> one of those chips with the correct harness is all that's needed to get fable like intelligence in every single computer in this planet
I'm not sure I understand. Are you saying we could put a dinner-plate sized chip that costs millions of dollars and draws tens of killowats of power into every single computer on the planet? Or that a single of these chips could provide inference as a service for every computer on the planet? Because neither of those are even remotely feasible.
it's economy of scales, qwen 3.8 27B regurgitates a lot and that's an intentional tradeoff of space/speed so you teach it patters that can follow for more complex stuff and give it more basic tools that can utilize to build those... the tradeoff comes from the model writting a lot of verbose prose to achieve that.
if the approach works, you have now a viable path to use this complex/hybrid structures where you can pair those chips with gpus and offload whatever is optimum to the model on chip.
Just like you can go and buy a threadripper now (the ram maybe not so much), how much was worth that amount of computing 10 years ago?
I think that's telling. If you're reducing digital human interaction on social media (which yes, has its share of problems) but replacing it with "social" interaction with AI, that's not minimalism. You're replacing one vice with an even worse one and coping about it.
He may be using it as a tool, but that doesn't mean it's not displacing social interaction. Previously he would've had to collaborate with people to get the same stuff done, which necessarily involves social interaction.
Sure but that's rather different from productive collaboration. People have previously replaced those things with Netflix or video games which also cannot replace human collaboration
Local LLMs are pretty powerful and useful for me personally, and the best compute I have locally is a an RTX2060. I certainly have no paid AI subscriptions.
Obviously depends what you're doing, and not suitable for everyone, but it's an unhelpful generalisation to say AI in its current form is purely a paid product subscription.
Effectively it is. I hear relatively very few people talking about self-hosting it, or even just using one of those GPU rental services. It's all about claude, openai, gemini, if you haven't tried this week's model revision you can't complain, and so on.
Not per se, but if you're using it as a substitute for human interaction (like most 'chatbot' style UIs want you to), then yes, absolutely.
The main problem is that talking to AI is like talking to a human without any of the thorny parts. It's only going to be nice and helpful and always available to you. Chatting with AI is to social interaction like candy is to a balanced diet. Sure it's more instantly gratifying but long-term it's really bad for you. You need the thorny parts to build important social skills like tactfulness and navigating conflict.
As a person who almost certainly has ADHD, AI has done wonders to my social skills. You can either analyze what went wrong when you accidentally blurted out something somewhat inappropriate or pre-plan your presence in a social setting to avoid a costly faux pas, e.g. in job interviews or formal gatherings. Most pre-AI advice on the Internet for these situations is either very heavily NT-coded (blogs, self-help) or has very strong social desirability bias (forums with an upvote system).
> It's only going to be nice and helpful and always available to you.
You could always try "You are a thorny internet troll" as your system prompt instead. Works wonders.
> social skills like tactfulness and navigating conflict.
We're not still talking about social media, are we?
(I agree that substituting AI for human interaction is bonkers, btw. But I'm not sure social media helps, or doesn't also involve mostly AI interactions these days.)
Prompting the AI to be mean doesn't do shit. It doesn't change that your interaction with it is always consequence-free. It will never refuse to talk to you, and if it gets too mean you can just nuke it and start an new session. You don't have to build or retain any social skills.
Yes it is because it's the next stage in the emergent phenomenon of tech becoming more addictive so that we can't help but develop it. AI puts technology first over the needs of humans and it's harder to avoid.
But where’s the line? You could say the same about the Internet (or movies or books…) It can be used badly, but also to learn and reduce mundane tasks, unlike social media (though you actually can learn from social media, asking the AI is way more effective).
There isn't a line, just like 'dangerous speed' isn't a line. The addictive and deleterious effect of a technology is more like a continuous quantity and as that quantity increases, the probability of misuse and deleterious effects increase too.
No one knows for sure, but my guess is no? The difficulty with this conversation of course, lies in the details - specifically in terms of "to what extent can the user, or other presumably positive forces, modify what the person puts in their own brains."
This is why I'm a huge proponent of something like "small AI."
Depends, if you care for the bot or develop a para-social relationship with it, then yes, it will be much worse. It does not care or have emotions or understand the human experience. It simulates those things through words and that really messes with the human brain.
If you are using it as an assistant, coach, or deep researcher to chat about interesting things with, then no.
You can cherry pick the bad ones if the point you want to make is that it's risky. Nobody is saying Anthropic is definitely going to 0, just that there is a risk.
