This is the reason why they're not going to move on device anytime soon. You can use compression techniques, sure, but you're not going to get anywhere near the level of performance of GPT-4 at a size that can fit on most consumer devices

I think we’ll see completely new architectures dominate in the near future, ousting the transformer. I am strongly suspicious that, while impressive, transformers use several orders of magnitude more compute than is “needed” for the tasks they perform—if for no other reason because the human brain performs similarly and it only draws 20 watts! And it isn’t even an engineered system, jus the product of a very, very long history of natural selection! I fully anticipate that we’ll see AI in the near future that achieves human-level performance on sub-human power budgets like the ones you’d be constrained by on a phone :)

"neat future" is very ambiguous. At the moment there is nothing even close to transformers in terms of performance. I suspect you are right in general but I'm not sure about the "near future" part, there needs to be a pretty significant paradigm shift for that to happen (which is possible, of course, I just don't see any hints of it yet).

RWKV is an attention-free architecture that's showing promising scaling at a similar level to Transformers right now! There's also recently been Hyena, which uses a new mechanism that's kind of a weird mix of attention, convolution, and implicit modelling all at once. It's shown promise as well. Remains to be seen if these competing methods will truly scale as well as Transformers, but I've got my fingers crossed. Only a matter of time!

I agree that "near future" is quite ambiguous though. If I were to disambiguate my claims, I think I'd personally expect a Transformer-killing architecture to arise in the next 4-5 years.

> but ain't no way we run models meant to run on 8 nvidia A100 on our smartphones in the next 5 years

m$ has been working on an AI chip since 2019 so i think we will.

They still can't beat the laws of physics. It's just the kind of density that doesn't seem feasible to put in a smartphone.

Human brains seem to be of the same order of magnitude in terms of size, and seem more nuanced and capable than GPT4.

The human brain is also three-dimensional, heavily interconnected, and has built-in thermal management at every scale. Chips are much faster, but still operate on the essentially linear memory cells, and this limits how many matmuls you can do per second. If we can figure out true connectivity without doing tons of matmuls, then we should be able to massively cut computational demands of models.

It's not apples to apples comparison. There's a multitude of tasks that human brains are very bad at.

GPT excellence is in raw knowledge and answering machine. You won't find a single human brain that can hold the same amount of knowledge

And? Nvidia has been working on AI chips for years.

An A100 is about the size of a brick, there is no way we're fitting those 8 bricks in a phone in the next five years, without even thinking about heat management

An A100 HGX server is ~6kW of power consumption (and associated heat), while an iPhone is O(1W). I agree that a 6000x increase in energy density or 6000x decrease in power consumption is unlikely in this decade.

I mean we could maybe put uranium in iphones...but not that many transistors

the only thing I can say to this is that Apple have seemed laser focused on tuning their silicon for ML crunching, that that focus is clearly now going to be amped up further still, and that in tandem the software itself will be tuned to Apple silicon.

GPUs on the other hand are pretty general purpose. And 5 years on a focused superlinear ramp up is a long time, lots can happen. I am not saying it's 100%, or even 80% likely. It'll be super impressive if it happens, but I see it as well within the realms of reason.

Apple's new M2 Max has a neural engine which can do 15 trillion flops. Nvidias's A100 chip (released almost 3 years ago) can do 315 trillion flops. Apple is not going to close this 20x gap in a few years.

> this 168x gap

FTFY, remember it takes 8 of those to even load the thing. And when the average laptop has that much compute, GPT 4 will seem like Cleverbot in comparison to the state of the art.

right, it's a huge challenge.

I think the tuning the models to the hardware piece is important, and of course there is much more incentive to do this for Apple than nvidia because of the distribution and ecosystem advantages Apple have.

But also, I don't know... let's see what the curve looks like! It's only been a couple of years of these neural engines. Let's see how many flops M3 can hit this year. And then m4 the next. Again, 5 years is a long time actually when real improvement is happening. I am optimistic.

At some point, they will put the models in silicon. I’m curious as to when… 5yr?

That doesn't sound likely with the current architectures. There may be some kind of specialisation, but NN is like the chip design nightmare. We can't do chips that that many crossed lines. It's going to have to keep the storage+execution engine pattern unless we have done breakthroughs.

"More specialised than GPU" is the game for now.

Not even with 3d chip-stacking?

Well, we'll see what the future manufacturing brings, but right now we're not even at thousands of layers (as far as I know... please link if there's been more), and we'd need to be in hundreds of thousands range. Given the rate of defects also adding up and the need for some way to dissipate the heat... Yeah, still lots of challenges there.

Try comparing M2 with an actual consumer level GPU, not a supercomputer...

Curious, why do you think that? My knowledge is limited to marketing material and my M2 vs my 3090, and my conclusion so far would be that’s in every hardware makers marketing claims the past couple years.

The M2 and the 4090 are both very general purpose. In fact, the 4090 allocates proportionally more silicon area to the tensor cores than Apple allocates to the neural engine.

The M series is basically the only "big" SoC with a functional, flexible NPU and big GPU right now, which is why it seems so good at ML. But you can bet actual ML focused designs are in the pipe.

Well, and gobs of RAM, which only top end cards can compete with.

I am really hoping Intel disrupts this status quo. The only thing stopping AMD/Nvidia is anticompetitiveness.

Does llama.cpp use the npu or metal? I thought it uses NEON instructions? So currently those units aren't even used.

I don't think so. M chips just happen to have a really good memory subsystem and good simd performance through accelerate, so the CPU performance is pretty good.

Some stable diffusion implementations can use the NPU or GPU, or (experimentally and unsucessfully) both.

> but ain't no way we run models meant to run on 8 nvidia A100 on our smartphones in the next 5 years

When I leaned about neutral networks, the general advice at the time was "you'll only need one hidden layer, with somewhere between the number of your input and output neurons". While that was more than 5 years ago, my point is - both the approach and the architecture changes over time. I would not bet on what we won't have in 5 years.

Vicuna (LLama) 13B is pretty good IMO. A 20B model can definitely fit in RAM on future devices.