The AI summer

Well, not just in the last five weeks, but certainly in the last six months.

And there was the Apple event two, three weeks ago, two weeks ago, which, I mean, I'm not going to do the detailed Apple post dive because A, it's too late, and B, if you care, you've seen it.

But it was interesting because it sort of pointed at all of the important questions.

And certainly I think, you know, if we were having this conversation, when we were having this conversation six or nine months ago, it would still get quite difficult to kind of tangibly pin down what were the kind of big structural questions.

And as we've got quite a lot further into deployment, even though this is still only sort of 18, 19 months old, the energy of AI is, that is, since ChatGPT 3.5, we're a lot further towards a tangible sense of like where the key questions are.

And so the Apple event is first of all, like the primary question in AI is, is this basically more software and features and capabilities and another platform shift and it will play out like machine learning and cloud and mobile and smartphones and everything else? Or are these things just going to be able to do all software? And so are these going to take over all the software you have and do a whole bunch of, and do all the stuff that you can't do as well? And even if you don't go all the way to ADI, and I wrote a thing about the Apple event last week where I called this AI maximalism.

There's an AI maximalist view that says LLMs are just going to do the whole thing.

They'll do everything.

They'll pull every complex task that might need five pieces of software, or maybe there is no software for it.

Now ChatGPT or Anthropic or Claude will just be able to do the whole thing.

Or is it going to be features? Which is obviously what Apple has said.

And of course, the incumbent always wants to make new things a feature.

So that's what they've done.

And so there's that.

Does it run in the cloud or does it run in the device or both? Are the models generic commodities that are sold at marginal cost or are they unique and have differentiation? How many models are there going to be? What's going to happen? Who's the end user? Yeah.

Seems to be a big one.

Are we selling to companies? Are we expecting consumers to use this on a daily basis?

Yeah, well, Apple's put Apple sales to consumers.

So that doesn't really apply to them.

But there's all and, you know, does NVIDIA have a moat? And obviously Apple is in a kind of unique position in that they can make their silicon and build APIs on top of it.

So Apple sort of pointed at all of these sorts of core questions.

And it seemed to me, and this is sort of the framing I'm using for my presentation, is that you could kind of do a Venn diagram.

Or the questions that matter inside tech and the question that matter outside tech if you're an engineering company or a car company or a bank or retail or something or a marketing company.

And there's an overlap in the middle.

And so there's a lot of questions that basically only matter to people in tech like you know.