ilaksh

Obviously there will be a person nominally administering it and officially running it. But some of them will be really minimizing the human involvement.

Funny, but 2027 will actually be the year of the AI Companies run by AI CEOs.

Obviously kind of early and hard to pull off still, but by the end of next year I think you will see a ton of people attempting it with at least partial success.

There are already plenty of less serious attempts.

What do you mean "influenced inference" almost instantaneously? My laptop from 6 years ago can run an agent in a very fast loop in a web browser. It's not going to take over anything though since it's Gemma 4 E2B with only 2 billion parameters.

AI that we have now is not a digital animal in capability or kind, and is not currently anywhere close to taking over.

But it's still in its present form very intelligent in meaningful and useful ways. And it is not too soon to talk about concerns of a potential existential threat in the future. Because it could sooner than we might realize, threaten our existence.

Because of the potential, we should have a culture of caution as we continue to rapidly improve AI.

Yes but it seems more feasible to try to avoid people starving in there in the medium term. Abolishing them is a good goal to pursue in addition. It's just more difficult.

My question is, are they actually deporting people efficiently? Because if they are accumulating people in these facilities, it could become dangerous. They have been dehumanizing them for months.

If there are constraints for food supplies, staff are not going to risk their jobs to keep prisoners fed.

There needs to be a law for video audits every X weeks. Or there is a real risk of starvation in some places. Some of them are for profit and the intense racial hatred is no joke.

Give credit to Thinking Machines for their recent open release. Also Google's Gemma 4 is pretty decent. Also thanks to Ideogram for open weight v4.

Stack Overflow abuse and the rage it generated in me is sort of nostalgic. Kind of sad in a way that I can't quite recall the rage anymore.

By the way, at least as far as what it is showing me，SO is an AI site now. Home page Ask a question goes straight to an LLM doing RAG against Stack Overflow.

Edit: actually they tricked me into asking AI. The text box to ask AI is right next to what is actually an Ask A Question button that skips the AI.

It's training data seems limited or out of date because it had no idea what Gemma 4 E2B was. It just said that wasn't in the SO data and could I explain wtf I was talking about please. I guess not Gemma 3 either. And no web search. So it's useless for me today.

Interestingly, the Stack Overflow AI answer is provided by OpenAI, and these may be the most useless responses I have seen any time recently.

Whaddyaknow. Got me a little twinge of the good ole' rage tingles after all!

Fair point as far as there being low success rate in some ways and over-enthusiasm. But 0% success is a dishonest exaggeration.

He's ramped his own AI spite up to a manic level.

I don't suppose this works in the browser?

It's as good as gpt 5.6 sol and _half_ the cost..

This showed up in my Youtube recommendations, I am not associated with them.

yeah, they are barely hanging on. they only raised $144 billion over 14 rounds. who knows if they will ever get any more. we should all chip in for a t-shirt.

:P

they have, click on "back to store" -- there is more merch.

Are you sure there isn't something qualitatively different from general purpose AI and robotics as opposed to the type of automation that Marx knew?

He's basically saying that even though AI capability is high and rapidly increasing, it is not reliable, creative or tasteful enough to replace humans. Further he implies that it will take decades before this is the case.

But we already do have have some kind of measurement of most of these types of side factors, and they actually aren't at zero and are increasing rapidly. So the implication that they will not be human level until decades from now is just (hopeful?) speculation or fuzzy thinking.

To me this looks like a really academic and official sounding version of the same quasi-religious hopium that usually defends the sanctity of the human. He is essentially saying that there is just something so special about humans that it will never be reproduced in a machine. It's very similar to dualism (and in many people actually is religious dualism). No AI is going to have human creativity or judgement. Not anytime soon. Why? Well, we all just _know_ that's not possible. Okay, maybe in a couple of decades (but they don't necessarily believe that anyway). Why would that take decades? Well we all can just _tell_ it's no where close, right? Because AI of today just isn't special like humans.

Aside from that worldview issue, I think that people still are not taking seriously or internalizing the concept of exponential improvement.

Computing efficiency gains can actually level off. In fact, they have many, many times before. But they always tilt back up again when we invent the next approach to get beyond the current level. This is how it has been for 90 years.

There are multiple ways that we continue to see huge gains in AI software, architecture, and hardware. There are huge efficiency gains available still as we move towards more radical fully compute in memory and/or analog approaches and other options like models implemented in hardware.

But Telegram hasn't engaged in that, some of their users have.

I think the issue might be that although Telegram has a lot of abuse takedown activity, they do not permit access or direct action by authorities. If I recall, they have reiterated many times that some level or types of messages always remain private.

