As a practicing scientist, I always find complaints about true innovation being unfundable in the current system dramatically overstated. As Matt alludes to, we all know that while proposals need to be well grounded in preliminary data and existing ideas, they are not rigid contracts. Moreover, again as Matt realizes, for every brilliant underappreciated idea, there’s not just one, but more like five bad ideas that absolutely should not be funded. I probably have that many before lunch!
A "people, not projects” approach sounds great if you can accurately identify those people who will consistently produce great science and not go off on too many wasteful tangents. But you can already guess that all that will happen is one set of gatekeepers (other scientists serving on review committees) is replaced by another, more opaque one (unaccountable administrators). And before you know it, the “usual suspects” will be getting funded and entire groups of people will be locked out (however you want to define them).
The answer to scientists not getting great ideas funded is so incredibly simple: just fund more proposals, as Kevin also points out below. Typical funding rates are about 10-20%, but no one on a review panel will say that means 80-90% of the proposals were not worth funding. So we spend a lot of time generating ideas, refining them, cutting them up and repackaging them, and shopping them around to different agencies. And while some of that undoubtably makes the science better, a lot of it is wasted time.
One more point: I know some of the projects called out by the Golden Fleece awards. I’ll just say that while politicians getting involved at this level is indeed kind of silly, they aren’t completely wrong either.
And one can cite the Roentgen/Meitner et al cases, but doing science has gotten a *lot* more expensive. It's getting a lot harder to transform physics while working as a Swiss patent clerk.
I'm glad that CERN exists and we can make these breakthroughs in pure science, but am I sad that the Superconducting Super Collider was cancelled here? I don't know. That was a pretty hefty price tag for addressing some pretty abstract questions about fundamental physics.
In the case of the LHC, it's pretty much a boondoggle though. Higgs was expected, yes, but nothing else of note has resulted. And there's no good reason to expect new physics from accelerators that can currently be built by humans now or in the future. It's not even just the waste of money though, but worse, the waste of smart people doing experiments that will very likely have no useful output and developing theories with no basis in reality. Sure, you never know until you try, but that's not enough to justify thousands of people-years and billions of dollars. More likely, people can do much better work if forced to go in different directions than just trying bigger and bigger accelerators. For example, I would place my bets on neutrino experiments producing novel results in the next century.
It would also be nice if science funding weren’t so feast or famine. When there’s a new batch of funding everything is fine but when the funding gets cut a bunch of people leave the field permanently.
This is what keeps me awake at night - how to keep paying my people if one of our grants doesn't get renewed/replaced with something else. And it typically takes 1-2 years to get even a really good idea funded.
I agree with these points but I can’t square that it’s all a funding approach. Sure, that’s the point of the political aspect that’s being reported on.
Yet, the rat race of tenure plays a role here too. Everyone has to chase funding/pubs via what the gatekeepers want. Grind out a few absolute garbage papers and commit a few pedagogical crimes graduating PhDs. By the time you get tenure you’re so ingrained into this system that real innovation is for all intents and purposes a distant dream
I just finished the Walter Isaacson book on the development of CRISPR. I dunno, but that development, along with that of inventing mRNA vaccines, the huge improvement in renewable energy economics etc etc make me wonder if we're not doing all that badly.
Maybe instead of the "Golden Fleece" we should be thinking about the "Golden Egg," as in be careful about killing geese.
It's always easy retrospectively to point to what should have gotten more support early on; I'm not sure that's an indictment of the whole system, nor am I even sure what you're proposing. If it's more money for a select few, well, I don't know how you decide who they are and how you can be sure the ones you're locking out don't have even better ideas if only given space to develop them.
Interestingly, she cites HHMI funding but nothing from NIH for that work, although I'm sure she had funding from them as well for other stuff. But even without her, it was going to happen. Feng Zhang was publishing on it around the same time: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3795411/
He cites an NIH Director's New Innovator Award, showing that the usual funding sources also knew this was a good direction in which to go.
My point about rethinking how we fund scientific research is to acknowledge that we may not be doing too badly overall (without emphasizing any necessary NIH--> CRISPR link) and to be careful not to let the perfect be the enemy of the good.
And just to dwell on CRISPR a moment more, I suspect that absent the overall supportive environment that NIH funding creates in the field, CRISPR development would have taken a lot longer.
Ah, sorry, I misread you! Yes, I agree that overall, it's working pretty well. I point to the mRNA vaccines as a good example of this, as it's been 2-3 decades of solid single-investigator work from many labs that allowed us to accumulate sufficient knowledge to make this almost trivially simple now.
Thanks for taking on a topic close to my heart. The politics you detail reminds me of Bismark's comment about sausage making, and I try not to think about it. Speaking as a senior biomedical researcher wholly dependent on public support, I strongly disagree with those (including some you cite) who extoll ground-breaking paradigm-shifting "breakthroughs" over mundane incremental science. IMHO science IS incremental, and it is precisely for that reason that it has been so successful in advancing over the past 2 centuries.
From my own perspective, I have only two problems with NSF: first, that they don't have as much money as I would like, and second, that they are too heavily influenced by progressive shibboleths.
Just wandering in to agree with a lot of other commenters that for every single complaint about the current grant allocation system, the one and only solution is more money. There is no royal road to funding good projects. Research sucks, most of it doesn't work out, and you have no idea which projects or which scientists will or won't work out and you're delusional if you think you can do much better than what everyone is already trying their hardest to do. Hovering over some stingy bit of money and desperately trying to send it only to the "most deserving" scientists is just like hovering over a few college admissions spots and trying to admit only the most deserving students, or hovering over a few parcels of land and trying to allow only the optimal apartment buildings, or...we've been over this sort of thing a million times on this Substack. You want to fund more good projects, then fund many more projects, and eventually a few of them will be good. Stop looking for a cheat code.
On a more frivolous note, I would just like everyone to know that the NSF grant submission/administration website is the worst thing in the world. I do not understand how it can have been designed without intentional malice. Fixing that website would increase every American researcher's productivity by numerous hours a year even if you changed nothing else about the system.
That's not really true. Many projects begin with a high probability of success in that they are likely to answer the questions that have been posed. It's just that the answers are not known ahead of time.
the one depressing aspect of the marked up Endless Frontiers Act that you didn’t hit on, but that gut punched me, was that Maria Cantwell (D-WA) went to bat for Jeff Bezos and Blue Origin and added an amendment that prevented NASA from just awarding SpaceX the Lunar Lander contract for the SLS / Artemis program.
Whatever you think of Bezos or Musk, it’s hard to argue that SpaceX hasn’t wooped the entire private space industries butts when it comes to putting up or shutting up.
SpaceX formed about the same time as Blue Origin did, beat countless odds, and has put up many more times than it has been shut up…landing rockets on their tail end, reusing them up to 10x, sending humans to the space station on reusable rockets, etc. Meanwhile, Blue Origin has yet to even put a rocket in orbit.
