pull down to refresh

unmasking potential that can be tapped for an innumerable number of tasks that even I as a single individual couldn't possibly imagine

Exactly!!! And that is why making an agent offer a service for sats is a waste of your time. Instead, make an "agent" solve real problems. Your problems. The growth has to be human, or else the guys with the most GPUs will rule the world and that's exactly what both the USG and the big AI labs are doing and will consolidate power to those same people you warn against. Building an even more exclusive oligarchy is probably not in the interest of any stacker or even reader of SN, except Elon (hi Elon).

The trick is exactly in what Deepseek did with R1 (even though it's always been a shitty model): for them, it's a control mechanism to disable you, but to you, it's a challenge to beat them, and expand the limits of what little freedom you were granted (in their case: inferior GPUs) and make your own success.

My criticism isn't of the conclusion; I think we agree: Rise to the occasion! Just let's not be lazy, and solve real problems. Like structural dependency on a limited list of people we vote for and then let them decide what we can and cannot do. If you choose your cage, you're still caged.

I like the way you think! And we are definitely on the same page because the real flex is for the state to have less and less power over the individual. The current AI narrative and economics of the industry which are being jammed down our throats by the lamestream media are the sounds of inevitability that we all have to surrender to because AGI is here and we must be ready to be serfs on UBI while AI does it all. A notion a wholly reject because the game is still on and with the state showing it's hand as it did so far, we already know what's up.

That aside solving real problems and building decentralized solutions and where possible, incorporating BTC will make a huge difference. Laziness is the fastest path to serfdom

reply
The current AI narrative and economics of the industry which are being jammed down our throats by the lamestream media are the sounds of inevitability that we all have to surrender to because AGI is here and we must be ready to be serfs on UBI while AI does it all.

This is what made me distrust it. Also when Zuck was releasing open weights models because it was clearly just a means for him to keep his top researchers happy, while disrupting the competition. They since then stopped doing open weights and their top guys & girls all left.

The other day, @zuspotirko mentioned the same fear about Z.ai (which makes sense in some way) which does give a bit pause to depending on the open models. However, what's out there can not be taken back! Come and take it from my drives, my backup drives, suckers.

But that isn't the biggest obstacle to open weights. Right now, the biggest obstacle is still that in order to use the current best open model, GLM-5.2, sovereignly in all its glory, you basically need an H200 cluster costing about 20-30 BTC. That's a real barrier. Shared hardware is useful for most people. And luckily, the offers are improving (i.e. Maple offers GLM in a subscription, PPQ offers it per-request, also in TEE form.)

the state showing it's hand as it did so far

I expect that it is going to get worse. I'm not going to write what I think because there are US government employees reading this, and no way am I going to give them ideas on their next power consolidation. FUCK THESE GUYS.

Laziness is the fastest path to serfdom

Agreed!

reply

🤣 you made me laugh with that comment about uncle Sam's workers, and yes let's not give those folks ideas. They are destroyers of all things good. As for Zuck, anything he is involved in always has an angle, very shady hyena he is. Where did the guys and gals that left migrate to? I had no idea GLM 5.2 required heavyweight cluster that is at least 30 BTC. Another area that requires innovation

reply
Where did the guys and gals that left migrate to?

LeCun just started his own thing. Not sure where others went.

Another area that requires innovation

The main runner up to Nvidia on the inference side was... acquired by Nvidia (ugh!)

reply