simianwords

What would that look like in an example? A SaaS company selling its services to other agents in another company that uses that SaaS?

I can’t imagine AI making sales by itself

It’s literally not true though. Big tech built AI which is probably the most important technology since internet

There’s this popular misconception that companies only chase stock valuation but it doesn’t look true.

Interesting article but I don’t understand this:

    When they can eventually work directly with one another rather than sitting inside human hierarchies, we will be dealing with a genuinely new kind of economic input.

What does AI working directly with one another mean? I don’t think it’s anything new - it’s the same AI working on a long horizon task. A big part of the article works on the assumption that this is qualitatively new way AI can work but I see it as a human operating an AI agent that does long horizon task that internally can spawn multiple subagents (but this part is abstracted from the user who only gives it a prompt).

Unless there are domain specific models, what am I missing?

Yes but my relation to work has changed significantly. I haven’t written any code myself.

You’d have probably sneered at the suggestion that this would be possible in 2024. You’d be wrong then and you’d be wrong again now.

What does this have to do with prompt injection?

It’s a good sneer but in a few years this would be normal and you’d forget that you were sneering at this. I saw this happening when people said they used AI for coding in 2024.

so is it normalized for people to sneer at folks who happen to be ahead of the curve now?

I was downvoted on the suggestion that trivial prompt injection attempts are mostly solved.

https://news.ycombinator.com/item?id=50162483

Not sure if trolling but nukes are not decentralized and are tightly regulated. They are also famously a guarded secret and many wars have been fought to keep it that way and not let our enemies get hold of it.

I don’t understand this position at all. How can decentralization help here?

Let’s take one doom scenario: drones operated by AI. We can only have this work if we trust AI to do the right thing and not kill us.

I would reasonably say that any company doing this should be regulated and must pass strict alignment checks and make sure the models won’t do something bad to us humans and take control.

You on the other hand are suggesting: no let’s deregulate. Let’s decentralize and any company should be allowed to sell whatever drones they want. And alignment is fake anyway, so don’t bother aligning it. Drones do bad stuff because they work using probability so there’s no point anyway. Decentralization is the answer?

Other than banning the use of AI or severely restricting it so that AI can’t do most things (like tool calls) what is your alternative?

One of the fundamental problems with this issue is that people think alignment as a concept is fake. I think this is naive. Alignment is a real science with empirical findings and there is a way to make AI do what you want it to do. I don’t buy nor should others buy this fatalistic phrasing

If you agree that AI is actually that powerful and has very legitimate concerns for safety, what are you even arguing about?

So there’s no lying by anyone here in fact it’s a good thing that OpenAI is making these things known to the public.

What do you really think is the way to solve the AI safety problem?

Ban AI?

Let the first big incident happen and then regulate?

I read your comment and it only suggests that alignment is that much harder because you can't selectively prune information. Your post actually convinced me to be even more fearful of AI

Motte and Bailey.

Motte (easy to defend claim) : asking for regulation might serve both purposes

Bailey (hard to defend claim) : companies are conspiring to make AI look way more powerful than it is and AI is actually not that powerful. They are just trying to hype up the stock valuation of their companies.

This is the reality but people would rather believe that all this is hype marketing to prop up the stock valuation.

Isn’t it 0.2?

Isn’t this prompt cached? Then it’s not that big of a deal

I used Grok Bot and found out that I could offset some stock value going down and get back ~$300. I gave it my bank statements, email access (which had my stocks monthly report as attachment) and a few other things.

I'd have honestly never been able to do it myself - its just too tedious.

Modernity has always been about abstraction and specialisation. If he's so humanist, he must also want to make his own food and his own car because after all, you are offloading your thinking to the underlying companies.

This is so frustrating to me. Ever heard of the planner agent reviewer agent and the tester agent framework?

I'm not sure if that's what you mean but I know many that are making their own flow like this and think that this brings in a higher level of performance..

Honest question, why shouldn't any agent SDK's API simply be a PROMPT.md with a runtime that runs the LLM with tools; why do you need excessive abstractions over that like langchain offers?

Is it to have boundaries of some determinism?

I’d bet on this

I keep seeing that MiMo is in the frontier in terms of performance and cost but I see very few talking about it and the only ones that do talk about it are doing it in context of AA

Wikipedia can gain a lot by doing a cleanup using AI instead. There’s a lot of incorrect information in Wikipedia that you can easily detect using AI with a prompt like “find something wrong in a few randomly selected Wikipedia articles”. Usually it does find something.

If it’s that easy for me to correct it, it should be automated.

The benchmarks are already saturated and the curves for Terra stick to the axes.

Old = good now? I don’t think it’s fruitful to romanticize farming. That’s the same problem with marxists with hammer and sickle

You mean patronizing? In that case yes it is patronizing

I'm surprised at people suggesting that the model layer can't be the moat. Having great models that are smart requires talent, culture and good judgement. Sure it can be commoditised to an extent but could it not be that Anthropic will always stay in the lead?

They only have to be marginally better and there will be tonnes of people who will pay a lot of money to have that marginal intelligence.

"knowledge economy arbitrage" is good actually. That's how almost all of tech industry works. And we are all fine with it.

Call it "colonialism" if you want but it screams "fake moralism"

Me neither, I was quoting someone else

Couldn't you just train a general purpose LLM on the output of this CNN?

Anyone wanna take a bet that the general purpose model will beat these specific models in a year or two?