Status Updates From What to Think About Machine...

For decades, the field of artificial intelligence suffered the syndrome of moving goalposts. As soon as an intelligence development target was reached, it was redefined, and no longer recognised as "intelligent".

Many imagine coldly objective future computers, but no one loves a know-it-all. People will prefer modest, polite computers that are deeply subjective. Our machines will not contradict our inanities: they'll gently suggest: "That's an intriguing idea, but weren't you also thinking that..."

[Written in 2015]

Chess was conquered by analysing more moves, Jeopardy! was won by storing more facts, natural-language translation was accomplished by accumulating more examples. These examples suggest that the secret of AI is that there is no secret. Like so much else in biology, intelligence seems to be a collection of really good hacks.

[Written in 2015]

Drones are designed to attack and surveille - but attack and surveille whom? With the right machines, we can expand literacy and knowledge wider into the world's population. But who determines the content of what we learn and appropriate as fact?

The Cambridge psychologist Michael Kosinski has shown that your race, intelligence and sexual orientation can be deduced fairly quickly from your behaviour on social networks. On average, it only takes four Facebook likes to tell whether you are straight or gay.

We can't deal well with a threat only now looking like a small, distant, dark cloud on the horizon: AIs that perform better than we do at the very highest levels. This cloud need not concern us now. It may never appear. Right now, we have trouble making an AI that passes the Turing Test.

[Written in 2015 by Gregory Benford, professor emeritus of physics and astronomy]

"The step from human-level AI to superintelligence will most likely be quicker than the step from current levels of AI to human-level AI. Superintelligence could well be the best thing or the worst thing that will ever happen in human history," writes Nick Bostrom in 2015.

At last, someone here who actually does know what they're talking about.

My one-sentence summary of the first hundred pages: the world's leading experts on AI, science, philosophy etc don't know shit, you'd be out of your mind to believe anything they said.

And if that's the smartest people in the world, just imagine how much worse things are with the rest of them.

My hot tip for this morning: if you're going to read a book on futurology, wait until it's been available for at least ten years. You'll enjoy it much more, and it will also make a lot more sense.

perhaps Pinker is being overly optimistic that AI systems built by alpha-male techbros won’t project alpha-male techbro anxieties, but I’m attracted to his reasoning.

Murray Shanahan on Conscious AI and the ability to suffer remains relevant “which if any attributes we associate with human consciousness is necessary in human level AI?”