A few days ago, Nortrhwestern hosted a conference in honor of Paul Milgrom winning the Nemmers Prize. (As reported here and here.) The speakers seem to be the A-list of Economists in Mechanism Design, and their talks slides, as well as a video of Milgrom’s talk, are on the web page. Computer Scientists were not invited, but it seems that Lance Fortnow was still able to tweet from there gems like:
Preston McAfee: Left to their own devices computer scientists would recreate the Soviet Union
Susan Athey: CS still needs to catch up in auction theory (on why Microsoft hires economists)
Turing's Invisible Hand 
A few computer scientists were invited, us Northwestern folks and David Parkes, but definitely dominated by economists. To be fair, Susan did apologize for her CS comments later that day.
Well it is not about apologizing or not. That is what she thinks. There is a difference between auction theory and auction design. CS researchers are more attracted to the theory, and limitations but these things are changing.
Btw – except being a strong woman, I don’t remember what exactly Susan has contributed to auction theory in the last 5 years or so.
I am not sure if it is healthy to discuss who needs to learn what, instead one should focus on the learning needs of the today’s society.
A few years back, I wrote a manuscript called automatic economics, with atomic as a substring of this word colored differently. What I pitched is that more and more economics is going to be run by computers instead of human, and since CPU cycles are cheaper than brain cycles, things would be done at their finest granularity. I applied this viewpoint to Search problem, Music problem, and many other problems such as CO2 economics. This is indeed happening in the world. Economics is becoming automatic. I do not think I was the first person to pitch this concept, and I do not think I was the last.
What it means is that we need both CS and Economics to create futuristic economics systems. Indeed, if you try to apply these two branches separately on a problem you sometimes get results not confirmed by the practice. For an example, VCG is truth revealing but not in Search auctions when you take the computational learning aspects of CTR into account. (There are many reasons known for VCG not being truth revealing, and this computational reason is yet additional one.) So indeed, the primary benefit of VCG claimed over GSP, does not hold water if you take computation into account.
Well to get such results you need to put both CS and Economics into a single brain. If you are pure economist you may end up implementing VCG (for clarification, GSP is not proposed by economists) and lose the compputational benefit of GSP. If you are pure CS you may end up implementing the first price auction.
So we need to put both CS and Economics in some of the brains. Now these brains could either come from CS side or Economics side is secondary. So indeed if CS needs to catch up with Economics, so is the other way round too. Economists have to learn some computation too, if they want to improve their impact in the future world. In some sense, the combine field is also a branch of economics. Economics have dozens of branches, why not consider another one. Look at the name, we call it Algorithmic Game Theory, and not call it Game Theoretic Algorithms. The first one says it it is basically game-theory (a branch of economics) with Algorithm’s as an adjective. The second is saying the same thing the other way round.