worldsayshi

I think these SFRs are basically indoor blimps.

And yet all this power in the hands of the very few are only there because the rest of us don't know any better than coordinating for their benefit rather than coordinating in any other way.

places like India

I had the impression that India is quickly turning rich enough to say no to things like this.

I've used Interrail every time I want to travel a longer distance across Europe by train. Spares me some of the stress of dealing with interruptions since I can often just hop on the next train. Unless it's fully booked...

Or they scan at the edge on the user's device.

So then the user can "just" install their own client.

Is scanning mandatory? - No — voluntary.

Voluntary for whom? The service provider? Can I opt out of getting scanned?

Does it touch encrypted messages? - No. End-to-end encrypted communications were never scanned but providers could deploy client-side scanning under this law.

So it circumvents e2e encryption?

---

How would these laws prevent me from just side loading my own open source client?

Why Oracle?

Also, all connections in iroh are end-to-end encrypted.

Couldn't this be a flaw in the attention mechanism? Like they need some kind of grounding. An awareness of what they fundamentally should care about and how the thing they are currently giving attention to relates to that?

It seems like the site is currently getting hugged so I can't look but I'm glad someone is trying to achieve a more minimalistic take on the UX of a software forge. It seems like an area with good room for improvement.

Such competition will only turn into a moat of the hardware suppliers never manage (or choose) to fully adapt to increased demand?

But then the real moat should be on the hardware side anyway?

Realistically they can't go that much above the actual cost for inference since the customer can always switch to self hosted or inference only providers. Or their models have to be significantly better than the open source models for the foreseeable future. They will never be able to charge much more for their lower tier models.

Don't you think that we will eventually get more specialized hardware that will greatly improve efficiency? Running neural networks on GPU:s seem like quite wasteful?

I see at least two potential positives in this:

- The frontier AI companies have realized they won't be able to count on gaining ground and earning more in the future through sheer moat. They have to start earning right now.

- The playing field on the market got a whole lot more even as a result. Now everyone is competing on cost and quality - while there are still a lot of competition. AI suppliers can't easily get away with subsidizing their own product and enshittify later.

I might be missing something obvious here? It feels to me that if the frontier AI companies thought they could gain a lot more moat they wouldn't raise their prices this much this early? And their current moats/head start doesn't seem insurmountable?

craftspeople take their jigs with them from job to job

Except for software gigs the software typically belongs to the customer so you'd need to rewrite it every time...

I guess it becomes different if instead of hiring more people to do more - all investment goes into more AI credits.

Then again, as long as there is more demand and there's a limited supply of compute you can still continue to hire people as well. If we assume that the market has infinite demand for whatever AI + humans can produce together both will have jobs.

If demand is limited and compute is plentiful it should make sense for a company to try to have AI do as much of the work as possible.

I guess AGI is the breaking point and superintelligence is everything above?

no one can stop me or control what is said no matter what

Can't they just triangulate the nodes and hack or unplug them? And put whoever objects into prison?

You're probably on point there.

The code it generated was awful.

I suppose you could solve that in two ways. Manually rewrite it as you did. Or formalize an architecture and let the AI rewrite it with that in mind. I suspect that either works.

I see two points:

1. AIs aren't yet good at architecture.

2. AIs aren't yet good at imagining technically exciting stuff to build.

And I agree that there's still space there to build a career in the short to medium term (plus Jevons Paradox). When both those points are no longer true we are certainly much closer to, dear I say it, agi. I suspect that (1) will be solved for somewhat limited domains in the near future using harnesses. And it could snowball from there.

And more intelligence should give an opportunity to increase explain-ability rather than just complexity. It can potentially explain the proof at the level of the listener. Make visualizations. Etc.

Why can't we (or AI) invent ways to explain information that makes it much more digestible? And the solutions simpler?

Why is it necessary to continue to increase complexity when we get better intelligence? Can't we find more simple solutions? Or at least more explainable.

The more expensive it gets, the higher the incentive for more competition in the hardware space.

It should be feasible to crowd fund training runs right?

I often think of go as a "better" python. As in, easy to learn and easy to use. But also performant and the module system and package manager seem to be a little neater. (sorry for flamebait)

But I wonder how well it can cover similar use cases? Go is great for devops and web backends. But what about AI and data science?

I absolutely love this vision. He's still working towards the goal. It seems that his vision has problems scaling up though. He seems to mostly still drive this himself.

What kind of health devices require reliable internet connection to prevent death?

So pi.dev + forgejo?

Maybe forgejo has a shot?