Read on Twitter
I was at a Machine Learning/AI conference in Seattle several years ago and there were several keynote speakers including one of from the Allen Institute for AI. I was struck by one of the comments made by one of the keynote speakers at that conference. 1/n
He postulated that while several in AI community would have you believe that we had reached the moon (this was around the time Amazon's Echo was already out but not Tesla's autopilot) we had essentially just climbed a tree. 2/n
I was reminded of this today when I asked Google Home when the San Mateo County stay at home order was supposed to end and it was not able to provide a useful answer. 3/n
I understand that this is just one query but over several questions asked of multiple "smart" devices, I have come to believe that the "AI" from the big three, Amazon, Google and Apple, still has a long way to go to challenge anything close to human intelligence. $AMZN $AAPL 4/n
As a programmer and someone that has used machine learning, my questions are often phrased in a manner that could help the "AI" and I often get better responses than others in my household. The responses just aren't good enough in many cases.
The power of AI was readily evident on a project I was working on where the machine learning models were able to come to the same conclusion in about 3 months that it took us to reach over nearly 2 years of traditional data analysis. 5/n
I was also reminded of just how dangerous it can be to fully rely on AI that may not be ready for "edge cases" on a drive to Sacramento a couple of years ago. I was on Tesla's autopilot on a freeway with light traffic. Most cars were traveling close to the speed limit. 6/n
I was on the left most lane and spotted the driver two cars in front of my swerve all of a sudden. This put me on alert and I took control back from auto pilot. When the car in front me also swerved, I finally saw what was going on. 7/n
There was ladder lying horizontally across the fast lane. It was fully within the lane and I switched lanes just in time to avoid the ladder that could have caused considerable damage to the front bumper. Who knows what it could have done to the battery pack. 8/n
This was an interesting edge case where my actions to avoid an obstacle were based on what I saw two cars ahead of me and I don't know what the AI would have done when the obstacle finally presented itself when the car in front of me swerved. 9/n
I called the highway patrol and told them about the ladder including the nearest exit from the ladder. I still think back to that situation and am reminded of the conclusion of the book The Second Machine Age. 10/n
The authors outlined a future that included both machines and humans with each doing what they do best. I wrote a review of the book back in 2016 here: https://www.insidearbitrage.com/2016/09/the-second-machine-age-by-andrew-mcafee-and-erik-brynjolfsson/ I think a future run predominantly using AI is a dream that may not be realized anytime soon. 11/11