Maybe that's the issue is that a lot of illicit activity is going on in private channels and whether or not their filtering addresses it at all, authorities see the activity and have no access for court cases or direct action against it, so they can imagine it is quite rampant.

hm. I should have written that more precisely, but I thought it was implied/obvious that it would go down to some degree or another. I didn't mean it would literally remain flat. it's a question of how much they drop though, and I don't think they know for sure, because there are new technologies that are really just being held back by manufacturing scale and anti-competition which otherwise could cause larger than anticipated pricing drops for older hardware. like.. how could you read what I wrote in that comment and conclude that I needed you to explain that the prices would drop?

Dumb question, but when the Nebius capacity dashboard says they have around 3 non-preemptible B200s available, does that mean _total_, or is it just how many I myself might be able to rent on demand?

One aspect of the profitability might be the utilization and the pricing a few years down the line for slightly older hardware. Already now it seems like the increased processing you get from newer devices versus the cost difference makes something like an H100 or even A100 significantly less desirable than newer more powerful ones. As an individual, I am happy to be able to get an H200 on demand, but the B200 or B300 can do so much more work with optimized software and models for only modestly more cost that if those become available then from a business perspective you really have to prefer that if you can keep it occupied.

Then with Vera Rubin being like 3 times more effective or whatever, that adds a new layer of gradual obsolescence. So the question is can they keep the pricing up on the older ones a few years down the line enough to fill out the end of those expected payback periods.

The real boogeyman for a neocloud that has heavily invested in expensive Nvidia hardware might be a variation of that beyond Nvidia with startups that have even more dramatic efficiency increases pushing the leading edge even further. For example, if companies like Mythic AI and d-Matrix could somehow rapidly rapidly scale, that would push prices down for all of Nvidia hardware that is significantly less efficient.

I guess so far it doesn't look like any startups with really big efficiency breakthroughs are even close to being able to scale like Nvidia though, especially with the manufacturing and power crunch. But I suspect some of that is because of favoritism and strong arming protecting investments rather than a free and fair ecosystem.

Luckily there is nothing about the current period that reminds anyone of Hitler's time.

AI discussions these days remind me of college when I agreed to "debate" some Christians. You can assign the camps however you want to assume. But the point is that there is not going to be a lot of constructive discussion.

What are the possibilities for adding more high level tasks like "pick up the [arbitrary thing]"? I assume it's 100 times harder to deal with hands and arms in a generic way. But maybe for grippers with two claws it could be more tractable to just output two force vectors per claw or something for the grasp and another two fir the drop. And maybe the SDJ could do reverse kinematics or something.

But one RGB image wouldn't work. So maybe one would need a depth camera.

Zulip has been open source since 2015.

From what I can see, the license for this (AGPL) is more restrictive than Zulip (Apache 2). So I would stick with Zulip.

if you don't find that then you could fake it with personaplex possibly but making another ASR/STT model just listen continuously and transcribe then send to an LLM with function calling. at least that would allow one direction easily.

Are there any open source full duplex models that are out besides PersonaPlex? There was a chinese open one, maybe Fun Audio chat or something, that said it was going to release a full duplex version but I am not sure if it did.

My dream would be open source full duplex with function calling or some kind of rudimentary text output. PersonaPlex is still interesting although it was looking like we would need to fine tune it to handle outgoing or avoid going off the rails easily.

Well, Taalas has that kind of technology, but the chip they demoed is probably 20-100 times smaller than necessary since it's only an 8b model.

But let's say they could someday scale that up to a much larger model, 72 large chips per wafer and each chip can do 1000 LLM requests at once (Vera Rubin?). So it's roughly the equivalent of an NVL72 rack.

You might be able to serve something like 50000-60000 requests at once. So I think it's more like handling a small city's worth of customers per wafer than the world if you had that.

I believe in less than 5 years we will get to that, but the model size and/or number of agents is going to keep going up also.

I think the profits depend on how well they manage their fleet purchases (or possible sub-leasing?) to get high utilization without overloading or idle racks.

Because accelerators like H200, B300 etc. are highly parallel and designed to run like 200 or maybe 300 sequences at once (depends on the model, just guessing). I assume they finance the hardware and that cost per device or rack is the same whether each unit is handling 10 requests or 150 requests (aside from electricity).

And probably international customers factor into it to get good utilization over more of the night time. And it likely is something that they look at quarterly more seriously than monthly. The biggest risk to profits might be a downturn in business that causes some portion of the financed AI accelerators to go idle or get low utilization for some weeks (that they can't sublease).

Shocking that a well executed AI tutor improves outcomes.

Hasn't computer assisted interactive learning already been proven for years? Why does there seem to be so much skepticism about enhancing it with AI?

Is this just something like, astoundingly slow adoption or poor execution? Being held back by paper textbook makers? Teachers unions dragging their feet?

How can interactive AI driven individually paced learning _not_ be obviously dramatically more effective?

They were using Sonnet 4.6 for some fre form responses so that could be applied to something subjective.