Saturday, September 30, 2017

Artificial Inanity

Long time readers know that I've been skeptical about Artificial Intelligence in general (and very skeptical about the specialized AI in self-driving cars) for a long time.

The Silicon Graybeard did the homework that I've been too lazy to do.  He has an outstanding post about why HAL won't be refusing to open the pod bay doors.  It's even covered with awesome sauce (I love me some Gartner Hype Cycle charts).

Will AI continue to give impressive demonstrations?  Sure.  But here's the key quote:
One source claims Deep Blue cost IBM $100 million. 
When are those algorithms doing [sic] to genuinely add $1 billion to IBM’s bottom line? Building still more specialized computers to beat humans at Jeopardy or Go is just creating more demos that solve no useful problems and do so in ways that humans don’t.
Yup.  Until AI produces results that return 100x the investment of building the AI, it's a nothingburger.


SiGraybeard said...

Thanks for the kind words.

My wife tells me that when she was three, her dolly was real to her. She kept it warm, put it to bed. All that stuff. Then something happened and the dolly was decapitated. Instead of being traumatized, he did more of a Spock-like "Fascinating. It's manufactured. Look at how the hair is knotted off inside. The eyes that blink are just little glass marbles. It's kinda like my Bic pen as a manufactured thing" (and this was years before Star Trek was on TV).

As a three year old.

People treat AI like she treated her dolly, not like a manufactured thing. To push the Star Trek analogy, people treat Commander Data like a human officer, not a man-made object. They're less objective than this three year old girl.

Michael said...

Watson and the work it is doing with oncology is probably a better model to look into than Deep Blue. Its still in a learning phase for sure and its only as good as its teachers can be, but its entirely possible that data analytics will become very important in the medical industry in the next decade.