One potential regression that I see is that the current generative models are abandoned, after being ruled as "infringing copyrights" by multiple countries. The tech itself won't disappear but it'll be considerably harder to train newer ones.

The most problematic part is however if one of them survives; likely Google. That would lead to a situation as in your second paragraph.

Nvidia 3060 to 4060 progress and recent ssd news be like

Yeah I think Google podcasts is getting killed in a few months.

Honestly they're barely hacks at this point, hacking implies some kind of social engineering, internal leak or mad computer skills. The last few major data breaches have been more along the lines of leaving things with default passwords or storing customer data in plaintext.

"Activate the flood."

It's really bizarre how so many business can exist while not turning a profit just because there's a profit potential because they rose in popularity really fast, Uber will be 15 years old this year.

I hate /r/technology. Thanks for reminding me

I just want good, cheap, and mass production ready solid state batteries of any sort. Right now, anything that's on the market either isn't good or isn't cheap, and none of it is mass produced... Often all three.

If we get over the hurdle of something we can mass produce for cheap that's as good as, or better than the existing lithium tech that we have, I'm in.

The US might, but I don't see the EU giving up on them without major trade concessions.

I like this one. I've been hoping for some host-your -own AI models that I can dump into a system with a bucket of TPUs and a decent GPU for processing and get my own version of something like chat GPT at home then train it on the entire collective works of documentation and help articles about the software I usually do support for so it can act as a defacto repository of "natural language" chat/search for troubleshooting.

The SSD price hike prediction is really fucking infuriating. Doesn't seem like we're aiming to replace HDDs ever at this pace.

You're a leap year.

Yes!

Or the complete opposite, we will realize that AI is hard and LLM s will probably not take over the world. Self aware AI is probably much further away than we think. But who knows! 🫥

I think we're still a bit far off from that. No doubt the models will be quite good, but they won't be anywhere near general intelligence.

It probably won't happen until we move to new hardware architectures.

I do think LLMs are a great springboard for AGI, but I don't think the current hardware allows for models to cross the hump to AGI.

There's not enough recursive self-interaction in the network to encode nonlinear representations. So we saw this past year a number of impressive papers exhibiting linear representations of world models extrapolated from training data, but there hasn't yet been any nonlinear representations discovered and I don't think there will be.

But when we switch to either optoelectronics or colocating processing with memory at a node basis, that next generation of algorithms taking advantage of the hardware may allow for the final missing piece of modern LLMs in extrapolating data from the training set, pulling nonlinear representations of world models from the data (things like saying "I don't know" will be more prominent in that next generation of models).

From there, we'll quickly get to AGI, but until then I'm skeptical that classical/traditional hardware will get us there.