SpaceX has taken 60% off the launch market by underbidding everyone else for launches by tens of millions of dollars. Even Boeing and Lockheed’s partnership (United Launch Alliance) has failed to compete…and now those companies are leveraging lobbying connections to basically get in on the action at the expense of NASA’s program timeline, etc.
A lot has been written about it, but the fundamental structure of companies like Boeing, Lockheed, and Blue Origin is just not capable of competing with a company structured like SpaceX. And instead of restructuring themselves and learning the lessons that SpaceX has taught everyone, people like Jeff Bezos just keep hiring the same old people to run things exactly as Lockheed and Boeing did, and he’s getting the same result.
Isn't she going to bat for Boeing, not BO, given that Boeing is based out of WA?
In any case, while Old Space's profligacy is well-documented, SpaceX will have eradicated from the market anyone who's not conducting reusable launches within ~5-10 years, so I'm not too worried about it. Sooner or later the savings will win out.
well, Blue Origin is headquartered in WA and the amendment gives just enough money for NASA to accept the next lowest big (BOs) and basically mandates that they do so…
The worst part is it mandates all sorts of things that ultimately delay the project many years and cost even more millions of dollars to do things like more testing at Stennis Space Center in MS that NASA doesn’t even say it needs to do.
They can flush a few billion more down the tubes for the next decade or so, but it doesn't really matter in the long run IMO. Dread it, run from it, SpaceX still arrives, and the rest of the industry will have to adapt or die. Still, definitely frustrating.
well, it’s the needless delays in the projects that only make them more likely to be cancelled all together that’s the real tail risk. It really isn’t about SpaceX, other than SpaceX proving that it can be done at much much lower cost, and be done better, to boot…
In my field, astronomy, I think it's rare that transformative research is blocked because of the review process for grants. And I think it's unlikely that tinkering with the review process will significantly reduce the chances of transformative work going unfunded.
In astronomy, breakthroughs are driven by technological advances in telescopes, and those advances are limited by funding. So if you want to increase the number of breakthroughs in astronomy, you should increase the funding levels for the research infrastructure (telescopes and the labs that build their instruments). Meanwhile, if you want to promote more good research in general and train more young scientists, you should increase the funding levels for the grant programs at the NSF, NASA, etc. There are plenty of good proposals that get rejected because of the limited funding available.
I found this post incredibly frustrating. Matt would make a point and I would ready my devastating response(*) in my head and then read the next paragraph and he has fully met my objection. This happened several times, to my irritation.
Thoughtful, well-reasoned posts are the bane of my existence.
I'm interested in these "secret Congress" dynamics. I live in Washington State where we just wrapped up our state legislative session. By all accounts it was pretty big year for the legislature and a number of big, and contentious things got over the finish line: a cap and trade bill; a new capital gains tax; and several police reform measures.
The session was also conducted virtually and didn't allow for the normal level of lobbying and advocacy and general daylight being cast on the things working their way through the process. It definitely wasn't a "secret" legislative session, but the relative productivity of the legislature seems like a real violation of the idea that big stuff doesn't happen without pressure, lobbying, advocacy, and lots of visibility.
I'm curious if similar dynamics played out in other state legislatures? Did others do things 100% virtually? Did this lower the visibility of what was going on, and in turn yield surprisingly substantive outcomes? Is my hunch about this completely wrong?
Half of MY’s articles argue essentially “Uncle Sam should spend more on X.” The obvious retort is “debt is bad, especially when the US just borrowed $6 trillion for the pandemic.”. I’d like to see Matt’s take on 4 issues:
1) How much inflation pressure will the newly deployed $6 trillion create? This is the biggest Keynesian experiment since World War Two, federal spending as a share of gdp approaches has approached World War One levels. There was serious inflation in 1918, why will this time be different?
2) What is the debt carrying capacity of the federal government? During the Napoleonic wars, British debt climbed to 250% of GDP and 1815 Britain was basically a subsistence economy with a few proto industrial sectors. The U.S. could probably borrow more than 250% of gdp without serious problems, but how much more?
3). How quickly should the US approach it’s debt carrying capacity? Should we leave a margin of error for a war or recession?
4) How do you convince swing voters and swing state senators that borrowing a greater percentage of gdp than 2007 Greece is a good idea?
the boring but correct answer is to "wait and see". Basically, we've been paranoid about inflation and debt for decades but things are fine so it's reasonable to spend a bunch and see what happens. If inflation starts growing, we can see it happening and start taking action. The way some people talk you would think that 4% inflation would be the end times.
I basically agree, the question is whether we should borrow 5% of gdp for 10 years or 50%of gdp in two or three years. The former approach is likely to spend money on worthier and more useful things.
I'm not sure I disagree, but we can't really plan like that. This is likely to be the only two year period of the 2020's with a Democratic trifecta in government. I think that's the basic argument for going big, especially the transformative family stuff that people will notice.
The short answer is that debt is bad *unless* the debt pays off in the long run because the stuff it funds creates economic growth that enables repayment of the debt. The same people that talk about the "national credit card" ignore the other obvious personal finance metaphor of a mortgage, which increases debt but has an obvious positive purpose.
In response to 1) "why will this time be different?" Probably because almost everything about the domestic and global economy is different than it was in 1918. I'm not smart enough to explain all the deflationary pressures that have basically killed inflation in America for the past 20 years, but there's a reason Bernanke and Yellen couldn't even touch 2% when they tried
As to 1: Not much inflation 1-5% government spending isn’t as inflationary as you would think. It is government setting of prices that causes Venezuelan inflation also being a closed economy without many important hurts, example oil supply shocks. During the Great Depression FDR devalued the dollar relative to gold and added 40% more to everyone’s bank account and there was no inflation. So Giving away 40% of M1 can cause 0% inflation every time I see that in history books I reread it. However, Running the economy hot can leave us vulnerable to a supply shock, if we have a trade war. That is what caused inflation in 1970s so that is the inflation risk. If your are going to cherry pick prices for inflation so will I. I just bout a microwave for $49 my parents bought a microwave in 1985 for $49 Costco hotdogs $1.5 same as 1986.
2 debt to GDP is situational. What are Interest rates are you a reserve currency etc. You are correct in saying waiting till the market says no more can be painful. Especially if it is other countries holding the debt. In the USA case the fed is holding the debt. the government issues the bond and the government buys its own bonds... Inflation is a much bigger threat for that reason. Based on history I, am only worried about inflation if the government starts mandating wages or we have a trade war. And no minimum wage up to 15 wouldn’t be too bad because look around most places fast food is already above $15 who even makes 7.25 anymore? In conclusion We should not worry about debt to GDP if the fed holds a big chunk of it and banks are required to hold it for a safe asset mix. I will worry about percent of debt to GDP when China and the EU own our debt. If that where the case I would worry about our current levels.
