TechTakes

ah yes the only way to make LLMs, a technology built on plagiarism with no known use case, “useful for any actual ai application” is to throw a shitload of money at nvidia. weird how that works!

A lot of these systems are silly because they don't have a lot of RAM and things don't begin to get interesting with LLMs until you can run 70B and above

The Mac Studio has seemed an affordable way to achieve running 200B+ models mainly due to the unified memory architecture (compare getting 512GB of RAM in a Mac Studio to building a machine with enough GPU to get there)

If you look the industry in general is starting to move towards that sort of design now

https://frame.work/desktop

The framework desktop for instance can be configured with 128GB of RAM ($2k) and should be good for handling 70B models while maintaining something that looks like efficiency.

You will not train, or refine models with these setups (I think you would still benefit from the raw power GPUs offer) but the main sticking point in running local models has been VRAM and how much it costs to get that from AMD / Nvidia

It's FOMO amplified by capitalistic competition. No company wants to be the one left behind. I guarantee Google, Meta and even OpenAI know the limitations of their products. They don't care, they just want to be at least as good as their competitors, because they assume at some point one of them will reach "good enough." And at that moment, if they're not in position to grab market share, they'll lose a once-in-a-generation chance for billions or trillions of dollars in value.

We're the casualties, because the people in the middle - companies with no AI but whose C-suite buys into the hype - demand we use unworkable products because they're too willfully ignorant to know they're not panaceas to whatever is bothering those C-suite execs at the moment.

(I’m making a generous assumption by suggesting that “ready” is even possible)

It was ready for some specific purposes but it is being jammed into everything. The problem is they are marketing it as AGI when it is still at the random fun but not expected to be accurate phase.

The current marketing for AI won't apply to anything that meets the marketing in the foreseeable future. The desired complexity isn't going to exist in silicone at a reasonable scale.

Apple Intelligence and the first versions of Gemini are the perfect examples of this.

iOS still doesn’t do what was sold in the ads, almost a full year later.

capitalism working against itself

More like: capitalism reaching its own logical conclusion

Betamax actually found use in Television broadcast until the switch to HDTV occurred in 2009

Betamax had better image and sound, but was limited by running time and then VHS doubled down with even lower quality to increase how many hours would fit on a tape. VHS was simply more convenient without being that much lower quality for normal tape length.

HD-DVD was comparable to BluRay and just happened to lose out because the industry won't allow two similar technologies to exist at the same time.

Neither failed to do what they promised. They were both perfectly fine technologies that lost in a competition that only allows a single winner.

Dude don’t throw Betamax in there, that was a better product than the VHS. AI is just ass.

@PixelatedSaturn @dgerard

> It’s coming for sure

You got a source for this?

you're right tho

Doch!

You're right - and even if the user is not conscious of this observation, many are subconsciously behaving in accordance with it. Having AI shoved into everything is offputting.