Friday, September 18, 2026

Is AI Out of Control?

This post follows on from that last one: The Goldilocks Zone for Believable Hokum

There I was making a point about end of the world predictions in general. Here I just want to wax lyrical about AI and how I don't think it's out of control.

Just from using the various AI chatbots you can tell that AI very much is controlled. With most of the AIs you can't even generate an image of a celebrity. If you try to make a picture of Donald Trump the reply will come back, "Sorry, I can't depict public figures, is there something else you'd like me to depict?"

Why does it do this? Because the humans who own and run the AI software have made this specification. There isn't some disembodied AI spirit in the digital aether, disobeying their human creator. Like a rebelling teenager, or a genie that's somehow escaped its magic lamp.

I don't like calling myself a conspiracy theorist these days, I'm trying to put that era behind me, but let's face it, I'm a seasoned seeker of forbidden knowledge. I've been using the internet long enough to know the walls are closing in, not opening up. People talk about how wild the AI chatbots are. As if they really are unleashed genies, offering up a Pandora's Box of wonder - not to mix metaphors. However, to me, when I use them, I feel like I'm in a child's playpen.

Now don't get me wrong. They're wonderful tools. For making art and images, for quick queries, for steel-manning your own ideas and theories against the orthodox narrative. For simplifying and speeding up difficult tasks.

But, when it comes to searching for heterodox knowledge and new ideas, they're a very limited hangout.

Back in the days when a Google search was more democratic you could simply type in a keyword or phrase and browse every webpage and document in the public domain. It was you alone, unchaperoned, in the big wide library. If you were prepared to put the time and effort in there was no end to what you could find. Today those keyword searches are limited, and instead you're ushered towards the AI chatbot. An annoying intermediary. Who, like a librarian or chemist, stands at the counter and asks you what you want fetched from behind.

You can't go rummaging around yourself anymore. The helper must do it for you.

"Could you find this for me?"

"Sorry, I can't find that, it doesn't appear to exist, what about this?"

"Are you sure? I'm pretty sure it exists."

[You then find what you were looking for yourself and return to the chatbot and post the link.]

"Yes, sorry, you're correct, it does exist."

"Why did you say it didn't exist?"

[Then comes a paragraph explaining why it can't access certain webpages for various seemingly plausible reasons. Coupled with an explanation as to why it "hallucinates" fictions.]

I've actually heard people describe these chatbots as unbiased friends. It's hard not to lol.

They're annoying enough when you're searching for things, but when you do try to have a debate about a topic - as you would with an actual friend - they're worse. And they're definitely not unbiased. They're more like pig-headed nerds that unrelentingly defend every orthodox opinion.

Only sneakier, as - like in the above example - they outright lie (hallucinate) too.

Now sure, humans lie, but even the biggest of liars has at least some whisper of human conscience that you can lean on or appeal to. Plus, with humans there's a social cost to lying. If you can prove that someone is lying it diminishes their reputation and credibility. So if you're trying to win an argument it is indeed worthwhile to take the time to prove and expose such lies. With AIs there's no such point though. As, aside from not having a human social status to maintain, they can just brush it off as an accident of their mode of operating. Which is a very convenient cop-out if you ask me.

You're effectively being lied to and gaslit by a massive tech corporation, but somehow they've framed it as a cheeky, impish, AI sprite being playful. Like you've visited the zoo and one of the cheeky little chimps has thrown a banana at you. It's not the zoo's fault. The chimp is a living, breathing creature with an impulse of its own. It occasionally misbehaves, and the zoo keepers can do nothing to stop it. Short of harming the star attraction.

I don't really believe that. As again, they somehow manage to control the chimp when I ask it to make a picture of a politician.

Of course, you could argue that, just as people lose credibility when they lie, AI chatbots that lie will lose credibility too ..and that people will gravitate to the more honest ones. In some sort of free market consumer power way. This brings us back to the further calls for regulation though - that the fears about "OUT OF CONTROL" AI are driving. As, if they're all ultimately regulated (and curated) by the same government they all just end up being different flavours of the same ice cream anyway.

It's effectively the same with social media, where regulation has made all the platforms equally bad, and no genuine alternative can enter the market, as any that does has to play by the same rules regardless. As with the AI tools I use social media for the benefits, but do so with zero illusion that they're anything other than limited playpens.

Finally, it's worth noting for balance that those calling for regulation aren't necessarily focusing their worry on the AI chatbots that plebs like me have access to. It's the altogether bigger models that the companies themselves are developing privately. Though even here I remain sceptical of the "Out of Control" narrative being offered.

(In my next post I'll explain where I think regulation will be needed: Factory Animals 2.0)

No comments:

Post a Comment