3 how quickly should we up our debt to GDP ratio on average the same as GDP growth but in a recession. well like 5-25% a year for 2-3 years max. Here is why you will eventually have economic growth to cut against the debt ratio growth during good years and our welfare state is needs based so spending automatically goes down in good times. Assuming you will have 4-6% economic growth is reasonable especially if you are dumping money into the economy. And before you say we have exceeded those percentages keep in mind the pandemic has been over 2 years. Finally, inflation combined with low interest rates helps the debt situation.
4: Swing Voters don’t really care about debt to GDP ratios they care about cancel cultural they care about abortion and gun rights. They care about the woke mob. I have 3 or 4 conservatives I listen too and that is what they go on and on about. True swing voters are like Joe Rogen he mostly doesn’t talk politics but he talks about trans in sports, cancel culture and that he doesn’t want his taxes going up. telling swing voters here is tax free stuff is popular. Also Joe wants legal weed so give swing voters that. As for swing Republican senators they are talking about the debt in bad faith you can get republicans on board with debt increasing polices but adding a tax cut to the rich that makes the debt worse.
I agree in the long term disagree in the short term. We contracted by 4% in 2020. We contracted because waiters had to get other jobs often jobs in higher productivity parts of our economy, productivity grew in 2020. Now in 2021 the entire hospitality hospitality industry will come back into existence and they will have to increase workforce participation to do so because all the waiters got slightly better jobs at Amazon. Given that 1% real annual GDP growth is reasonable or -4% 2020 +6% but only for 2021. But yea over a period of 3-5 years 6% real GDP growth will not happen baring insane policy changes.
And in response to 4) I think Matt's typical retort (correct IMO) would be: "you help them a lot and play down culture war stuff so they like Democrats"
To address #4, voters don't care about debt or the deficit and never have in the modern era (as evidenced when the GOP's "deficit hawks" blew up the deficit to cut corporate taxes) so I wouldn't bother worrying about this one at all.
A whole bunch of US science funding mechanisms were created either at the beginning of the Cold War (like NSF) or immediately after Sputnik appeared to show that the USSR was ahead in science and engineering (DARPA). It's not surprising that the China rivalry is also driving US science funding.
On grants: besides the problems described here, the big problem imo with NSF (and other!) grant-making procedures is the amount of work that is spent applying for them. Academics spend a large portion of their time applying for grants, and that in turn reduces the amount of time they spend doing the research the grants were supposed to support. Since NSF grants tend to be very competitive, this means that every grant made induces a lot of science *not* to happen, as for each scientific project funded, many scientists have been incentivised to spend a large amount of time not doing science.
(This isn't just an NSF or just an American thing, of course, it's a product of the incentives and limited perspective of grant-making bodies. But the NSF is a really notable case.)
In the context of a world where uncontroversial bills can pass through regular order like this, I don't think a watering down of ambitions like you saw with Endless Frontier is necessarily the end of the world. It might actually be a good thing! Some parts of this bill will work and others won't. If you think you'll get a second crack at this once we've learned from our experience, the rest of the money can be more appropriately targeted. I don't personally think it's a good idea to just throw $100 billion dollars at the NSF as it stands IF you believe congress can return to this issue and finish the job.
I think it's a consequence of the omnibus/reconciliation world of legislating where every law feels the need to 'solve' whatever problem it is addressing in full and in perpetuity. You see that with the ARP, which objectively has a TON of wasted and misdirected money in it. But it is defended correctly on the grounds that it is 'better to do too much than too little.' Left unsaid here is the assumption that you only get one shot at it. Obviously it would actually be much better to do too little and then a month later to do the right amount while taking into account what worked and what didn't.
If you believe in the definition of creativity as the synthesis of new ideas from disparate ones (i.e., connecting non-obvious dots) then the current funding model is designed to thwart creativity aggressively.
If scientist from Field A rubs elbows with scientists in Fields B and C and gets an interesting idea, their only hope of funding is to collaborate with people from Fields B and C. If they go through the front door, their proposals will be refereed by experts in Fields B and C, who will see the lack of familiarity with the norms of their field as a sign of amateurism and will be extremely skeptical than an idea they didn't think of, and don't understand well, is going to solve big problems in their field. (They're usually right; these are high-risk endeavors by definition.) But if the scientist from Field A tries to team up with scientists from Fields B and C, they will run into the same basic problem of trying to sell weird-looking ideas to skeptical experts, because they will have to apply in an existing call, which is always going to be based around known problems and potential solutions based on known ideas. (There is a feedback loop from program officers attending conferences and wanting to fund already-successful labs that is difficult to break through.) If they manage to get some funding, they will face the same problem all over again with editors and referees at the top journals. In the end, the disincentives to pursue truly creative research are so strong that only a few scientists at well-funded, high-profile institutions will even try.
In practice, the result is that a handful of labs define what is popular and then seed the academic landscape with their progeny, who fulfill that prophecy. Those who stick to the defined tracks, know their place and maintain good relationships will find friendly program officers and calls that were effectively written by their former advisors. They will write papers that get refereed by people who are primed to recognize that they are publishing an acceptable level of new results in their narrow sub-field, which will perpetuate the cycle. (It is hard to overstate how hard the system pushes back when you try to move out of your lane.) Much good science gets done this way, but when you under-fund science, it essentially all ends up flowing into this gatekeeper model, which is nigh incapable of producing actual fundamental breakthroughs because it rewards competition on papers, awards and profile (fame) rather than ideas.
We're essentially mining the breakthroughs of the 20th Century without seriously investing the approaches to science that created them. The diminishing returns is not because science is getting harder or more complex, it is the result of too many people are standing on the shoulders of giants rather than becoming giants themselves.
The best way to get "scientists from Field A to rub elbows with scientists in Fields B and C" is to support America's research universities. They gave us the 20th century scientific revolution, as well as the internet era and the genomic revolution.
And one way to help universities is via grant funding. But other ways matter, too, since they are big places that do more than science. (And their fostering of a hundred disciplines helps them break out of ruts.)
So: more money for research labs! But money for universities in general.
That makes sense to me. It seems like micromanaging too much of the funding through centralized federal gatekeepers has created a whole ecosystem built around gaming that system. From my limited anecdotal knowledge, it seems like the system we have now basically depends on and empower professors who are best at acting as rainmakers for their university, not necessarily those with the best ideas.
Maybe it would be better to distribute a large portion of federal research money essentially as block grants to major research universities in each state.
The sad thing is that the NSF is way better in this regard. They tend to fund people with good ideas with very little micromanaging. But most of the money flows through DoD, which is the epitome of micromanaging. It's also a good ol' boy network with opaque review processes that, ironically, is one of the better places for fundamental research funding; fundamental research in narrow topics that is done by people who can break into the network.
I once saw a chart of government funding to my university. The part that works on defense projects dwarfed the rest roughly 3:1!
It's a different way of doing science, that's for sure. Their projects practically are contracts, including with explicit milestones they need to meet to get the next increment of funding. And like you say, it can be ironically fundamental stuff (the most "out there" project I've worked on was with those folks!).
And empower university administrators instead of scientists? That would be significantly worse.
