Discussion about this post

User's avatar
Alistair Penbroke's avatar

Interesting analysis but I don't quite agree with it.

First problem - the decline of corporate R&D is over-stated. These graphs are showing relative shares, but all that means is that governments flooded universities with money which then spent it on expanding their quantity of output. Quality, however, is often missing. I've been in the tech industry for a long time and seen a lot of R&D happen, but the reliance on academic research is pretty minimal even in the AI space. Part of my job involved reading academic papers for a few years, but I eventually gave up because the ROI was zero. Lots of papers that made enticing sounding claims but when examined carefully had caveats that made them useless.

Second problem - the distinction between research and development. Very important for academics, not important at all for companies. When I think about the successful big tech projects I've seen, there was no clear line delineating the two. Experiences and results of development informed research, and research was often the same thing as development, just on more speculative features. The same people would do work that smelled like research for six months, then development of those research ideas into a product launch for another six months, then back again, with no change in job title or function. I myself have done such projects in the past.

Thirdly - the assumption that "papers published at academic conferences" is the same thing as output. Very few people in the corporate world care about publishing papers. It's just a distraction. Unless they came from academia and publishing has become a part of their identity, or they want to keep the door opening to returning, papers just aren't going to happen. The only peer reviews you care about are actual reviews by your actual peers, i.e. managers, because that's what determines if your R&D turns into promotions or bonuses. Google AI research is really an odd duck in this respect in that they seem to care a lot about getting published but that's really rare.

Obviously if you just plot by papers published then universities will look very productive and it'll seem like there's a pipeline or supply chain or something here because that's what universities are paid to do and what they optimize for. If you've actually been in real companies doing cutting edge work with tech, like I have, then it all looks like a shell game over the definitions of arbitrary words. Companies invest a ton into "R&D", just not in a way you can neatly sum and plot on a graph. Often they struggle to even measure it for their own purposes!

Finally - a nationalized energy company that can fund research? I know this is a left wing academic blog but come on. That would wreck the energy markets and deliver absolutely nothing in the way of usable research output. The west already dumps way too much money into public/academic energy R&D labs, as well as VCs pumping up private firms, and then Google also funded lots of energy R&D in the past too (see here: https://www.google.org/pdfs/google_brayton_summary.pdf ). There's very little to show for it because doing proper R&D that matters requires experience in the field, so it all gets done by oil companies, wind turbine manufacturers etc.

Expand full comment
DougAz's avatar

Excellent but some observations.

The transistor is not basic research like Cosmology. It was based on basic research done a half century earlier by Bohr, Plank and Einstein.

Some misses. The greatest industrial research center in America for 200 years was the Dupont Experimental station. And they very much capitalized!! Nylon. Gunpowder. Lycra. Synthetic fibers like polyester. As well as Agriculture.

Sarnoff labs... TV and Radio development but not the basic research... Marconi et al

GE Schenectady- plastics basic research for wire insulation led to polycarbonate and others. Electrical..well developments not really basic

Still Bell Labs was the premiere research organization of the world.

Bell labs invented fiber optics. In the late 50s. Used as a super top secret sensor deep underwater for detection of submarines. The laser also needed in that scheme they coinvented. But the basic research for both was based on basic optics of Newton and Maxwell, and Einstein.

Expand full comment
98 more comments...

No posts