Progress in AI Isn’t as Impressive as You Might Think

technologyreview.com

19 points·mutor··4 comments

You can tell that by just looking at various AI startups - insane claims, nothing working to show, some servers crunching numbers but solving no real problem And the only thing that seems to be there is telling cats from dogs, but we've already solved that

And possibly related - what is left of the 'Big Data' fever? Where are we, what did it bring to the economy, did map-reduce change the way we live?

The URL for the original report is mangled in this article; the correct URL is http://aiindex.org/2017-report.pdf

I agree that AIs that can generalize to solve out of sample problems are a long way off and we are in a bubble.

I manage a machine learning team. My hope is that the tooling improves so it will be quick and inexpensive to build models and applications. In other words, make up for lack of AGI with great tools and well trained practitioners.

The difference this time is that artificial intelligence practitioners are not interested in general artificial intelligence. The AI Index report is claims its goal is to "ground the conversation." It does not address social issues. This continues the comfortable conversation practitioners are already having. It does not change its course.

Giving up the goal of general artificial intelligence is a strategy for taking ethical considerations as a starting point. As is always the case, today's AI will be tomorrow's ordinary programming technique (e.g. A). It's just that this time, "When I squint it's not really* evil" is the governor on the engine.

Reply on news.ycombinator.com