As for professors acting as "rainmakers", I'm not entirely sure what you mean. Believe it or not though, the most well-funded scientists also tend to be doing great science as well.
No doubt it varies by field, but I'm not sure that the administrative and management skills necessary to be in charge of a large lab full of rats, mice, grad students, research assistants, etc, optimized to churn out large amounts of small-ball research papers and grant proposals necessarily correlates with the skills and creativity needed to to make scientific breakthroughs. Sometimes it obviously does but what about all the times it doesn't?
No one gets grants because of their administrative skills (just ask the folks working in the big labs!). And if all they churn out is small-ball research, their proposals will eventually have trouble getting funded with the critique that its "only incremental".
Having been in big labs and having been a small lab myself, what I can say is that the big ones are significantly better positioned to make breakthroughs. This is for a number of reasons, including more cross-pollination of ideas between people, more resources to try and/or develop innovative approaches, and (to contradict my own post above a bit), more intellectual freedom, at least for the postdocs and students (who are actually the true innovators anyway).
I'm sure you know more about it than me so I won't argue! My impressions are just extrapolated from a few anecdotal perspectives from conversations with people. I wasn't suggesting, though, that people are getting grants because of administrative skills despite lacking science skills. More the other way around, whether the system is losing the benefit of absent-minded professor types who might be scientifically brilliant but lack administrative skills.
I think it's unlikely that the government, ruled by political processes and acceptance by the body politic, will ever be truly daring when it comes to funding out of the mainstream scientific research.
But if Elon Musk et al want to throw a few billion bucks at crazy ideas, I'd say go for it.
You majorly TLDR’d here dear friend. But my long morning bowel movement allowed me the time to read it.
I’d love to note your following point:
“ We're essentially mining the breakthroughs of the 20th Century without seriously investing the approaches to science that created them”
Do you have thoughts or opinions about whether or not you would apply this conclusion to the current trend of AI/ML? I would say it fits nicely. These practitioners often have no understanding of the problem they’re trying to solve nor any reading of the robust literature surrounding it. This leads to approaches to solve problems by throwing out research, and throwing in tons of data and overly accessible software packages to generate an “AI solution.” That... rarely ever bears much fruit.
I am no expert in AI/ML, but I have dabbled with it and sat through many, many talks where people are trying to apply it to various problems in physical science. I think, though, that it is another good example of mining old breakthroughs, since (AFAIK) they mostly happened in the 80's and were just waiting around for computing power to catch up.
It's also not clear to my what problem ML actually solves. It is unquestionably good at optimizing certain tasks at scale, but the best AI is usually judged by its ability to do something that a human can already do, though often less efficiently (in the literal sense that it requires more people). How is it going to break new ground upon which future breakthroughs will be built?
But isn’t the actual work there done by quantum chemistry that was developed int the 60’s? ML can make the process more efficient, but it is clueless about what it is doing without QMMC/DFT, etc. to validate the results (or evaluate fitness in a genetic algorithm).
Everything builds on something else in science, of course, but the main advances from AlphaFold really come from just how much resources (people + GPUs) they threw at the problem. Others would have gotten there eventually though. Apparently the company DeepMind lost almost $650m last year alone!
That they lost that much money is great example of how private industry is willing to invest in research if they sense the payoff is significant enough.
"Others would have gotten there eventually though" - to what does this not apply?
I don't want to come across as dismissive of ML; I think it's great. It's like the transistor, which was revolutionary and a major scientific and technological accelerant. But the underlying breakthroughs that was based on weren't so clearly built on something else in science. In fact, the understanding of electrons and solid-state materials that enabled the invention of the transistor were radical departures from existing science that challenged settled science. Those sorts of bold, non-intuitive assertions and discoveries seem to me to be less frequent, in part, because of the incentives of modern funding models.
Of course, for all I know the scientific revolution in the life sciences is producing earth-shattering, paradigm-shifting breakthroughs on a daily basis akin to the quantum revolution a century ago. I wouldn't know because the acronyms alone make that literature impenetrable to mere mortals such as myself.
I do think that applying AI and ML to a problem, while often extremely difficult, is about as low on the creativity scale is one can go. The bottleneck there is the paucity of experts; if you can plausibly apply your domain knowledge to an ML-driven project--and find people to implement it--money will rain from the sky. It doesn't even have to produce anything useful, since it is likely sufficiently novel that the implementation itself is taken as proof of success. (Plus people are afraid to ask questions, lest they reveal their ignorance of your jargon-filled presentations.)
I mostly support the amendment taking some $ from the increased NSF to increase DoE R&D funding. My experience is that DoE programs in energy efficiency and advanced manufacturing are more engineering /application focused than NSF. I think that some of the big developments will need some patient money to cross the chasm to market-ready technologies. I think that some of the DoE programs are better set up than NSF to support those pre-commercial developments.
I agree the ambitions shrinking is bad and more money for science is great! I think ARRA roughly doubled NSF funding for a year and they had no trouble finding worthy projects to fund. I however don’t think it’s a big issue that some of the money went to DOE labs, which I think are doing some of the same technical work anyways that the bill envisioned. Also, with regard to the total amount of funding, I think there is also the issue of where in the research/development chain the funding is going to. If the funding represents primarily an increase in basic research funding that is very good because that is the kind of research that is least likely to be funded by the private sector. If however the funding is split more evenly between research and development I think this is much less good. Also, just because Congress authorized a certain amount of research in this bill doesn’t mean Biden couldn’t get the federal bureaucracy to allocate some more money to research than it already does, or to set research priorities that are particularly worthwhile. He also has the power to tackle some of the science bureaucracy issues you point out, though paradigm change is hard. And I hope with the return of earmarks that Congress becomes less interested in individual spending decisions and more interested in the broader outcomes government achieves. There was a real move in the last half of the 20th century to small ball Congressional thinking, and generally much less ambitious legislating. A Senator should be criticized, not praised, for worrying about $84,000 in spending. Lastly, I’d like the government to do some more prizes where the reward is received once the discovery has been made. This could incentivize a broader set of folks to look into important problems, and it has the benefit of avoiding the grant bureaucracy you rightly criticize.
As a practicing scientist, I always find complaints about true innovation being unfundable in the current system dramatically overstated. As Matt alludes to, we all know that while proposals need to be well grounded in preliminary data and existing ideas, they are not rigid contracts. Moreover, again as Matt realizes, for every brilliant underappreciated idea, there’s not just one, but more like five bad ideas that absolutely should not be funded. I probably have that many before lunch!
A "people, not projects” approach sounds great if you can accurately identify those people who will consistently produce great science and not go off on too many wasteful tangents. But you can already guess that all that will happen is one set of gatekeepers (other scientists serving on review committees) is replaced by another, more opaque one (unaccountable administrators). And before you know it, the “usual suspects” will be getting funded and entire groups of people will be locked out (however you want to define them).
