There are Five conundrums that AI should solve.
1 Intelligent Agent have limited resources.
As we imagine that Agent should handle questions by itself, but how many cores, memories it can use it under a setting, how to scale it, how to expand it smoothly?
That’s one of the question.
2 Computation is local, but problems have global constraints.
Local computation is basic compute unit, it can only solve one kind or specific questions. Solve one kind of questions can not solve every same kind of, each problem have it’s own feature.
3 Logic is Deductive, but lots of problems is not.
We setup logic in every unit with AI to solve problems, we think all problems have it’s own logic and can be calculated, but the truth is, lot’s of problems are adductive or inductive in nature, it’s cannot be computed.
4 The World is dynamic, but knowledge is limited.
AI started with all knowledges we already know, how to let AI agents to get the knowledges they don’t know, no experiences as a beginner to learn to address a new problem.
5 Problem solve, reasoning and learning are complex, but explanation and justification are even more complex.
With AI we get what we want to get, how it’s be down? what’s the meaning and how it’s be down with the decision and also the target is be done?
by Gavin Wang after learning from machine learning courses at 11.05.2019 in Dresden