Hacker News Reader

I think Unsupervised Learning is fundamentally connected with the mission/goal of the AI. Whatever the mission is give to AI, it need to start learning information landscape by itself and make classification and use that knowledge to make predictions and act accordingly to optimized the outcome of the mission its given.

View on HN

But you can't even 'give an AI a mission' without a model of the world! And, if you expect the AI to achieve its mission in a way that's acceptable to you, you better be supplying it with a big enough model for it to reasonably be able to hit that tiny target in the space of possibilities.

Take self-driving cars as an example. Do you not think that any AI that will be able to do this will be essentially 'given' a huge set of knowledge before it's expected to learn anything by itself?

But, when you revise the model based an analyzing the predictions of the previous run and adapt to the positive result, isn't that supervised learning with back-propagation?

Reply on news.ycombinator.com