The answer to scientists not getting great ideas funded is so incredibly simple: just fund more proposals, as Kevin also points out below. Typical funding rates are about 10-20%, but no one on a review panel will say that means 80-90% of the proposals were not worth funding. So we spend a lot of time generating ideas, refining them, cutting them up and repackaging them, and shopping them around to different agencies. And while some of that undoubtably makes the science better, a lot of it is wasted time.
One more point: I know some of the projects called out by the Golden Fleece awards. I’ll just say that while politicians getting involved at this level is indeed kind of silly, they aren’t completely wrong either.
And one can cite the Roentgen/Meitner et al cases, but doing science has gotten a *lot* more expensive. It's getting a lot harder to transform physics while working as a Swiss patent clerk.
I'm glad that CERN exists and we can make these breakthroughs in pure science, but am I sad that the Superconducting Super Collider was cancelled here? I don't know. That was a pretty hefty price tag for addressing some pretty abstract questions about fundamental physics.
In the case of the LHC, it's pretty much a boondoggle though. Higgs was expected, yes, but nothing else of note has resulted. And there's no good reason to expect new physics from accelerators that can currently be built by humans now or in the future. It's not even just the waste of money though, but worse, the waste of smart people doing experiments that will very likely have no useful output and developing theories with no basis in reality. Sure, you never know until you try, but that's not enough to justify thousands of people-years and billions of dollars. More likely, people can do much better work if forced to go in different directions than just trying bigger and bigger accelerators. For example, I would place my bets on neutrino experiments producing novel results in the next century.
Sabine Hossenfelder has been on this beat for a while now: http://backreaction.blogspot.com/
There are more things in heaven and earth than are dreamt of in your [science]. :)
It would also be nice if science funding weren’t so feast or famine. When there’s a new batch of funding everything is fine but when the funding gets cut a bunch of people leave the field permanently.
This is what keeps me awake at night - how to keep paying my people if one of our grants doesn't get renewed/replaced with something else. And it typically takes 1-2 years to get even a really good idea funded.
I agree with these points but I can’t square that it’s all a funding approach. Sure, that’s the point of the political aspect that’s being reported on.
Yet, the rat race of tenure plays a role here too. Everyone has to chase funding/pubs via what the gatekeepers want. Grind out a few absolute garbage papers and commit a few pedagogical crimes graduating PhDs. By the time you get tenure you’re so ingrained into this system that real innovation is for all intents and purposes a distant dream
All I can say is that while that may be true for some, that hasn't been my experience at all.
I just finished the Walter Isaacson book on the development of CRISPR. I dunno, but that development, along with that of inventing mRNA vaccines, the huge improvement in renewable energy economics etc etc make me wonder if we're not doing all that badly.
Maybe instead of the "Golden Fleece" we should be thinking about the "Golden Egg," as in be careful about killing geese.
It's always easy retrospectively to point to what should have gotten more support early on; I'm not sure that's an indictment of the whole system, nor am I even sure what you're proposing. If it's more money for a select few, well, I don't know how you decide who they are and how you can be sure the ones you're locking out don't have even better ideas if only given space to develop them.
I checked Doudna's 2012 Science paper on CRISPR: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6286148/
Interestingly, she cites HHMI funding but nothing from NIH for that work, although I'm sure she had funding from them as well for other stuff. But even without her, it was going to happen. Feng Zhang was publishing on it around the same time: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3795411/
He cites an NIH Director's New Innovator Award, showing that the usual funding sources also knew this was a good direction in which to go.
My point about rethinking how we fund scientific research is to acknowledge that we may not be doing too badly overall (without emphasizing any necessary NIH--> CRISPR link) and to be careful not to let the perfect be the enemy of the good.
And just to dwell on CRISPR a moment more, I suspect that absent the overall supportive environment that NIH funding creates in the field, CRISPR development would have taken a lot longer.
Ah, sorry, I misread you! Yes, I agree that overall, it's working pretty well. I point to the mRNA vaccines as a good example of this, as it's been 2-3 decades of solid single-investigator work from many labs that allowed us to accumulate sufficient knowledge to make this almost trivially simple now.
Man, I dream of a Democratic party that feels confident enough to say their American Families Plan reduces abortions
Safe, legal, and rare!
This is more important to the USA than Israel Palestine or the Middle East.
Thanks for taking on a topic close to my heart. The politics you detail reminds me of Bismark's comment about sausage making, and I try not to think about it. Speaking as a senior biomedical researcher wholly dependent on public support, I strongly disagree with those (including some you cite) who extoll ground-breaking paradigm-shifting "breakthroughs" over mundane incremental science. IMHO science IS incremental, and it is precisely for that reason that it has been so successful in advancing over the past 2 centuries.
From my own perspective, I have only two problems with NSF: first, that they don't have as much money as I would like, and second, that they are too heavily influenced by progressive shibboleths.
Just wandering in to agree with a lot of other commenters that for every single complaint about the current grant allocation system, the one and only solution is more money. There is no royal road to funding good projects. Research sucks, most of it doesn't work out, and you have no idea which projects or which scientists will or won't work out and you're delusional if you think you can do much better than what everyone is already trying their hardest to do. Hovering over some stingy bit of money and desperately trying to send it only to the "most deserving" scientists is just like hovering over a few college admissions spots and trying to admit only the most deserving students, or hovering over a few parcels of land and trying to allow only the optimal apartment buildings, or...we've been over this sort of thing a million times on this Substack. You want to fund more good projects, then fund many more projects, and eventually a few of them will be good. Stop looking for a cheat code.
On a more frivolous note, I would just like everyone to know that the NSF grant submission/administration website is the worst thing in the world. I do not understand how it can have been designed without intentional malice. Fixing that website would increase every American researcher's productivity by numerous hours a year even if you changed nothing else about the system.
That's not really true. Many projects begin with a high probability of success in that they are likely to answer the questions that have been posed. It's just that the answers are not known ahead of time.
the one depressing aspect of the marked up Endless Frontiers Act that you didn’t hit on, but that gut punched me, was that Maria Cantwell (D-WA) went to bat for Jeff Bezos and Blue Origin and added an amendment that prevented NASA from just awarding SpaceX the Lunar Lander contract for the SLS / Artemis program.
Whatever you think of Bezos or Musk, it’s hard to argue that SpaceX hasn’t wooped the entire private space industries butts when it comes to putting up or shutting up.
SpaceX formed about the same time as Blue Origin did, beat countless odds, and has put up many more times than it has been shut up…landing rockets on their tail end, reusing them up to 10x, sending humans to the space station on reusable rockets, etc. Meanwhile, Blue Origin has yet to even put a rocket in orbit.
SpaceX has taken 60% off the launch market by underbidding everyone else for launches by tens of millions of dollars. Even Boeing and Lockheed’s partnership (United Launch Alliance) has failed to compete…and now those companies are leveraging lobbying connections to basically get in on the action at the expense of NASA’s program timeline, etc.
