Hacker News Reader

Comment by ClintEhrlich | Hacker News Reader

I found some of the almost mournful passages in this piece off-putting. As the author ultimately recognizes, the drift of robotics talent from publicly funded universities to private enterprises is an indication that the technology has advanced to the point of commercial viability.

In other words: Companies finally think they can build robots that are good enough that people will pay for them! To me, that is exciting news, not an occasion to lament the sudden emptiness of university laboratories.

Perhaps in other fields, there is a strong tension between the basic research that government funds and the commercial applications that private companies pursue. But in robotics there are usually foreseeable uses for the technologies that researchers are pursuing. I don't foresee the direction of the field being dramatically changed, particularly given the amount of government money that will still pour in through the DoD.

View on HN

Uber and other companies (including Google) have been hiring a lot of robotics faculty as well as newly-minted PhDs. This is a good thing for robotics research.

There's generally a shortage of tenure-track jobs relative to the number of qualified PhDs, so when companies poach faculty, it improves everyone's career options. This is one of the reasons why CS PhDs have a much higher chance of getting good tenure-track faculty jobs than, say, physics PhDs.

Of course, when there's a faculty exodus it causes a temporary leadership vacuum--e.g. this has been the situation at Stanford for the last couple years, where most of the AI faculty has left to start companies. But most of the students at Stanford seem to do fine, and universities can usually fill the void in a few years and hire some new enthusiastic faculty.

Yes, the editorialist here seems to be bemoaning the 'professionalization' of the research topic, as if the companies and researchers owe some sort of loyalty to a university which profits off the cheap labour. It is roughly equivalent to chastizing NFL and NBA teams for taking all of the NCAA's best players, and paying them more.

More, in this case, meaning "at all."

Nah, the CMU Robotics Institute folks they're talking about were almost certainly paid.

Not by Silicon Valley standards, but they did receive actual money. These were faculty and employees, not students.

I was referring to the NCAA athletes :)

> Not by Silicon Valley standards

Ph.D. students don't get "poached", they "graduate or drop out" and then "get jobs". The people being poahced are professors and scientists in permanent positions.

I think people tend to under-estimate a full professor's salary at a top CS deapartment.

The average salary of an associate professor at CMU is $138,000, full professor $194,000. Before consulting, money from research spinoff startups, etc.

Even not adjusting for cost of living, I'm not sure how that's "not by Silicon Valley standards".

And adjusted for cost of living, Pittsburgh is in one of the cheapest cities in the US (e.g., it's not uncommon for CMU grad students to buy houses in reasonable areas of the city on their $30-40k/yr PhD stipends).

I'm not sure about non-faculty research scientists and so on, but I would imagine the $50-$100 range is about right. Which, again, adjusted for cost of living, is right there with SV.

Edit: E.g., according to the obviously-take-with-a-grain-of-salt CNN COL calculator, the average CMU professor would have to make $324,000 to have a comparable salary in SF. So I guess "not by SV standards" is probably actually accurate, just probably not in the way you meant

Hm. I know the guy who ran the Master of Software Engineering department at CMU West (in Mountain View.) And he didn't make the "average full professor" amount you're claiming. And one would assume he'd be making at least average for a professor.

That would be crazy money in Pittsburgh. Are you sure that figure is accurate?

> Are you sure that figure is accurate?

1. I don't have first-hand information about CMU, I got the data from here:

http://www.american-school-search.com/faculty/carnegie-mello...

Other data indicate that the six figure range is definitely accurate for CS associate and full professorships at research universities:

https://www.higheredjobs.com/salary/salaryDisplay.cfm?Survey...

And it's not insane to assume that the top schools pay 30-50k more.

2. It is an average. Lots of compounding factors: Was the person running the program a Full Professor? (Most professors are not full professors, and a lot of faculty aren't professors. Note the 50k+ discrepancy between assistant and full professors.) Did he have a research lab and how does CMU factor grants into compensation? Etc. I don't know the specific details about that particular person (or even CMU), but I also don't have any reason to doubt the numbers given above.

I'm pretty sure the guy running the department was a full professor, yeah.

They don't say anything about how they get their data, which is a little odd.

Yeah, the numbers on the second page seem more realistic to me. They give more like $95k for a full professor -- which, after Silicon Valley adjustment and the extra $30-$50k you mention is more like $130k-$175k for a full professor. That's much closer to what I've seen than $195k.

So I think the first site is just wrong in how they're estimating, at least for CMU.

> So I think the first site is just wrong in how they're estimating, at least for CMU.

Those numbers are FOR CMU, and they are the only numbers we have for CMU. So I'm not really sure how this gets resolved.

> They give more like $95k for a full professor

No, they don't. They say $111,262 (see CS, not overall). Average. (And that's just salary -- a CS professor at a top school has plenty of built-in options for making money on the side.) So you're are assuming that one of the top CS schools in the country pays 15k below national avage.

Your numbers for PhD stipends are a bit high. I wouldn't be surprised if SCS students were paid that much, but the typical stipend for the rest of us is more like 24k/year.

