AI, is pretty much the application of Machine Learning... which a nice buzzworthy term... but in reality its “predictive” analytics... or pattern recognition, and application....
The problem with it is, and always will be, there is no real higher “intelligence” going on... there is no inferred reasoning... even when you allow the machine to learn independently and don’t train it.. its simply doing data analytics... and it doesn’t stop, step back and say, my conclusions seem suspect... it just says, that’s the conclusion, and any system based upon it, will act on that conclusion.
You get a statistical probability that given certain inputs, that the the output is X, based upon an algorithm that was trained to predict, or that was allowed to self train...
The classic example philosophical questions of life and existence aren’t there... whether the car its driving decided to plow into the old man, or the pack of small children playing when it has a no win situation, is purely a result of the underlying model.... Nothing more.
The UBER murder, and yes it was a murder folks, those cars should have never been allowed on the road, could only have happened based on a few scenarios... 1) the woman who was hit was not “seen”... and this is highly unlikely that the lidar did not detect a large object well ahead of the car moving across the road slowly... and yes, it relies on lidar, a form of radar, whether the road was dark or not is irrelevant.... or 2 and far more likely, the predictive analytics the car was utilizing decided not likely that this thing moving across the road is something I need to worry about, keep going.
I know syfy always shows the AI turning against humans out of intellectual conclusion we are inferior or whatnot... The real danger I don’t think is that some mass AI juggernaut will declare war on humans... But there is little doubt that a lot of people can wind up dead because the criteria those predictive analytics are using are incomplete, or in the case of self trained, may not even know what they are.
Say for example a toxic gas must be vented, there are 2 choices, vent to an enclsed area full of 500 people, or vent to the atmosphere... venting to the atmosphere may cause no immediate threat to human life, but the gas is considered ecologically damaging, and the algorithms weighs that more important than human life?
We are being sold a false bill of goods on the state of all this, self driving cars that are truly safe are multiple decades away... IBM makes big ads bragging about Watson diagnosing diseases, but the reality is, in practice it fails once it exist the test conditions....
Nice analysis. IMO, those chasing AI have stars in their eyes. The gist is that they largely are into the "cool" aspect of AI and glamorizing the "look at what we can do" aspect of AI to the point of literally being dangerous. Case in point.
I have watched the competitions of autonomous robots with simultaneous emotions of wonder and disappointment. The REAL use and focus of these autonomous robot effort should be in relieving human effort and in commercially viable areas.
For instance, the introduction of a robotic arm that welds the frames of automobiles makes the most sense. The arm could make the same exact welds at 8am all the way through 5pm without any breaks. Whereas a person doing the same thing would make inconsistent welds, be burned, get blinded and get smoked out by toxic fumes.
What I am saying is that the effort should be made in robotics/AI to replace Mike Rowe's Dirty Jobs, not autonomous cars. FWIW.
Great post, thank you.
AI isn’t sufficiently intelligent until it can explain to its owner why its proposed course of action is right.