A lot has been written about it, but the fundamental structure of companies like Boeing, Lockheed, and Blue Origin is just not capable of competing with a company structured like SpaceX. And instead of restructuring themselves and learning the lessons that SpaceX has taught everyone, people like Jeff Bezos just keep hiring the same old people to run things exactly as Lockheed and Boeing did, and he’s getting the same result.
https://arstechnica.com/science/2021/05/congress-fires-warning-shot-at-nasa-after-spacex-moon-lander-award/?amp=1
Isn't she going to bat for Boeing, not BO, given that Boeing is based out of WA?
In any case, while Old Space's profligacy is well-documented, SpaceX will have eradicated from the market anyone who's not conducting reusable launches within ~5-10 years, so I'm not too worried about it. Sooner or later the savings will win out.
well, Blue Origin is headquartered in WA and the amendment gives just enough money for NASA to accept the next lowest big (BOs) and basically mandates that they do so…
The worst part is it mandates all sorts of things that ultimately delay the project many years and cost even more millions of dollars to do things like more testing at Stennis Space Center in MS that NASA doesn’t even say it needs to do.
They can flush a few billion more down the tubes for the next decade or so, but it doesn't really matter in the long run IMO. Dread it, run from it, SpaceX still arrives, and the rest of the industry will have to adapt or die. Still, definitely frustrating.
well, it’s the needless delays in the projects that only make them more likely to be cancelled all together that’s the real tail risk. It really isn’t about SpaceX, other than SpaceX proving that it can be done at much much lower cost, and be done better, to boot…
In my field, astronomy, I think it's rare that transformative research is blocked because of the review process for grants. And I think it's unlikely that tinkering with the review process will significantly reduce the chances of transformative work going unfunded.
In astronomy, breakthroughs are driven by technological advances in telescopes, and those advances are limited by funding. So if you want to increase the number of breakthroughs in astronomy, you should increase the funding levels for the research infrastructure (telescopes and the labs that build their instruments). Meanwhile, if you want to promote more good research in general and train more young scientists, you should increase the funding levels for the grant programs at the NSF, NASA, etc. There are plenty of good proposals that get rejected because of the limited funding available.
I found this post incredibly frustrating. Matt would make a point and I would ready my devastating response(*) in my head and then read the next paragraph and he has fully met my objection. This happened several times, to my irritation.
Thoughtful, well-reasoned posts are the bane of my existence.
(*) Including googling "Golden Fleece Award."
I'm interested in these "secret Congress" dynamics. I live in Washington State where we just wrapped up our state legislative session. By all accounts it was pretty big year for the legislature and a number of big, and contentious things got over the finish line: a cap and trade bill; a new capital gains tax; and several police reform measures.
The session was also conducted virtually and didn't allow for the normal level of lobbying and advocacy and general daylight being cast on the things working their way through the process. It definitely wasn't a "secret" legislative session, but the relative productivity of the legislature seems like a real violation of the idea that big stuff doesn't happen without pressure, lobbying, advocacy, and lots of visibility.
I'm curious if similar dynamics played out in other state legislatures? Did others do things 100% virtually? Did this lower the visibility of what was going on, and in turn yield surprisingly substantive outcomes? Is my hunch about this completely wrong?
Half of MY’s articles argue essentially “Uncle Sam should spend more on X.” The obvious retort is “debt is bad, especially when the US just borrowed $6 trillion for the pandemic.”. I’d like to see Matt’s take on 4 issues:
1) How much inflation pressure will the newly deployed $6 trillion create? This is the biggest Keynesian experiment since World War Two, federal spending as a share of gdp approaches has approached World War One levels. There was serious inflation in 1918, why will this time be different?
2) What is the debt carrying capacity of the federal government? During the Napoleonic wars, British debt climbed to 250% of GDP and 1815 Britain was basically a subsistence economy with a few proto industrial sectors. The U.S. could probably borrow more than 250% of gdp without serious problems, but how much more?
3). How quickly should the US approach it’s debt carrying capacity? Should we leave a margin of error for a war or recession?
4) How do you convince swing voters and swing state senators that borrowing a greater percentage of gdp than 2007 Greece is a good idea?
the boring but correct answer is to "wait and see". Basically, we've been paranoid about inflation and debt for decades but things are fine so it's reasonable to spend a bunch and see what happens. If inflation starts growing, we can see it happening and start taking action. The way some people talk you would think that 4% inflation would be the end times.
I basically agree, the question is whether we should borrow 5% of gdp for 10 years or 50%of gdp in two or three years. The former approach is likely to spend money on worthier and more useful things.
I'm not sure I disagree, but we can't really plan like that. This is likely to be the only two year period of the 2020's with a Democratic trifecta in government. I think that's the basic argument for going big, especially the transformative family stuff that people will notice.
The people who are most listened to in the public sphere are much more likely to be affected by inflation than they are to face unemployment.
The short answer is that debt is bad *unless* the debt pays off in the long run because the stuff it funds creates economic growth that enables repayment of the debt. The same people that talk about the "national credit card" ignore the other obvious personal finance metaphor of a mortgage, which increases debt but has an obvious positive purpose.
In response to 1) "why will this time be different?" Probably because almost everything about the domestic and global economy is different than it was in 1918. I'm not smart enough to explain all the deflationary pressures that have basically killed inflation in America for the past 20 years, but there's a reason Bernanke and Yellen couldn't even touch 2% when they tried
As to 1: Not much inflation 1-5% government spending isn’t as inflationary as you would think. It is government setting of prices that causes Venezuelan inflation also being a closed economy without many important hurts, example oil supply shocks. During the Great Depression FDR devalued the dollar relative to gold and added 40% more to everyone’s bank account and there was no inflation. So Giving away 40% of M1 can cause 0% inflation every time I see that in history books I reread it. However, Running the economy hot can leave us vulnerable to a supply shock, if we have a trade war. That is what caused inflation in 1970s so that is the inflation risk. If your are going to cherry pick prices for inflation so will I. I just bout a microwave for $49 my parents bought a microwave in 1985 for $49 Costco hotdogs $1.5 same as 1986.
2 debt to GDP is situational. What are Interest rates are you a reserve currency etc. You are correct in saying waiting till the market says no more can be painful. Especially if it is other countries holding the debt. In the USA case the fed is holding the debt. the government issues the bond and the government buys its own bonds... Inflation is a much bigger threat for that reason. Based on history I, am only worried about inflation if the government starts mandating wages or we have a trade war. And no minimum wage up to 15 wouldn’t be too bad because look around most places fast food is already above $15 who even makes 7.25 anymore? In conclusion We should not worry about debt to GDP if the fed holds a big chunk of it and banks are required to hold it for a safe asset mix. I will worry about percent of debt to GDP when China and the EU own our debt. If that where the case I would worry about our current levels.
3 how quickly should we up our debt to GDP ratio on average the same as GDP growth but in a recession. well like 5-25% a year for 2-3 years max. Here is why you will eventually have economic growth to cut against the debt ratio growth during good years and our welfare state is needs based so spending automatically goes down in good times. Assuming you will have 4-6% economic growth is reasonable especially if you are dumping money into the economy. And before you say we have exceeded those percentages keep in mind the pandemic has been over 2 years. Finally, inflation combined with low interest rates helps the debt situation.
