[TimPC]

Most tech people know very lot about very little. Most programmers know next to nothing about neural nets. Most neural net researchers know very little about self driving cars. Most self-driving car researchers are experts in incremental changes and know very little about what grand breakthroughs are likely in the next 2-3 years. To give you an idea, most go players and most AI researchers were shocked by AlphaGo. The expertise didn't help.

I think for technologists in particular expertise can be very dangerous as we tend to think we know a lot more than we do. Predicting the feasibility of an autopilot on a 1-2 year time horizon is a highly specialized difficult analysis. It's definitely not close to something most tech people can do well.

[marvin]

Is it true that most AI researchers were shocked by AlphaGo? These are the people who say it's impossible we'll get some qualitative breakthroughs yielding human-level or super-human general intelligence in the next few decades.

So if the majority of the group is that shocked by an incremental improvement, it would be better if we just interpreted their long-term forecasts as meaning "no idea" instead.

[6gvONxR4sf7o]

It would be best to interpret most long-term forecasts as meaning "no idea" instead. This goes for all groups, not just just ML researchers. Now, they have a little edge, and a big edge in the short term, but it's closer to "no idea" than anything else in the long term.

[afthonos]

It’s quite true. A take I found insightful (I think it might have made the top page when it was first posted):

https://intelligence.org/2017/10/13/fire-alarm/