Well, howdy there internet people, let's bow again. So today we are going to talk about

Harvard and AI and polling and something that happened up at Harvard at the Kennedy School that is

interesting in and of itself but the reason it's happening is more

interesting. So up there they conducted some

experiments let's call it, where they used AI, large language models, and kind of trained AI agents to be either conservative or liberal and then polled them, conducted

polling of the AI. Now overall it actually seems like it did okay.

However there was a major misread at one point, and that deals with Ukraine.

The conservative AI responses, well they were hawkish.

They were in favor of supporting Ukraine and intervention.

The liberal AI, well, I was more opposed to it.

That's the exact opposite of what's occurring.

If you missed it, Trump has basically promised

to cut aid at this point to Ukraine if he is elected. So why did that happen? Because the data set that the AI was used or the AI

used to be trained with, it ended prior to the political positions being established. So again it's a good

reminder that AI is not AI the way you think of it from science fiction. It is a large

language model. It is scraping the information that is on the internet,

collating it, and using that and kind of giving it back to you. If the data set isn't there, well, it's

going to give you a bad read. That would be incredibly important when it comes to polling

because a lot of issues are, well, they're recent events, their political events.

So the data set that the AI was trained with,

it may not include it. And then it also can't really factor in

other things that are important to polling as far as things like name recognition and stuff like that.

...