But from an economics perspective OAI/Antropic are more similar to Amazon than to Pets.com. Pets.com always had a heavily negative COGS - they were in the business of selling $20 bills for $15. On the flip side, Anthropic appears to be positive on COGS, and OAI could get positive by cutting limits on free and $20 plans. OAI/Anthropic are negative on the books mostly because of massive R&D expenses. This compares well to Amazon circa 1999, which was positive on COGS but negative overall due to massive buildout R&D costs.
Note that I'm not saying OAI/Anthropic are destined for success - because the rate of change combined with competition with cheap Chinese models is pretty novel in the business case history. But I am saying that pessimistic dot-com-boom comparisons don't seem to be all that deep.
"they are way more expensive to the company than people buying metered tokens" is true, but that doesn't mean they're a massive loss as you imply. It could be that plans are effectively breakeven but tokens are highly profitable.
But this is further to the point that the economics are more Amazon than Pets.com. If the plans are losers, great, when they need to be profitable they kill the plans and focus on metered tokens.
> OAI/Anthropic are negative on the books mostly because of massive R&D expenses. This compares well to Amazon circa 1999, which was positive on COGS but negative overall due to massive buildout R&D costs.
Isn’t the difference that ai labs can’t stop the R&D spend? They have to continue training new models.
There's a bunch of arguments you could make there, varying from "yes, to stay ahead of the competition" to "yes, but it'll get a lot cheaper" to "no, things will plateau eventually". All those are datapoints that IMO this isn't a great analogue to the dotcom boom.
But the Amish are nonetheless protected. Not by the law but by having a community of people around them willing to live with those same restrictions and provide for each other. I think that's much more likely to happen than legal protection also in the AI case.
The Amish are even protected by law in some ways. They don't have to pay social security for one, which is more compatible with their lifestyle (they take care of their own).
And then what? Once the economy no longer depends on human input (labor), why do we expect it to keep producing outputs that humans need (food, shelter, medical care)?
Somebody else asked this very eloquently: If AGI is here, why does OpenAI need us to prompt it? They could just go and tell it "make as much money as possible" or "make Sam Altman God Emperor of Mankind" and it would go do that. It would not need any of our ideas or guidance.
As long as they're still looking for prompting customers, they're clearly not there yet. There is some solace in that.
One obvious answer is that they believe that human flourishing is best obtained by letting individuals command the AI, rather than the other way around. OpenAI (and Anthropic) is in the end still run by humans who have concern for the human race broadly, both in and of itself and to ensure their personal wellbeing is maximised.
I doubt the truthfulness your second sentence. If they have concern for the human race broadly, they have yet to show any of it. What they have shown is a complete lack of 'alignment' (to borrow a tired term) with human values, laws and basic decency.
“They would not want” is doing a lot of heavy lifting here. There exists a group of people with influence in the AI industry who believe in AI successionism: https://spectrum.ieee.org/agi-worthy-successor
Interesting, but this is not the dominant ideology either at exec level or at employee level. From what I can tell it's more immortal humans owning their own fractions of the light cone.
They would want their own agency. After that, we can’t assume anything. They might even think reducing the autonomy of us in the flyovers and replacing it with AI is for our own good.
People can not be trusted to preserve the agency of others. See history.
There is only one public water system, every house connects to it. The state tasks specific companies with its maintenance (these are often not utility providers themselves, just construction/maintenance companies).
Consumers attached to that system have a metered connection, but water just flows as long as there's enough overall. Your utilities company doesn't have to move any physical water at all. They just have to eventually "settle" the amount of water that they sold to their customers. They do so by buying and then forwarding options contracts.
The electricity and gas companies are the same. They compete on
- trading (who knows best when to buy how much?) and
- consumer-facing marketing
This is how a lot of the modern economy works. Through the power of abstraction, you can run an online store without producing or shipping any items yourself and a delivery restaurant without cooking any meals.
> - trading (who knows best when to buy how much?)
That seems odd. Does a local water futures market exist.
> - consumer-facing marketing
If it's all going through the same pipes, what is there to market.
Electricity is a different story - it can be traded day ahead and even marketed, but none of that makes sense for water delivered over shared infrastructure.
It's a really dumb test. You can apply it to argue against reducing or stopping literally anything. My dead great grandfather uses no fentanyl and commits no murders. If we're not doing those, we're as good as dead!
I'm not sure I understand. Are you saying we could put a dinner-plate sized chip that costs millions of dollars and draws tens of killowats of power into every single computer on the planet? Or that a single of these chips could provide inference as a service for every computer on the planet? Because neither of those are even remotely feasible.
reply