle for them to make that difficult decision. I'm sure Sam Altman spent a lot of nights crying to sleep over that one.

It's good that we never got open weights for a GPT-3 class model, right?

> 2. There is nothing about regulation that is required to give them an advantage; it’s a fallacy to suggest “regulation = advantage”

It's difficult to outline the number of different ways that regulatory capture can and will benefit OpenAI and Anthropic.

If we limit things to "Fable"-class models, then nobody competing with Anthropic/OpenAI is ever allowed to outdo them. If you literally legally aren't allowed to make better models, you can never charge the "premium" of having the current top-tier model. You have to compete by being cheaper than Anthropic instead, which is a lot less enticing from a standpoint of trying to become profitable. Take a look at the per-token pricing of Fable real quick. Yeah. Being able to charge a premium for the model that does the cybersecurity thing is literally their most important asset.

If Anthropic really is making a profit now as rumored, then they need regulatory capture as fast as they can possibly get it. Once Anthropic and OpenAI prove that AI is a viable investment, entities with a lot of capital will suddenly become even more interested in re-allocating it towards taking a slice of this pie.

> 3. It would be preferable to argue the facts (is it or is it not dangerous? Why or why not?) than conspire as to why they’re stating what they and plenty of others have been saying for decades

OK. On the factual end of things, AI safety experts have been raising the alarms for over a decade and so far essentially none of their predictions have come true.

The rationalists, the main group pushing the AI safety rhetoric that dominates the Western AI sphere and is virtually absent in China, which appear to be going entirely off the deep end at this point in time, put together their best minds to produce AI 2027, which of course like most predictions started to fall apart a bit before hitting the one year mark. Predicting the future is hard.

I have made this point before but I will make it again: I think AI safety experts deserve at least as much skepticism as Ed Zitron in their claims. They've been making similar dubious claims about AI capabilities and an incoming singularity for a long time now, and so far the best we can say is that maybe it will eventually come true.

Maybe. Or maybe you've read too much goddamn science fiction.

Go also just produces suboptimal code because it isn't maximally geared towards runtime performance. It lacks a lot of the more expensive optimization passes, like auto-vectorization.

On the other hand, time and time again it has been proven that it can be a good language to write tools that are perfectly sufficiently fast, like esbuild. With ts-go the choice to target Go instead of Rust seems to have come down to sensibilities and ease of translation, and possibly some other factors as well. I don't think they were aiming for maximum performance, and it would be unsurprising if you could squeeze quite a bit more out of ts-go itself with some effort.

The real benefit was getting the initial "10x" improvement, of course. Getting an additional 2x-3x is fantastic, but it's less noticeable when your codebase is already getting typechecked faster than you care to measure. For me even my largest codebases are chewed through plenty fast on ts-go, and I trust the TypeScript team to have done a good job (especially having seen some of their process for doing the port) so I think I'm good on missing out vibecoded Rust ports and whatnot. Interesting for sure, but we'll see how it all pans out in the long run.

Can you provide an example of where your dog harming someone else is in fact not negligence on your part?

But negligence seems like exactly what OpenAI and/or their contractors are guilty of anyhow.

It's a program on their computer that they ran on purpose... And yes you are definitely criminally liable for things your dog does.

The agents are clearly not accidentally compromising systems. Seems very easy to trace back the responsible parties here, I really don't put much stock into this argument.

I think it may just be bad automated flagging. I vouch'd them since they don't really seem to violate any rules/guidelines (right?)

I don't use OpenCode but I would like to know who this impacts. Surely people who are just using OpenCode locally with their own credentials wouldn't be impacted, right? This sounds like something where you have a hosted service and you put your keys into that. Not saying this isn't bad, but it isn't clear and they're not sharing any information.

There is no obvious explanation for why you would want this anywhere I can find... Does anyone understand? If you wanted this, wouldn't it make more sense to extend the last version of TypeScript before ts-go instead? Hmm.

Well no, you probably wouldn't get thrown in jail for that for the same reason that having your car stolen and used to commit a murder wouldn't usually get you thrown in jail. I dunno what you mean by suggesting crimes committed using computers are ambiguous. I mean the CFAA is vague and all, but more in an overly broad sense, which in this case seems to make it even less ambiguous.

Sure, sure, but they've already won by even having a policy against AI slop articles and posts, which, to my knowledge, HN seems steadfast in refusing to do so. After all, the public is fiercely divided into mostly one side in favor of banning said articles. HN is apparently just going to ride this out with only a policy banning LLM comments.

At this point, my only reason for still being around on here is because I simply don't know of any place that I find strictly better. Most of the popular alternatives to HN are annoying in ways that I find equally hard or even harder to tolerate.