Learning Machines and the Future of Academics | Hacker News Reader

Learning Machines and the Future of Academics

stochasticresonance.wordpress.com

39 pointsgoodweeds4 comments

The idea of an 'open-source' replacement academia is compelling. Currently academia generally has partial transparency and a closed competitiveness between research groups, where replicating results is non-trivial (at least in the comp sci research field I am in where sharing source is not standard).

The analogy of a git-hub for experiments makes sense; research is about expanding knowledge freely. In contrast, the partial disclosure and under-sharing of tricks of the trade between research groups is about the entrenched interests of particular groups.

The article makes a strong case for a system that removes the inefficiencies of the status quo, that disrupts the current university system, because that institution is no longer serving its initial purpose as well as it could be.

However, what is less clear, is what financial system would provide incentives in the academia-github system? Would the traditional advisor/advisee relationship provided in PhD programs also be substituted for something else? Would more people be doing volunteer research instead of being paid for it?

Those are great questions.

I'm not sure the answer is clear.

What I do know for sure is the current system for paid research and funding is broken on several levels.

The advisor/advisee relationship and the whole idea of PhDs is a medieval system.

There are amazing PhDs doing amazing work. Norvig, Thrun and Ng certainly are, but there are some who have made a career out of publishing recycled variations on marginal insights, while it is increasingly difficult for young PhDs with fresh ideas to even get consideration for grants.

Open vs Closed is orthogonal to paid vs volunteer. Sure there is a lot of unpaid open source work, but a ton of the oss in use was written by people who were paid.

There is also an implication of what it means to develop and claim intellectual property, which I think most people on HN would agree is also somewhat broken in our current system.

I'm not the one with the answers, but I'm definitely going to keep asking questions.

I just want to finish with a Star Trek movie quote someone said the post reminded them of:

Captain Jean-Luc Picard: The economics of the future is somewhat different. You see, money doesn't exist in the 24th century. Lily Sloane: No money? You mean, you don't get paid? Captain Jean-Luc Picard: We work to better ourselves.

Yeah, I think the financial incentives lie at the heart of it. While there are people who are just jerks, a lot of the jerkish behavior is because it's incentivized: if you're thrown into a hyper-competitive environment where you're pitted against other researchers for a limited number of tenure slots and grant funding, it doesn't encourage sharing. You could share that great data set you collected, but all you'll get in return is some fuzzy thanks, while if you keep it proprietary you might get another 2-3 journal papers out of it, which'll get you tenure+grants.

I think fixing those incentives is the key to fixing that. Alas, I see it mostly going in the other direction: universities are moving more and more towards a quantified business-type model, where professors are judged wholly on money brought in and publication metrics, and the competition is only getting fiercer. If someone wants to be open/sharing/generous with research and have a job, it's not impossible, but it's fighting against the system more than with it.

Rational agents at a Nash Equilibrium.

Reply on news.ycombinator.com