How to Write a News article on the Dangers of Artificial General Intelligencepost by maximkazhenkov · 2020-02-28T02:14:48.419Z · LW · GW · 1 comments
I noticed that a lot of news articles/op-eds on AGI follow the same scheme. Not sure if this has been written about before so I'm just going to put it out there.
1. The strawman
"Let me explain to you how AIs today are totally not AGI by pointing out all the simple things even a human child can do with ease that AIs struggle with. Have I mentioned we don't have real AGIs yet? How silly of you to still think that we already have AGI! No, really, I assure you AlphaGo is definitely not an AGI. It's not even as smart as a five year old/lab rat/retarded cockroach."
2. Impress me.
"Talk to me once you've solved natural language processing/causal reasoning/AI doctors/*insert any AGI-complete problem*. Anything before that is just fear-mongering/sensationalism"
3. Serious problems for serious people
In the mean time, we need to focus on real issues serious researchers worry about, such as:
- Trolley problems for self-driving cars (How convenient that human drivers are too slow to make a decision anyway)
- Adversarial examples for image classifiers (But when humans fail, it's because the image is genuinely confusing, you see)
- Pranksters altering traffic signs to confuse self-driving cars (How about I just steal the sign? Checkmate.)
- Racist AIs (Not gonna touch that issue with a ten-yard pole)
Of course this is just a caricature of AI articles in the news media not to be taken too seriously; there are also plenty of articles with more nuance.
(Someone train GPT-2 to write AGI news articles)
Comments sorted by top scores.