Ostfriesentee

Maybe just another two years, but the trajectory is promising and hardware will improve, as well. Strata is good for inference, at least.

Those 120 years went by in a flash.

I don’t quite follow. In the current scenario he has committed both treason AND was unfaithful, no? How is that a smart move? That’s pretty stupid, if planned with intent.

GrapheneOS reigns surpreme. We should all use it. But I am not buying a phone for €2000. Count me out.

I’m not sure making the tool you’re selling less useful is a smart idea. Hmm, yes liability is an issue. But other models are just gonna allow it anyway, so people will just stop using Claude for that usecase. Reverse engineering won’t stop. Claude will just not be used for it anymore

I think yawning, mold and the movement of sand dunes are pretty interesting.

“Anthropic is scared that local MoE models are going destroy their subscription-based LLM business model so they try to scare governments into regulating local inference”

I pay zero, not even electricity because it’s being generated in this building, for my agent. It’s running entirely on my computer and it’s decent enough. The LLM subscription companies are eventually going to fail because everyone and their dog is going to be able to run their own models.

That’s okay, I guess. Disclosure is important.

Are you still bright, BrightCandle?

Same, a 35B model is running on my gaming rig. It’s good enough for most things, even coding.

Also note Apple’s, AMD’s and NVIDIA’s (Spark) unified memory solutions for local LLMs/agents. AMD advertises 300B and NVIDIA advertises 200B models running locally. The question is whether Anthropic/OpenAI are going to be relevant at all in the future, with local models becoming increasingly capable and affordable. However: There are rumors the Chinese government wants to restrict access to its models. Not set in stone and also not sure whether this is true. Could this entail no more access to new open-source models? Who knows. But even then, energy and resources are going to be a bottleneck. I have not looked into the concrete numbers and this is not my field of expertise, but I find it highly unlikely that they are going to be able to capture the entire market.

Teeth regrow trials are coming up. Probably gonna hit the market in the 2030s.

Isn’t that what Mistral is kinda pivoting towards? I think they now also host third-party models. That’s essentially providing cloud computing for open-source models. Not entirely sure.

I still do that and never did it any differently. Mainly to conserve energy, since my parental household was poor and we actually could not afford to leave appliances running when not in use. This became a habit and now I’ve been incidentally preventing data harvesting for years. I do need to get an old 90s alarm clock though, since I cannot always turn off my phone at night when I have plans in the morning. Also GrapheneOS.

Nor sure if it’s FOSS but HERE WeGo is pretty okay.

With continued optimization everyone is eventually gonna host their own LLM locally. Then OpenAI will be irrelevant. You can already host some flagship level LLMs with tools on your humble gaming rig with the right settings.

VRChat still has a little old internet. Some worlds are weird, but some communities engage in stimulating discourse. Of course, avatars and such are monetized to some extent, but not the spoken content (please don’t monetize speech).

I mean, you can chill in VRChat if you want.

deleted by creator