Notices by Jason Gorman (jasongorman@mastodon.cloud)

It's basically like if they decided to put a coin-operated toaster oven in every vehicle. You don't actually want toast, but if you don't keep feeding it coins, the car won't start.

@dalias The thing is, if products are redesigned so they won't work without the LLM (like putting a coin-operated toaster in every car and designing the electrical system to not work without it), then whether or not we actually *want* toast will be largely irrelevant.

@dalias And then expect productivity to double because they read it in Forbes

LLMs are a near-perfect rentier technology. The cost of training them to the point where they're even marginally useful is so prohibitive that only people with very deep pockets can do it. The goal then is to make us all reliant on them, by shoehorning them into as many products as possible.

And it doesn't matter whether or not we actually need them if products won't work without them.

There's a small window of opportunity here for software engineers at Meta where they might be able to get another job in this industry.

Engineers at X? That ship's sailed, I'm afraid. But presumably the concentration camps will need IT.

The "Here's a list of people I admire who just happen to have lots of followers" LinkedIn post is a classic of the genre.

Weirdly, when I suggested to the same executives who are currently pouring $gazillions into "A.I." that they could achieve bigger productivity gains by investing a fraction of that in their people and teams, they were like "Meh. We're good, thanks." It's almost as if it has nothing to do with productivity...

I've thought a lot over the years about what makes "work" feel like work for me.

Mindless/reasonless? Yep. Not a fan.

Repetitive? Sure. That's a chore.

Hard? It's better when it's easy, so I'll try to make it easier.

But I think by the far the biggest factor in whether something feels like "work" is when someone is telling me what to do. Suddenly, I don't feel so much like doing it.

Weird, right?

That's an actual thing, though. Search for "autonomy and intrinsic motivation"

"But Jason, how do we stop people from doing things that work against the interests of the organisation?" Well, off the top of my head, NOT REWARDING THEM FOR DOING THAT might be a good start. If incentives are aligned with organisational goals, autonomous people will tend to find ways to give you what you want.

Playing GPT-x at chess reveals a major limitation of Large Language Models: a complete lack of dynamic reasoning.

So many tasks require the ability to plan ahead and evaluate potential outcomes (e.g., is this a good chess move or a bad chess move?)

It's arguably the whole point of intelligence.

This is coupled with a complete lack of temporal reasoning. LLMs have no sense of history or ordering of events.

"Can Europe build a $1 trillion company?"

No. Because market caps in Europe are usually based on actual business performance.

There are a bunch of European businesses with revenue larger than some of these US $trillion businesses, though.

Off to Cornwall tomorrow for a... er... Gorman Xmas. 30 minutes' walk from the cottage hospital where Mum died a decade ago, weeks after being diagnosed.

I've never been back, and with good reason. Took me a year to get back on the horse. I don't know WTF Dad's thinking. But I'm going to find out.

A useful statistic would be what % of AI experts were once estate agents. Second one I've seen today.

The conversation with A.I. stans typically goes like this

Me: "I'm not happy with the results I've been getting with the coding assistant. Often needs considerable coaching and then rework."

Them: "Oh, you must be using it wrong."

Me: "What's the right way to use it?"

Them: * describes exactly what we've been doing *

Me: "That's what we've been doing, and the results are poor" * shows them examples *

Them: "Okay, but does the hotel booking *have* to be in the future?"

Every time I pull on the "Oh, you must be using them wrong" thread with "A.I." coding assistants, it seems to lead to a lower quality bar. That's the "prompt engineering" secret sauce, apparently. Care less.

I did actually say I was afraid this is how Gen A.I. would play out over a year and a half ago. I wasn't afraid that they'd be able to do my job; they clearly can't.

What I was afraid of is that execs and investors won't care that they can't.

If only we hadn't spent the last 70 years training customers to accept software that's hard to change and doesn't really work anyway

Fun fact: "Agile Software Development" is 77% "Software Development"

@thirstybear What really irks me is the opportunity cost - economic and social - of the tech industry focusing all their attention on science fiction while real problems that this kind of investment in time, money and resource *could* solve right now go unattended.

@aral Unless you have $44 billion to hand, of course