Highest Rated Comments


dangersandwich17 karma

also a lot of brilliant technical people figuring out how to beef up encryption and security to make it way harder for agencies like the NSA to spy on everyone en masse.

Shameless plug for /r/netsec.

dangersandwich12 karma

In some fields it even does already (Watson on Jeopardy, Alpha.go, Deep Blue...). Currently Computers still look & are inferior, but again, double this every two years!

Your examples are of sandboxed applications of pseudo-AI (neural network learning). These technologies benefit from raw computing power, but Moore's Law doesn't change the fact that computers are only good at routine, codifiable tasks. Advances in machine learning (ML) expand this capability to include some non-routine tasks, but with limitations: ML algorithms need existing datasets to help the classifier "learn" recognition to adapt to outliers, since it is based on statistical analysis of inputs. So what can ML do when there is no data?

You still need humans to develop new ideas, designs, and abstractions, which is the core of what engineers do. Again, comparative advantage. While computing power will undoubtedly continue to reach new highs, it's only one variable in the equation and doesn't even begin to address the comparative advantage of human labor, or whether it is substituted or complemented in various labor markets.

dangersandwich10 karma

Thanks for doing this AMA!

I am convinced that in time engineers will be replaced by computers. This may be faster than we expect (Moore’s law of doubling approximately every two years), and will probably take us by surprise.

What time scale are we talking about here? 25 years? 100 years (a generation)? Three generations?

As an engineer with an interest in economics, there is currently little (if any) evidence in the economics community that automation will be able to completely eliminate demand for human labor, especially the type of labor that requires critical thinking and abstract thought processes. Automation technologies have historically acted as a multiplier on human productivity where humans have a comparative advantage, and it's very hard to imagine this changing pre-singularity.

Skill-biased technological change (SBTC) will require humans to continually develop more comparative advantages through more education and training, and automation technology can help us do that by freeing up more of our time to work on improving the best technologies while automating whatever we can, which results in a continuous cycle of increased human productivity.

I can see simpler engineering design tasks, like for example, a) designing enclosures for sensitive electronics; b) designing simple machined parts; and c) reducing part count (DFMA) getting replaced by automation technologies, but this is an example of technology acting as a complementary to engineering, not as a substitute. Certainly some engineers that exclusively do these types of simpler design tasks may get replaced, but this is a case of wage/skill mismatching rather than automation being a perfect substitute for all engineers.

dangersandwich3 karma

What kind of computer do you use at home, and what kind of portable devices do you use, if any?

dangersandwich3 karma

What should I do as a recent engineering graduate if I want to work for JPL?