The Goldfish Chronicles

I’m looking forward to the scifi novel which bases its premise on spam filters and spam bots increasingly reaching artificial intelligence by passing the Turing test. It seems to me that both can be viewed as the two halves of this test and are playing out a socio-genetic algorithm to get there.

There are a few interesting areas to deal with in this vision:

1. Spam filtering isn’t only concerned with whether something sounds like a human, but rather whether it’s talking about something that the human it’s defending is interested in. To that end it’s a higher bar than a Turing test and humans will start to fail that test (to the point where the human checking their spam folder does not recognize that email as not spam). We’re probably here already.

2. People don’t generally respond to spam. So it’s a one-off Turing test rather than the ability to hold a conversation. Spam is generating believable messages, but is not parsing human text (one assumes). So you’d have to make the ‘one item of non-reality’ that you put into the system the notion that spam companies start writing IM spam to have entire converstions and going down the Eliza route.

3.  The people who do reply to spam are not the smartest. The AI that this socio-genetic loop will create will be equally not smart.

Yes it’s true… the singularity is coming, and when it arrives it will mean nothing more than our stopping caring about machine intelligence and caring more about whether the intelligence talking to us passes our bar (and our automated agent’s bar).

[PS: I’m sure all of this is in multiple novels and short stories already… there is nothing new under the sun ]

This entry was posted on Wednesday, April 22nd, 2009 at 10:02 am and is filed under Opinion. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.