Machine Learning, AI and the Internet of things

Was browsing through a forbes article about how IOT will bring about revolutionary AI. Here are my 2 cents

The article does make some good points that by becoming more and more omnipresent IOT is providing more data points. With data science being the magic buzzword and learning behaviors becoming evermore omnipresent, we can indeed see more "smart appliances" which could, in theory in a connected grid coordinate more... but this is not really specific to IOT. However, if this means an all omniscient "electric dreams" type AI, with HAL at the helm... not really.

First learning systems are relatively limited and though they can do some pretty neat tricks, with some really useful... Even in the case of unsupervised learning, they can only give great results in the fields they were programmed for. And even then, it is important to validate results, make sure that they don't just generate some beautiful hype-worthy massive dataset that are basically useless.

Furthermore, I dont expect - at least in the short term - the actual devices themselves to be very smart. First, embedded systems don't really have that much power. Second, people have expectations of the objects they use... the reason I have a backpack is to carry things, not have a dialog about what shoul go in there. The primary reason to many devices we use is practical. Finally, there is cost and maintainability... those two factors make a great case for simplification of devices

Yes, data science is awesome, yes machine learning allows to make awesome discoveries and with internet of things around the corner some really cool applications could emerge but in the end, people should expect evolutionary features, nothing revolutionary... especially in the short term