4: Swing Voters don’t really care about debt to GDP ratios they care about cancel cultural they care about abortion and gun rights. They care about the woke mob. I have 3 or 4 conservatives I listen too and that is what they go on and on about. True swing voters are like Joe Rogen he mostly doesn’t talk politics but he talks about trans in sports, cancel culture and that he doesn’t want his taxes going up. telling swing voters here is tax free stuff is popular. Also Joe wants legal weed so give swing voters that. As for swing Republican senators they are talking about the debt in bad faith you can get republicans on board with debt increasing polices but adding a tax cut to the rich that makes the debt worse.
I don’t see 4-6% real gdp growth as achievable and I don’t care about nominal growth.
Also great questions
I agree in the long term disagree in the short term. We contracted by 4% in 2020. We contracted because waiters had to get other jobs often jobs in higher productivity parts of our economy, productivity grew in 2020. Now in 2021 the entire hospitality hospitality industry will come back into existence and they will have to increase workforce participation to do so because all the waiters got slightly better jobs at Amazon. Given that 1% real annual GDP growth is reasonable or -4% 2020 +6% but only for 2021. But yea over a period of 3-5 years 6% real GDP growth will not happen baring insane policy changes.
And in response to 4) I think Matt's typical retort (correct IMO) would be: "you help them a lot and play down culture war stuff so they like Democrats"
or you make the debt even worse by adding a random tax cut to the ultra rich.
tax cuts for the rich aren't inflationary though (they are bad for other reasons)
To address #4, voters don't care about debt or the deficit and never have in the modern era (as evidenced when the GOP's "deficit hawks" blew up the deficit to cut corporate taxes) so I wouldn't bother worrying about this one at all.
Tax revenue went up after the tax cuts.
Also the tax cuts eliminated a lot of exemptions.
So corporate taxes didn't really go down much overall
A whole bunch of US science funding mechanisms were created either at the beginning of the Cold War (like NSF) or immediately after Sputnik appeared to show that the USSR was ahead in science and engineering (DARPA). It's not surprising that the China rivalry is also driving US science funding.
On grants: besides the problems described here, the big problem imo with NSF (and other!) grant-making procedures is the amount of work that is spent applying for them. Academics spend a large portion of their time applying for grants, and that in turn reduces the amount of time they spend doing the research the grants were supposed to support. Since NSF grants tend to be very competitive, this means that every grant made induces a lot of science *not* to happen, as for each scientific project funded, many scientists have been incentivised to spend a large amount of time not doing science.
This would also be ameliorated by increasing the amount of money and hence decreasing the level of competitiveness!
(This isn't just an NSF or just an American thing, of course, it's a product of the incentives and limited perspective of grant-making bodies. But the NSF is a really notable case.)
In the context of a world where uncontroversial bills can pass through regular order like this, I don't think a watering down of ambitions like you saw with Endless Frontier is necessarily the end of the world. It might actually be a good thing! Some parts of this bill will work and others won't. If you think you'll get a second crack at this once we've learned from our experience, the rest of the money can be more appropriately targeted. I don't personally think it's a good idea to just throw $100 billion dollars at the NSF as it stands IF you believe congress can return to this issue and finish the job.
I think it's a consequence of the omnibus/reconciliation world of legislating where every law feels the need to 'solve' whatever problem it is addressing in full and in perpetuity. You see that with the ARP, which objectively has a TON of wasted and misdirected money in it. But it is defended correctly on the grounds that it is 'better to do too much than too little.' Left unsaid here is the assumption that you only get one shot at it. Obviously it would actually be much better to do too little and then a month later to do the right amount while taking into account what worked and what didn't.
If you believe in the definition of creativity as the synthesis of new ideas from disparate ones (i.e., connecting non-obvious dots) then the current funding model is designed to thwart creativity aggressively.
If scientist from Field A rubs elbows with scientists in Fields B and C and gets an interesting idea, their only hope of funding is to collaborate with people from Fields B and C. If they go through the front door, their proposals will be refereed by experts in Fields B and C, who will see the lack of familiarity with the norms of their field as a sign of amateurism and will be extremely skeptical than an idea they didn't think of, and don't understand well, is going to solve big problems in their field. (They're usually right; these are high-risk endeavors by definition.) But if the scientist from Field A tries to team up with scientists from Fields B and C, they will run into the same basic problem of trying to sell weird-looking ideas to skeptical experts, because they will have to apply in an existing call, which is always going to be based around known problems and potential solutions based on known ideas. (There is a feedback loop from program officers attending conferences and wanting to fund already-successful labs that is difficult to break through.) If they manage to get some funding, they will face the same problem all over again with editors and referees at the top journals. In the end, the disincentives to pursue truly creative research are so strong that only a few scientists at well-funded, high-profile institutions will even try.
In practice, the result is that a handful of labs define what is popular and then seed the academic landscape with their progeny, who fulfill that prophecy. Those who stick to the defined tracks, know their place and maintain good relationships will find friendly program officers and calls that were effectively written by their former advisors. They will write papers that get refereed by people who are primed to recognize that they are publishing an acceptable level of new results in their narrow sub-field, which will perpetuate the cycle. (It is hard to overstate how hard the system pushes back when you try to move out of your lane.) Much good science gets done this way, but when you under-fund science, it essentially all ends up flowing into this gatekeeper model, which is nigh incapable of producing actual fundamental breakthroughs because it rewards competition on papers, awards and profile (fame) rather than ideas.
We're essentially mining the breakthroughs of the 20th Century without seriously investing the approaches to science that created them. The diminishing returns is not because science is getting harder or more complex, it is the result of too many people are standing on the shoulders of giants rather than becoming giants themselves.
The best way to get "scientists from Field A to rub elbows with scientists in Fields B and C" is to support America's research universities. They gave us the 20th century scientific revolution, as well as the internet era and the genomic revolution.
And one way to help universities is via grant funding. But other ways matter, too, since they are big places that do more than science. (And their fostering of a hundred disciplines helps them break out of ruts.)
So: more money for research labs! But money for universities in general.
That makes sense to me. It seems like micromanaging too much of the funding through centralized federal gatekeepers has created a whole ecosystem built around gaming that system. From my limited anecdotal knowledge, it seems like the system we have now basically depends on and empower professors who are best at acting as rainmakers for their university, not necessarily those with the best ideas.
Maybe it would be better to distribute a large portion of federal research money essentially as block grants to major research universities in each state.
The sad thing is that the NSF is way better in this regard. They tend to fund people with good ideas with very little micromanaging. But most of the money flows through DoD, which is the epitome of micromanaging. It's also a good ol' boy network with opaque review processes that, ironically, is one of the better places for fundamental research funding; fundamental research in narrow topics that is done by people who can break into the network.
