News for hackers. Stuff that matters.
Re:Show HN: I've built a locally running Perplexity clone by wg0 on Thursday April 04, 2024 @05:38AM (#39928291)
In five year's time - by 2030, I foresee that lots of inference would be happening on local machines with models being downloaded on demand. Think docker registry of AI models which is pretty much Hugging Face already there.

This all would be due to optimisations within model inference code and techniques, hardware and packaging of software like the above.

Don't see billion dollar valuations for lots of AI startups out there to materialise into anything.

Re:Show HN: I've built a locally running Perplexity clone by openquery on Thursday April 04, 2024 @05:46AM (#39928327)
> I foresee that lots of inference would be happening on local machines with models being downloaded on demand

Why? It's much more efficient to have centralized special purpose hardware to run enormous models and then ship the comparatively small result over the internet.

By analogy, you don't have a search engine running on your phone right?

Re:Show HN: I've built a locally running Perplexity clone by Sammi on Thursday April 04, 2024 @05:59AM (#39928400)
You currently can't have a search engine running locally on your phone. Google search is possible the single largest c++ program every built. And nevermind the storage needs...

But in a few years we might be able to have LLMs running on our phones that work just as well if not better. Of couse as you mention the LLMs running on large servers might still be much more powerfull, but the local ones might be powerfull enough.

Re:Show HN: I've built a locally running Perplexity clone by vachina on Thursday April 04, 2024 @05:51AM (#39928358)
A more appropriate analogy would be driving your own car vs. taking the bus.
Re:Show HN: I've built a locally running Perplexity clone by bufferoverflow on Thursday April 04, 2024 @11:24PM (#39938401)
No, a more appropriate analogy would be driving your own billion-dollar super-yacht vs driving your own car.

Will not happen any time soon. Consumer hardware can't even run GPT-4 locally, and won't be able for a looong time. Each GPT-4 instance runs on 8 A100. The cost of such system is ~$81K. Not even in the ballpark of what most consumers can afford.

Re:Show HN: I've built a locally running Perplexity clone by dns_snek on Thursday April 04, 2024 @07:49AM (#39929134)
> Why?

Privacy, security, latency, offline availability, access to local data and services running on the device, just to name a few.

Re:Show HN: I've built a locally running Perplexity clone by ilc on Thursday April 04, 2024 @10:58AM (#39931268)
Big Tech + Countries: Those all sound like great reasons to centralize all access to AIs!
Re:Show HN: I've built a locally running Perplexity clone by ThinkBeat on Thursday April 04, 2024 @06:23AM (#39928533)
"640kb will be enough for everyone." (Gates)

I think that the models will evolve and grow as more powerful compute/hardware comes out.

You may be able to run scaled down n versions of what state of the art now, but by then the giant models will have grown in size and in required compute.

The 6 year old models will be retro computingish.

Somewhat like how you can play 6 year old games on a new powerful PC but by then the new huge games will no longer play well on your Old mach

Re:Show HN: I've built a locally running Perplexity clone by lobocinza on Thursday April 04, 2024 @11:00AM (#39931283)
There will be demand and supply for both cases.
Re:Show HN: I've built a locally running Perplexity clone by bufferoverflow on Thursday April 04, 2024 @11:19PM (#39938375)
Unfortunately training is insanely expensive. StabilityAI is struggling with staying alive. And Anthropic wants to spend $100 BILLION building a supercomputer just for training.

That's why I think these private companies will have the best AIs for many decades.