In this week’s episode, Kyle Polich interviews Pedro Domingos about his book, The Master Algorithm: How the quest for the ultimate learning machine will remake our world. In the book, Domingos describes what machine learning is doing for humanity, how it works and what it could do in the future. He also hints at the possibility of an ultimate learning algorithm that is capable of deriving all knowledge — past, present, and future - and doing anything we want, before we even ask.
Enter to win a copy of the book. Details here.
Domingos mentions that there is much to be optimistic about if there were a master algorithm: predicting less poverty, greater happiness, better relationships, more humane war and humans still in charge. But what if the master algorithm exists, but something about it is quite intractable? For example, it's either inefficient, or finding it is worse than NP.
During the interview, Kyle mentions the phrase "fuzzies and neats.” But what he means to say is "scruffies and neats,” which refer to two different schools of thought in the AI community. Although Kyle might seem dismissive of the idea at first, Pedro points out that these two types of AI research are actually useful labels. Hence, for those unfamiliar with the definition: "neat" is someone that wants to approach AI in a rather rigorous, exacting mathematical way. A neat would want to derive all their ideas from first principles and prove optimality criterion, establish lower bounds and things like that. It’s unclear as to what it would mean to "solve AI." However, with a "neat-solved AI,” one would most likely be able to write it out as a beautiful equation. A "scruffy," on the other hand, isn't particularly interested in such precise theory. Perhaps this is because they have more of a "hackers" mentality of trying to just take things apart, fiddle with them and inspect the results for something novel rather than systematically designing something to complete the task. Also, “scruffier” may believe that intelligence is too complex a system to be described with the elegance desired by “neats.”
We’d like to remind listeners about how you can win a free copy of The Master Algorithm. During this week’s episode, we’ve announced that we’re doing a little experiment at Data Skeptic. We need you to help us build our training data set, so we're asking listeners to send us their resumes. If you're a person deeply concerned about privacy that's fine too— you can leave out whatever information you consider private, such as your name or email with some anonymous entry. However, we’d appreciate you sharing your real data in terms of education experience and so on. So if you're willing to share that info, please visit dataskeptic.com/ma. There, you can upload your resumes in PDF formats. We're trying to make this all work out very nicely in an automated fashion, so for now, we’re accepting only PDF formats.
Everyone who submits their resume to dataskeptic.com/ma is going to get access to a report of exploratory data analysis we do. In other words, we’ll be sending some interesting stuff to help them navigate the data science job market. However, this would all depend on our sample size of course. So please help us out here and send your resume. We've got some other fun things planned that you can opt out of. Most importantly if you submit your resume by the end of March 2018, you’ll be considered in our random drawing of two winners for a free copy of The Master Algorithm.
You can also find a copy of Pedro's book on Amazon.