[duke_sam]

This is one of the many initiates where the process and details matter. What’s the due process here, can people see the prompt that judges them to be supporters? Is AI just highlighting “supporting” content for human review?

Even if you support the idea in principle, the actual practical implementation should be treated with skepticism.

[guappa]

I think it all boils down to: if sounds_muslim(name) → expel()

[BoredPositron]

It’s probably just the family guy skin color chart.

[jimvdv]

Even if it is intended as a hint to human observers, during a big scandal in my country a few years ago turned out all the gov officials where blindly following the computer advise. Many where falsely accused of fraud by tax authorities, almost all where from ethnic minorities.

[bigolkevin]

Or Israel reportedly ordering assassinations in much the same way.

https://www.972mag.com/lavender-ai-israeli-army-gaza/

[rwmj]

There's no "due process" for US visa applications. The process is, possibly deliberately, kafka-esque. Edit: I wrote about my visa process here: https://news.ycombinator.com/item?id=23942466

[pbalau]

> can people see the prompt that judges them to be supporters?

Not all AIs are LLMs...

[ben_w]

While true, the general principle remains important irregardless of implementation specifics.

If it were GOFAI, you can rephrase as "can people inspect the decision tree implemented by if-else blocks?"

[permo-w]

I’m not sure why you're being downvoted for this. the article does not state that the AI tool proposed to be used is an LLM. in all likelihood it is, but why assume?