Even so, there's a PhD student in my program who bought a house last year, and I thought seriously about it myself in my second year. Housing truly is cheap in Pittsburgh.

> I wouldn't be surprised if SCS students were paid that much

24 sounds like you're excluding the summer months?

I know in Math and other fields it's pretty common for "no guaranteed summer funding" to be actually meaningful words. But Most CS PhD students at top schools have no problem pulling a summer stipend, making 30 common. I never knew anyone in grad school who had trouble getting summer funding if they wanted it.

And usually CS grad students opt for at least a couple internships; even low-paying CS grad student internships can have 2x+ compensation compared with the phd stipend.

In any case, if a CS grad student at CMU/Stanford/MIT/Berkely is making less than 30k, it's purely by choice or really bad planning/luck.

24 includes summer, and that's pretty much standard for engineering programs outside of CS, even at Stanford, CMU, MIT and the like. I'm in the materials science program.

> It is roughly equivalent to chastizing NFL and NBA teams for taking all of the NCAA's best players, and paying them more.

That's characterizing industry as the more advanced form of academia (NFL/NBA are more advanced than NCAA). But that's not really true. Even at the companies with the most well-funded research teams, the actual kinds of research being done in industry are often different from the sorts of research those same individuals would be doing in academia.

And while it doesn't make the headlines as often (for obvious reasons), it's hardly unusual for people in industry to return to research in academic institutions.

We don't need to think of academia and industry as zero-sum, no, but we also don't need to think of academia as merely a stepping stone to industry, even if that's how it is used by some people.

> That's characterizing industry as the more advanced form of academia

It's not a perfect analogy, but I think he was mostly trying to hit on the amateur/professional split.

> It's not a perfect analogy, but I think he was mostly trying to hit on the amateur/professional split.

Right, and I'm saying that it's inaccurate to claim that either academia or industry research is 'amateur'. They're both professional, just with different funding sources and different goals.

That analogy breaks down a little - one of the reasons to lament this is because public universities publish their research for others to consume and build upon. Research within private companies stays inside those companies.

But in that sense you might as well lament when private companies hire recent grads too, since it's competing with the talent pool for academia.

I could also make similar arguments for the NCAA, as it builds the skills of many new and young players (far more than will ever play in the professional leagues), providing an indispensable resource for the development of talent and new techniques. In addition the NCAA teams are a valuable part of the colleges they are in.

In truth, I am completely ambivalent as to the value of NCAA; but I do not believe that professors and students have any obligation to be 'loyal' to university research labs, as the universities pay the researchers as little as possible, and would be more than happy to sell the researchers down the river for a small increase in their endowment or government grants.

Companies publish plenty of research. And unlike universities, they actually make products people can use, which is even better.

Yes, they do publish plenty of research, and thanks to the Bayh-Doyle act they make a lot of money off publicically funded entities too. Those great discoveries are patented, and those licensing fees are making products expensive.

I'm not bashing the act; government wasn't getting the job done.

I would rather see something like a prize system in tech, and medical discoveries. For instance, we need a cure for a certain disease. The government could offer, say a 1 billion dollar prize, to the first company that solves the problem.

The discovery would then be required to open sourced to U.S. companies? Hell, it might prevent some companies from relocating overseas to avoid taxes? So instead of outrageous drug prices, generic drug companies would bring the price down due to competition?

(a little off topic, but relevant? Maybe not relevant towards gadgets, but relative to expensive medical cures?)

http://www.ucop.edu/ott/faculty/bayh.html

The government could offer, say a 1 billion dollar prize, to the first company that solves the problem.

There are a number of challenges with the "prize" system. Who judges whether or not an invention qualifies for the prize? Things aren't black and white in biotech. If you're drug cures a disease but has bad side effects, does that deserve a prize? What if you don't cure the disease, but you massively reduce the burden of the disease. Do you deserve a prize?

Second, the prizes would have to be much larger than $1B. Gilead (who sells the hepatitis C cure Sovaldi and Harvoni) sold $5B worth of the drugs in Q2 of this year. Their margins are likely >50%. Even relatively small drug have NPVs of several billion.

The other issue with the prize system is it makes the 2nd to market impossible. It's a disincentive to competition. Under the old system you could have two, similar drugs sold and they battle it out for market share. Even if you split the market it can be profitable. Under the prize system, if you're 2nd to market, well, tough, you get absolutely nothing. A lot of drug development is incremental. Lipitor was the 5th statin to market (and the best out of all of them). With the prize system you'd get one drug and that's it.

> unlike universities, they actually make products people can use

Universities also produce lots of things that people actually use directly in the form of Free or Open Source Software. BSD, GNU, Mach, etc. Many of the architects of the Internet were working at universities too. It's not fair to dismiss them as you did.

And some companies publish plenty of research. Microsoft is very impressive. Google good but less. Apple not so much.

No, that's wrong. The same amount of research can still be done in public universities if the talent pool is increased. You now simply have a lot more people working on the problem. Have a dozen companies with a staff of a few hundred researchers creating more knowledge, even if it's temporarily private, is a huge win. Patents expire and other companies are still capable of figuring out the competition's solution.

