What’s the Matter with Boston Tech?

Wow! Can I quote you on that?

I sort-of agree. I think that someone who wants to be a career scientist should go back and get a PhD-level education. That said, the opportunity cost (a million dollars at my age, and I'm only 31) is so severe that the social signal is that society doesn't value this skill set.

You don't need a graduate school experience to target ads or be a financial quant, but you should have a PhD in physics if you're going to be doing theoretical physics for 40 years.

That said, if there were truly a "talent shortage", people wouldn't have to put their careers on pause for 5 years to get the degree, because employers would bend over backward to help people through the process. In the 1970s, BA-level research positions existed and there was usually an expectation that, after 2 years or so in one, you'd go back to grad school on the company's dime (while working full-time in the summer) and they'd help you a great deal with admissions by making sure you had opportunities to publish (on the clock). You wouldn't be making a full developer salary (in today's terms, closer to 80k than 120k) because you'd be working part-time, but you'd be living well and progressing toward a degree. That is what you see in a talent shortage. In 2014, it's pretty rare for an employer to pay for graduate school because, well, there isn't a talent shortage relative to the work that employers will actually pay for. There's a talent surplus, which is why you see PhDs filling BA-level research positions.

I like knowing there's at least one technology community where "I published three papers in NIPS/USENIX/POPL/$YOUR_FIELDS_CONFERENCE for my MSc/PhD/post-doc" actually gets more appreciation than, "I read an Android book and coded up something like Angry Birds, only more different."

I agree. I've got no problem with what you're saying here.

Excuse me, but it's currently AI/ML Summer, and quite likely to remain such for a little while longer

I hope this isn't summer, because it still seems that a lot of of "data science" is business bullshit dressed up as being more interesting than it is. Cleaning data so it can be fed into an off-the-shelf SVM package isn't very interesting. That's a 40-degree day, to use Stringer Bell's terminology. Being optimistic about the very long term, I like to think of it as an AI March (which means there are a lot of 40-degree days out there).

Yeah, but targeted ads and financial quant-games don't actually constitute real innovation or advancement of... much anything. The fact that this sort of crap pays so very much better than doing actual research that genuinely advances mankind's fundamental capability to arrange the universe to our benefit is, to me at least, a low-level itch or irritant of moral indignation that I am unfortunately condemned to experience regularly due to growing up in the neoliberal era ;-).

>There's a talent surplus, which is why you see PhDs filling BA-level research positions.

In the life sciences, perhaps. In physics and mathematics, the PhDs go to finance. In computer sciences and some engineering fields, the PhDs can get jobs, even sometimes academic ones. I've interned for at least one Silicon Valley company started by a PhD in computing who published at USENIX (HEY GUYS WE SHOULD BRUSH UP THAT PAPER FROM LAST YEAR AND RESUBMIT HALP HALP MY PUBLICATION RECORD IS TOO SHORT!).

>I hope this isn't summer, because it still seems that a lot of of "data science" is business bullshit dressed up as being more interesting than it is. Cleaning data so it can be fed into an off-the-shelf SVM package isn't very interesting. That's a 40-degree day, to use Stringer Bell's terminology. Being optimistic about the very long term, I like to think of it as an AI March (which means there are a lot of 40-degree days out there).

I think you're looking at industry, while I'm looking at research (where I enjoy living). Neural Turing Machines (paper published out of DeepMind), for instance, can learn a general, arbitrary function on arbitrary-size data structures from input-output examples. That is a phenomenal advance in learning: generalized induction of functions, up to complexity limits. And then there's some of the stuff that gets published at, as I alluded to, NIPS.

I agree 100%. Our society has fucked-up priorities. Everything is based on the next quarter, not the next 50 years. Most of us are employed on semi-charismatic executives' private-sector political campaigns (sorry, I mean "careers") rather than advancing the state of the world.

I think you're looking at industry, while I'm looking at research (where I enjoy living).

Fair. But the low pay in research and extreme difficulty of getting in once you're out (PhD admissions get hard after 30; returning to academia if you leave after a PhD is nearly impossible) are indicators that our society's disinvestment in academia is continuing unabated.

Expanding industries take chances on people and train people up. Besieged and contracting industries (like academia) develop the pedigree bigotry and petty politicking.

Neural Turing Machines (paper published out of DeepMind), for instance, can learn a general, arbitrary function on arbitrary-size data structures from input-output examples. That is a phenomenal advance in learning: generalized induction of functions, up to complexity limits.

Yeah, there's a lot of great work that's being done. AI "winter" doesn't mean otherwise, and nor does this mediocre spring that (one hopes) is the precursor to something better. As I said, the current state is an acceptable March but, if this is summer, that's depressing.

It's not a lack of great work being done that's the problem. It's the fact that there's so little funding for it, and consequently so very few people are doing anything stimulating or important. There are always a few who can do great work in spite of the system, but we shouldn't have to bet our society on that.

It wasn't a lone hero that put people on the moon; it's was a large set of people working together (who weren't worrying about whether they'd be employable after 40, because they were taken care of). I don't think you can form that kind of an all-around strong team with academia and industry both in their current, degraded, resource-starved state.

Right now, you need top-0.25% talent and to sacrifice your life to get an academic or basic-research position. It'd be an improvement to get to a state where one of the two sufficed.