A Boring Dystopia

Ever notice that they're just doing what Clavicular or whatever his name is doing? They're inventing lingo to sound like its not bullshit.

Its not hitting yourself with a hammer its looksmaxxing. Its not standing around being a dork, its mogging. Its not a context window, its the chat scrollback. Its not asking chatgpt its "silicon sampling".

They're making it seem legit by making its own terminology and in-group lingo.

Bullshit artists.

Making shit up, but with extra steps.

The thing is, logically the distribution of opinions or individual situations/beliefs which lead to those opinions, has been baked into the model when the data used to train it was captured, which means that at best and if the entire principle of the thing works (which itself isn't mathematically proven in any way form or shape) they're still getting only poll results for the past and which will not actually change beyond some random noise until the next time data is captured and the model is retrained.

It's like repeatedly using an old picture of a street to make realtime claims about the traffic there.

The ideal would be that clients who actually want useful information will stop paying the pollsters for their useless crap.

The reality will be that slack will be more than picked up by people who want sham poll results to back up their agenda.

I've read a lot of fucked up shit in the last few years - that's the first time I've thrown my phone in response!

(My phone's fine - I just threw it onto the sofa beside me, but still..)

Tempted to start a AI survey company, only take big corp customers. Then always give them the opposite of intended survey results I assume they want.

Really if its fuckin "AI" doing the survey why even go through to steps. Just make some graphs, demographic breakdowns and what not, then let a random number generator do the rest. Don't even need gen ai.

So, use AI to just make shit up, then report that as information? I wouldn't expect anything less in our post-truth era. However, come on Axios, I thought you were better than that.

What could possibly go wrong?

Surprisingly unrelated to the recently compromised npm package.

Their enshittification was inevitable once they were acquired by Cox media

GFC.

This is dumb. But, there's a hint of an interesting idea in there. If LLMs sample all human text and produce statistical averages from it, there's a sense in which they contain a statistically average opinion.

It's basically like how if you use Google to search for "how many calories are there in a" it will suggest the next word. The word it suggests is the statistically average way to finish that sentence. That also means it's the food item for which people most want to know the calories. At least, it's the item they most type into a Google search box. It's just matching text patterns, but it reveals something about people that say a fast food company might find useful.

If you scale up the population of humanity to 8 trillion people and have thousands of years of data in these LLMs, maybe you actually do get useful insights about what people care about. And, maybe that's how you get psychohistory from Asimov's foundation series.