Profile picture
Simon DeDeo @SimonDeDeo
, 31 tweets, 5 min read Read on Twitter
OK: following, my thoughts on Google Brain and similar institutions. Disclaimer: this is my opinion, as someone who doesn't need their funding, and has built a career without needing to flatter them. (An anti-disclaimer, if you will.)
I grew up in the Bayesian era—I watched @DavidSpergel and his band of merry scientists change our view of the world with a few simple, theoretically-motivated equations.
That's what I brought to the table when I went out to study living and thinking systems. Around 2010, of course, the deep learning revolution became impossible to ignore.
It was exciting stuff. We'd have people visit the Institute and tell us about decision trees, random forests,—all sorts of wonderful things. I tried to get a handle on it but (honestly) there was so much we could do with simple tools that it was never a priority.
When I got to IU, I was hired as a prof in the informatics department, @IUSoICE (informatics==the future of CS—i.e., forget quicksort, let's work out what these machines are doing to human life). I was on a hiring committee, and we were keen to get a deep learning hire.
I took all the candidates out to breakfast (I was a naughty hire, and skipped meetings and committees to spend my time with research students and undergrads, this was the one gap I had).
I tried to work out what deep learning was about. Most of the candidates were too sleep deprived to dissemble. Basic answer: every sexy project we do—flying quadcopters, getting another 0.1% on the MNIST—is basically one graduate student.
You work out the topology of the neural net. Then you find the weights. How? The answer: "graduate student descent", a little pun to giggle over floppy croissants at the student cafe—in short, there's no good answer, a human being sits there and twiddles things about.
Machine learning is an amazing accomplishment of engineering. But it's not science. Not even close. It's just 1990, scaled up. It has given us *literally* no more insight than we had twenty years ago.
"Deep learning implements the renormalization group!" Yeah, I heard that too. If you have a system where information is organized spatially, is it really a surprise that the neurons group information together spatially?
I'd get invited to meetings at Google Research, or wherever. They had security like crazy—worse than a hedge fund. A security guard would follow you to the bathroom.
Each scientist at my "grade"—i.e., the equivalent of a junior faculty member, someone who should be out on the edge of knowledge—was, instead, managing a team of ten people doing graduate student descent.
Google can beat University of Kansas for the sole reason that they can hire ten times more graduate students per researcher. The difference, of course, is that a graduate student at UK has the chance to do something intellectually significant. Not true at GR.
They had no idea what they were doing. They had the manpower (word chosen advisedly) to apply deep learning to anything, simulating the Schrodinger equation, drug design, anything. Their main goal was to find the scientific field they could have the maximum impact on.
I've visited probably fifty Universities. I love it. Everywhere I go, I get new ideas. It's one of the best features of my job. There's one exception: commercial "research" labs.
If you want to build machines that monitor people and sell them more ads faster, go for it. If you want to find problem where you can take a working-class job, model the man or woman who does it, and build a net to put them out of a job without compensation, be my guest.
Have we done science with something Google Research has built? Absolutely. We have a great paper coming out where we use word2vec to help build a theory of puzzle solving.
But we could have built a system of equal utility ourselves. There's zero intellectual contribution there. I'm not joking, and I'll go head-to-head with anyone who says I am.
I got a nice cold-call from a top-flight Masters' student in CS, as I do sometimes (please keep them coming, I can pay). I flew him out and we started working on a problem in the emergence of social cooperation. He wanted to do DL.
Two weeks in we were a step beyond what Google Brain was doing. I don't mean technically—they had amazing YouTube videos of sprites in a landscape. I *do* mean intellectually. Their demos were like 2018 meets something out of the 1980s.
They said they did social science, but it was nothing of the sort. It was homo economicus spread out over 50 GPUs. At best, a devastating proof-by-example of need for academia. Buy a copy of Bowles and Ginits, Cooperative Species, and you'll learn more than they did, in a week.
Can you do cool research at Google Brain? Honest answer: no. You will be on the cutting edge of machine learning, yes—an engineering discipline whose basic goals are set by large corporations. But you will not be a scientist.
I get that you may need to make money. You can make a lot there, and all the jobs at Renaissance Technology are taken. Go for it—you have all my respect. Academia sucks.
But if you want, at some point in your flourishing career, with your mind and your soul, to join the two-thousand year old parade of intellectual progress, you are not going to do it at Google. Certainly not at Facebook.
If you want to do that, I have a suggestion. It's not the only path, by any means, and I've had amazing fellow-travellers who haven't. But here it is.
Go to graduate school. Do a PhD. With us, here at CMU/SDS, if you like—but we're not the only place that does computational social or cognitive science. You won't get paid much, but you will mentors who legitimately care about the development of your mind.
It's difficult to overestimate the difference between a good PhD program and industry. It is literally shameful, if you're a good PhD advisor, to interfere with the intellectual development of a student. At Google, it's a business plan.
None of this is a joke. This is ten years of experience. Graduate school applications are coming up in the Fall. Think about it. Make sure you're getting a good deal (you shouldn't go into debt for a PhD, and you should get healthcare).
In short: corporate "research" is a business proposition. Whatever true intellectual progress comes out of there happens in spite of management. Given how good these companies are at monitoring their employees, that gap is now miniscule.
Last anecdote, then I'm done. We visited Google Research, arranged by a contact. The people were unbelievably smart. We brainstormed all sorts of wonderful things to work on. The last day of the meeting, the academics were like, OK! Let's go to the pub! Let's hash this out!
Their response: this was vacation for us. We're behind on our real work. We have to work this weekend. (Not "we feel guilty", but "we have to".) For the academics in the room, this *was* work. Suddenly, I realized that this was vacation for them.
Missing some Tweet in this thread?
You can try to force a refresh.

Like this thread? Get email updates or save it to PDF!

Subscribe to Simon DeDeo
Profile picture

Get real-time email alerts when new unrolls are available from this author!

This content may be removed anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member and get exclusive features!

Premium member ($3.00/month or $30.00/year)

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!