[gruseom]

There are many forms of wishful thinking. Apocalyptic disaster is also one.

You seem nice; forgive me if I get too dismissive here.

As a software practitioner it seems to me obvious that we are so many light years away from the kind of software the Singularity people are talking about that the whole thing is all a fantasy club, and a little embarrassing. It's like the detailed debates 19th century radicals used to have about society after the Revolution.

Also, the Singularity people always seem to do that moving target thing where as soon as you say one thing, they go: But that's not the Singularity, that's a misunderstanding of the Singularity. Leaves me thinking: it must be awfully subtle.

[jmmcd]

A software practitioner, you say? As a researcher in machine learning, I think you're wrong. I agree that recursively self-improving AI is not around the corner, but I think it could happen in a few decades.

Even if it has a 1% chance of happening in the next century, I would still allocate resources to thinking about how to mitigate the risks (and maybe look foolish if it never happens) than leave things to chance and end up regretting it. By which I mean, a runaway AI destroys humanity or other bad outcomes.

As for the moving target -- since a singularity, if it happens, would be very important, it definitely makes sense to talk about lots of different scenarios. Anyone who dismisses a plausible scenario just because that's not what Vernor Vinge said, or what Ray Kurzweil said -- well, they can safely be dismissed.