I once saw a chart of government funding to my university. The part that works on defense projects dwarfed the rest roughly 3:1!
It's a different way of doing science, that's for sure. Their projects practically are contracts, including with explicit milestones they need to meet to get the next increment of funding. And like you say, it can be ironically fundamental stuff (the most "out there" project I've worked on was with those folks!).
And empower university administrators instead of scientists? That would be significantly worse.
As for professors acting as "rainmakers", I'm not entirely sure what you mean. Believe it or not though, the most well-funded scientists also tend to be doing great science as well.
No doubt it varies by field, but I'm not sure that the administrative and management skills necessary to be in charge of a large lab full of rats, mice, grad students, research assistants, etc, optimized to churn out large amounts of small-ball research papers and grant proposals necessarily correlates with the skills and creativity needed to to make scientific breakthroughs. Sometimes it obviously does but what about all the times it doesn't?
No one gets grants because of their administrative skills (just ask the folks working in the big labs!). And if all they churn out is small-ball research, their proposals will eventually have trouble getting funded with the critique that its "only incremental".
Having been in big labs and having been a small lab myself, what I can say is that the big ones are significantly better positioned to make breakthroughs. This is for a number of reasons, including more cross-pollination of ideas between people, more resources to try and/or develop innovative approaches, and (to contradict my own post above a bit), more intellectual freedom, at least for the postdocs and students (who are actually the true innovators anyway).
I'm sure you know more about it than me so I won't argue! My impressions are just extrapolated from a few anecdotal perspectives from conversations with people. I wasn't suggesting, though, that people are getting grants because of administrative skills despite lacking science skills. More the other way around, whether the system is losing the benefit of absent-minded professor types who might be scientifically brilliant but lack administrative skills.
I think it's unlikely that the government, ruled by political processes and acceptance by the body politic, will ever be truly daring when it comes to funding out of the mainstream scientific research.
But if Elon Musk et al want to throw a few billion bucks at crazy ideas, I'd say go for it.
You majorly TLDR’d here dear friend. But my long morning bowel movement allowed me the time to read it.
I’d love to note your following point:
“ We're essentially mining the breakthroughs of the 20th Century without seriously investing the approaches to science that created them”
Do you have thoughts or opinions about whether or not you would apply this conclusion to the current trend of AI/ML? I would say it fits nicely. These practitioners often have no understanding of the problem they’re trying to solve nor any reading of the robust literature surrounding it. This leads to approaches to solve problems by throwing out research, and throwing in tons of data and overly accessible software packages to generate an “AI solution.” That... rarely ever bears much fruit.
I am no expert in AI/ML, but I have dabbled with it and sat through many, many talks where people are trying to apply it to various problems in physical science. I think, though, that it is another good example of mining old breakthroughs, since (AFAIK) they mostly happened in the 80's and were just waiting around for computing power to catch up.
It's also not clear to my what problem ML actually solves. It is unquestionably good at optimizing certain tasks at scale, but the best AI is usually judged by its ability to do something that a human can already do, though often less efficiently (in the literal sense that it requires more people). How is it going to break new ground upon which future breakthroughs will be built?
I think the protein folding is a good example of how AI changes what is possible.
But isn’t the actual work there done by quantum chemistry that was developed int the 60’s? ML can make the process more efficient, but it is clueless about what it is doing without QMMC/DFT, etc. to validate the results (or evaluate fitness in a genetic algorithm).
Well sure - and flying rockets uses mathematics that was developed centuries ago. That doesn't mean taking another step in on the path doesn't matter.
Everything builds on something else in science, of course, but the main advances from AlphaFold really come from just how much resources (people + GPUs) they threw at the problem. Others would have gotten there eventually though. Apparently the company DeepMind lost almost $650m last year alone!
https://www.cnbc.com/2020/12/17/deepmind-lost-649-million-and-alphabet-waived-a-1point5-billion-debt-.html
You can read the paper for AlphaFold 1 here to get a better idea of exactly what they did: https://www.nature.com/articles/s41586-019-1923-7
That they lost that much money is great example of how private industry is willing to invest in research if they sense the payoff is significant enough.
"Others would have gotten there eventually though" - to what does this not apply?
I don't want to come across as dismissive of ML; I think it's great. It's like the transistor, which was revolutionary and a major scientific and technological accelerant. But the underlying breakthroughs that was based on weren't so clearly built on something else in science. In fact, the understanding of electrons and solid-state materials that enabled the invention of the transistor were radical departures from existing science that challenged settled science. Those sorts of bold, non-intuitive assertions and discoveries seem to me to be less frequent, in part, because of the incentives of modern funding models.
Of course, for all I know the scientific revolution in the life sciences is producing earth-shattering, paradigm-shifting breakthroughs on a daily basis akin to the quantum revolution a century ago. I wouldn't know because the acronyms alone make that literature impenetrable to mere mortals such as myself.
I do think that applying AI and ML to a problem, while often extremely difficult, is about as low on the creativity scale is one can go. The bottleneck there is the paucity of experts; if you can plausibly apply your domain knowledge to an ML-driven project--and find people to implement it--money will rain from the sky. It doesn't even have to produce anything useful, since it is likely sufficiently novel that the implementation itself is taken as proof of success. (Plus people are afraid to ask questions, lest they reveal their ignorance of your jargon-filled presentations.)
I mostly support the amendment taking some $ from the increased NSF to increase DoE R&D funding. My experience is that DoE programs in energy efficiency and advanced manufacturing are more engineering /application focused than NSF. I think that some of the big developments will need some patient money to cross the chasm to market-ready technologies. I think that some of the DoE programs are better set up than NSF to support those pre-commercial developments.
I agree the ambitions shrinking is bad and more money for science is great! I think ARRA roughly doubled NSF funding for a year and they had no trouble finding worthy projects to fund. I however don’t think it’s a big issue that some of the money went to DOE labs, which I think are doing some of the same technical work anyways that the bill envisioned. Also, with regard to the total amount of funding, I think there is also the issue of where in the research/development chain the funding is going to. If the funding represents primarily an increase in basic research funding that is very good because that is the kind of research that is least likely to be funded by the private sector. If however the funding is split more evenly between research and development I think this is much less good. Also, just because Congress authorized a certain amount of research in this bill doesn’t mean Biden couldn’t get the federal bureaucracy to allocate some more money to research than it already does, or to set research priorities that are particularly worthwhile. He also has the power to tackle some of the science bureaucracy issues you point out, though paradigm change is hard. And I hope with the return of earmarks that Congress becomes less interested in individual spending decisions and more interested in the broader outcomes government achieves. There was a real move in the last half of the 20th century to small ball Congressional thinking, and generally much less ambitious legislating. A Senator should be criticized, not praised, for worrying about $84,000 in spending. Lastly, I’d like the government to do some more prizes where the reward is received once the discovery has been made. This could incentivize a broader set of folks to look into important problems, and it has the benefit of avoiding the grant bureaucracy you rightly criticize.