>>11553427I'm a relative beginner in the AI research field, but we are so far off of a general intelligence that, at this point, it's not worth considering.
What machine learning experts are stuck on nowadays is developing an algorithm that can recognize objects in varied and changing environments. This is really important for self-driving cars, image processing, and other ventures. And it would be even better if this model could do so without millions of labelled training data examples, as is typically conventionally necessary for building state of the art models.
As of yet, there's no capacity for logical reasoning or strategy involved on the computer's part. We have no idea how to do that. In fact, neuroscientists have a much better understanding of that than AI researchers do. Our best performing image recognition models have roughly the visual reasoning capabilities of a rat. And to produce them cost a whole lot more.
Suppose that in a hundred, two hundred years we have killer robots on the loose. Oh well. At least global warming didn't get us first.