Basically, the entire premise is stupid. We are adding a few thousand highly skilled people into the workforce. There's no way that's a problem.

Actually, the universities will get more funding from now-wealthy alumni who made their money in industry.

You can't honestly think that the department will continue doing the same caliber of research immediately after losing 50 of its members. It will take years to regain the amount of know-how that just vacated the premises.

In the meantime that team just got a huge budget increase to continue their research.

CMU is a great university. I image there are 2-3 PhD students at the top 50 school who will be looking to work there. CMU might even be able to lure away a couple other professors from the top 25 who'd love the chance to rebuild the program.

Public companies are not a closed loop. The iPhone pushed cellphone UI rather far compared what academics did over the last 20 years.

On the other hand, it also meant that we're still stuck with capacitive touch screens.

If you think that's bad you should have seen the resistive ones.

I'd take the resistive screen of my old N900 over my current N4's capacitive any day.

Which were invented by government and university labs.

>There's generally a shortage of tenure-track jobs relative to the number of qualified PhDs, so when companies poach faculty, it improves everyone's career options.

Good for the PhDs; bad for the schools, who must now find more funding to retain the same level of talent, thus making future academic research (e.g. the next gen of robots) more challenging to accomplish

Maybe they can divert money from their athletic departments to pay their PhDs $20k more per year?

This is not an argument that applies much to, say, the mighty Carnegie Mellon Tartans.

Even at schools with actual athletics programs and budgets, most PhD students are paid with grant money won by their advisors, or through government fellowships. Not by the university. (I'm a PhD student at CMU paid with NSF grant money.)

Also of note, CMU just introduced a new presidential fellowship to fund undergraduate and graduate students in all fields [1].

1: http://www.cmu.edu/news/stories/archives/2015/september/pres...

You are assuming that there is a talent shortage and that universities know how to rank the talent. Not really true, considering how many PhD seats there are and how few faculty seats there are.

I have to agree with the more 'mournful' passages from a general perspective. I agree, it's promising to see the hard work of researchers finally enter the market, but the comments the author is making addresses a deeper issue.

Funding for basic research has dropped considerably over the decades across North America, yet the long-term value has shown itself to be unimaginably vast. It's incredibly short-sighted for a government to primarily invest in research that has direct industrial applications. The returns on basic research on both a financial and intellectual level, as a whole, are massive. We need to keep investing heavily in researchers that are addressing fundamental topics that some might otherwise see purely as 'academic'.

People have been building robts for years, they're just becoming more consumer-facing. What's not good is that a few companies buy up all the talent and control the market in a monopoly or oligopoly-style situation. Whatever goes on in comercial labs will be heavily patented; all large tech companies play that game (out of necessity until such time as there is meaningful patent reform) and Uber has applied for patents on things like their surge pricing mechanism, even though the concept of equilibrium pricing is a matter of basic economics.

Your main problem seems to be with the use of patents to protect market dominance. I agree with you that this is a problem, but disagree on the cause. The patent system has been causing this kind of issue ever since its creation; look at when the Wright brothers patented aircraft roll control, and a lawyer patented cars.[1][2]

[1] https://en.wikipedia.org/wiki/Wright_brothers_patent_war

[2] https://en.wikipedia.org/wiki/George_B._Selden#The_Selden_pa...

That is not the case, and I apologize for the lack of clarity. I am not totally opposed to patents, and I think your examples undermine your argument since they describe innovations that are clearly patentable (whereas my mention of Uber applying for a patent on surge pricing is highly questionable under the Prior Art standard).

My main problem is the use of capital to establish market dominance by indirectly buying out the public commons. Taxpayers arguably subsidized the development of Cornell's robotics department, with the benefits of their research being available across the whole economy (possibly through patent licensing), but by hiring away all the staff at once Uber has indirectly appropriated a chunk of that public subsidy.

Now all the individual scientists who worked there are independent agents who presumably liked the terms of the offers they received and also wanted to continue working with each other. One could say that the human capital in the lab was undervalued by the public funding option, but one could equally argue that Uber has helped itself to a large consumer surplus (qua hoarding the supply of available robotics scientists) with the joint effect of denying their availability to competitors, and increasing the price the public will have to pay to re-establish a comepetitive department, if it can do so at all.

There is no guarantee that this will result in an increase in robotics innovation from the private sector. It might suit Uber equally well to pay the robotics scientists to sit around and do nothing or otherwise fritter away the investment, as long as the scientists' absence from the marketplace of ideas disadvantages Uber's competitors. In other words, you don't need your own robot car, you just need to make it harder for anyone else to build robot cars that could compete with you.

I do not want to suggest that Uber has evil motives here, by the way - I have no idea what the firm's plans or goals are. I'm just pointing out that acquisition of a resource does not automatically lead to greater economic efficiency. We've all seen examples of ;arge firms buying up potentially disruptive competitors in order to bury them.

This is great, as it frees up public research funds to be invested in other perhaps more forward-in-time projects.

Reply on news.ycombinator.com