The application layer of AI is a lot harder than a demo
Don’t get me wrong I think AI is real, but in order to make it happen in the real world, we have to get past the hype, which means getting past the demos.
Developers/coders tend to be pro AI, because of a belief that coders will build AI not have coding jobs taken by AI.
There is certainly some low hanging fruit at the application layer of AI, such as task oriented gig work. Who does not want a worker who does not get tired working 24:7 and never complains about the price of labor?
It is not about how much intelligence is needed. Take something like driving a car, which most adults can do, yet the goal of Full Self Driving still eludes so many smart well-funded companies. Apart from the technology, the obstacles are instructive:
- Data. The domain must have lots of easily accessible open source data and a business model that incentivizes companies to share proprietary data.
-Consumer sentiment. We are more forgiving of human mistakes than robot mistakes and regulations are based on this consumer sentiment.
Data is a relatively easily fixable problem, but changing sentiment costs a lot of time & money.
For most ventures at the application layer of AI, the big question is who will be motivated to train AI the robots to replace them. Most knowledge workers now have more cloud and have seen what happened to workers who trained offshore outsourcing companies to replace them.
Search might be the killer app of AI but that is different story as I think of search as middleware not application layer.