While we're all having fun watching people play with GPT-3 right now, I'm feeling the same itch on the back of my head I felt back in the mid-2000's when ordinary people started piling into social media and carrying the Internet around in their pocket. The nagging thought that this toylike tech could end up having an unpredictable effect on the world. What is the "Cambridge Analytica" moment of neural network tech going to look like?

My money is on GTP-3 and it’s descendants eroding the ability to discern fact and even deliberately written fiction, from machine generated junk.

As these models get used to drive summaries and other text processes their pseudo-meaning will creep out from a kind of textual uncanny as everyone gets used to it... eventually it becomes dangerously difficult to glean actually useful knowledge from repeatedly reprocessing the now polluted corpus of “real world text” and we get some real world consequences, much like the failure of engineering at NASA to convey the risk of launching Challenger in the cold, when forced to summarise it to a single slide it conveyed the opposite meaning to management. When we stop doing the summary ourselves were in for trouble eventually.

I think people generally give a lot of undue clout to what Elon says, just because they pattern match "Elon Musk" with "smart guy". Granted, I think the hate bandwagon against him is just as silly.

I just think he's a fallible human like the rest of us, as unrelatable as he may be for various reasons. And he clearly has trouble keeping it together sometimes if his twitter is any indication. Him being terrified of something is all the more reason to be skeptical, because fear indicates a heightened emotional state, which is something he has a problem acting rational under.

That said, I firmly believe AI is gonna be a big fucking problem. Not necessarily because they'll gain sentience and enslave us, but because of how humans will use them. I trust that I don't need to go into detail about the utterly terrifying, civilization destroying disaster associated with anybody being able to create entirely fabricated, fake video, indistinguishable from reality, on a whim.

A synthetic, autonomous agent as generally intelligent as human, or an order of magnitude moreso, could indeed be terrifying.

But who's going to give such an intelligence any degree of autonomy?

Something 100x smarter than a human is going to figure out how to get a body, or something equally useful.

Why would it be given and not taken?

"Power concedes nothing without a demand" -Frederick Douglas

Because you need some autonomy to take more autonomy.

Next he'll accuse Watson of being a pederast.

Nationalistic flamebait is not welcome here, so please don't post like this to HN. It leads to nationalistic flamewars, which are equal parts repetitive and nasty.

https://news.ycombinator.com/newsguidelines.html

How about both? China using AI to take over the world?

Personally I think that WWIII is likely if there isn't some kind of resolution to the extreme divergence of culture and political structure between the US and China. AI is one of the ways that China could actually become a near-term military rival.