Where are we in AI? It’s been a while since I wrote so here it goes. You can read more about this in the first part of the current state of Gen AI.

Infrastructure

There’s been huge amounts of investment in GPUs / data centers for AI. You can see this by the size of the AI infra companies and the sense of who have been the winners in this first phase of the cycle. I hear much discussion these days on how we’re going to power up those datacenters (electricity) and on whether we’ve gotten ahead of ourselves on the infra side. I’d love to see this quantified more.

Developer tooling

It seems like an all out battle out there to win the hearts and souls of developers. Tons of tooling being packaged (I don’t mean just APIs for models). LLM development tools, data cleansing / DataOps, developer productivity, etc. I think there’s a sense that those who win developers have a beachhead into bigger territory.

Applications

The buzz on GenAI applications has subsided. It was hot a year ago and now it’s quieter. I get the feeling those doing well prefer to be quieter while they take market share and everyone is busy with channel development and enterprise deployment. This is the hard work of enterprise sales.

Models / Research

There’s a lot of development here. Compute costs continue to come down drastically if you don’t want to always stay on SOTA models (state of the art). Open source are becoming realistic contenders (albeit still somewhat behind commercial models). The state of research is still very hot. Everyone is looking for the next game changer improvement in models.

Consumers

We’re moving from AI being used by a visionary segment in the technology adoption curve to the early majority for weekly usage of core GenAI tools. People are becoming more habituated to leaning on these tools. Consideration is high when trying to do work. The early majority is pragmatic and has found tremendous value. Expect a solidification in the user base and increasing minutes / user. Multimodal tools are not yet productized or top of mind. That should come later in the year.

Enterprises

Organizations are now more solidly working through GenAI. There’s more swim lanes towards production. And teams now have more time with the tools and working together across the org (tech, business, risk, legal, etc). We’ve moved from enterprises scrambling to run pilots to more methodical programs of work. This is good.

What are you seeing? What’s your viewpoint on the current state of Gen AI?