This moment in time is a little fun, Anthropic IPO around the corner, OpenAI maybe next year, blabla. I'd like to record my thoughts not because there's any merit but for myself to look back at, later.

Two things are happening at once, around me. The first is that my boss thinks AI is very fun, and we are doing a few such projects for clients, got a department for it and such. We are given subscriptions. At the same time, the talk on the internet is mostly around how insanely high the figures thrown around by the AI companies are, how anthropic is a bad investment due to their datacenter obligations, how much everything is propped up by borrowed money. It's all very "fun" to think about...

Weak work culture

I think we are very unopinionated at work, to a fault. "We" as in the company. Employees are given extreme freedom in choosing their tooling, which i appreciate, as i am one. But it also means the burden is on us - no guidelines, no recommended procedures. Same with the AI tooling - it's there for us if we want it. The official policy is that you may use it if the client agrees. But at the same time, the corridor talk is different. "We're all managers now". "Nobody will write software again". "We need to speed up to stay competitive". Who's saying this? The boss, the managers, the AI department.

Financials

To me this all seems like a few people's frantic pursuit. A very expensive pursuit. They can't slow down now, if the project stops swimming it'll asphyxiate. Everything that will happen has to happen NOW, and by god we will make it so.

But i've pointed and laughed at other "obviously unsustainable" economic machines before, awaiting their imminent collapse, but that's not what happens. I never get to say it, i never get my i-told-you-so moment. The thing dies by getting subsumed by something else, becoming uninteresting, background noise.

What's da prognosis

So what i currently think, no grand prediction or anything that i thought too much about, just some intuition: it'll go over. Some neoclouds will fold, datacenter buildout will drastically slow, maybe stall. New models will still be produced, i think of google mostly here, things that are already institutions that can use them. Some people just want to burn money. I have too little knowledge of the chinese space to say anything there. But these LLM's do serve enough purposes for people that they'll be worthwhile. They're good for spam, SEO, and i can't deny coding. It'll be interesting to see where this all lands. What flavors will exist, and where they'll be put to use...

But i do hope we all stop making a spectacle of each new release. It's not that interesting anymore, we know what these are! If we strip the hype and the novelty, they have a specific shape, obvious limits. They'll be a boring corpo component to put in the boring corpo machine, and the current AI rockstars will be out, and we all have something else to care about.

Perhaps? Maybe?

Weh

Again let this be as naive as it may be. I just want some datapoints for when i look back at this in a few years time, or whatever.

No thoughts here on copyright, i don't know what will happen there. Personally i don't believe in intelectual property. But industrial replication of other people's work isn't good either. Hmm...