Re: Oppose AI in crime solving

Anders Sandberg (asa@nada.kth.se)
02 Jun 1999 01:14:14 +0200

"Eliezer S. Yudkowsky" <sentience@pobox.com> writes:

> Prosecutor: You're accused of killing eight nuns with an icepick.
> Defendant: What's the evidence?
> Prosecutor: Our neural network says you did it.
> Defendant: Why does it say that?
> Prosecutor: Um, nobody knows.
> Defendant: That's it? That's your evidence?
> Prosecutor [to jury]: Hey, look at all these cool blinking lights!
> Technology! Whoo-ah! It's never wrong!
> Jury: Guilty.
> Judge: I sentence you to death.

Hmm, sounds a bit like the problem with "truth" machines. How can we be sure they are never wrong?

Of course, using neural networks for sifting through databases and getting an opinion can be very useful. The trouble seems to be that people tend to offload thinking and responsibility too easily.


Anders Sandberg                                      Towards Ascension!
asa@nada.kth.se                            http://www.nada.kth.se/~asa/
GCS/M/S/O d++ -p+ c++++ !l u+ e++ m++ s+/+ n--- h+/* f+ g+ w++ t+ r+ !y