Iris.ai

Iris AI AS, Jun 23, 2017
Commentary by Stephen Downes
files/images/iris.jpg

The idea behind Iris.ai is that you provide it with the URL of a research paper and it provides you with a set or related papers (from a database of 60 million open access papers)  organized by category. For example, I gave it New Models of Open and Distributed Learning and got 259 related papers grouped by concept.Some of them didn't relate directly. Now, given that it didn't ask me to log in, I assume you will get the same results I did. Which raises the question: does the content of a research paper (or anything, actually) determine the best set of associated resources? Probably not. I think that the easy AI question is to associate things based on their properties. That's where we get algorithms like Nearest Neighbor (NN) and the like. But the hard AI question is to associate things based on the already existing set of associations (for example, the fact that I've already read such-and-such a paper, or the fact that George cited it in a paper he wrote in 2014, etc).

Views: 0 today, 519 total (since January 1, 2017).[Direct Link]