This is the best paper written so far about the impact of AI on scientific discovery aidantr.github.io/files/AI_innov…

Tons of stuff in the new House Select Committee on China report that came out today!

A few highlights... 1) Get $$ to our tech/science agencies

It's notable that even after the passage of CHIPS, very little of the increased funding has actually been appropriated and is currently stuck in limbo. Fund NSF, NIST, and DOE!

Very excited to announce a new series of short papers on NIH reform from an event @IFP co-organized with @BrookingsInst and @GoodSciProject.

With an annual budget of ~$47 billion, the NIH is our most important health R&D funder, but we can make it even better. The NIH plays a fundamental role in nearly every aspect of biomedical research.

In addition to being the largest funder of basic research, ~50% of patents linked to drugs approved by the US Food and Drug Administration directly cite NIH-funded research.

Lots of bad thinking out there about how "stealing technology" works.

I'm begging people to read "How Technology Grows" by @danwwang. Something Dan hints at, but I want to make explicit, is that process knowledge is not really an attribute of any particular individual, but instead a phenomenon at the institution / community / ecosystem level.
danwang.co/how-technology…

"What the VC mindset tended to miss was the extent to which the entire 'take big swings and hope for the best' mindset was itself significantly downstream of macroeconomic conditions rather than being some kind of objectively correct life philosophy."
slowboring.com/p/how-jay-powe… I am similarly torn in both of these directions. The vibes are unclear.

Rebuilding state capacity won’t be as simple as turning the government off and back on again.

I have a new piece with four practical strategies to make an entrepreneurial state, centered around the idea of creating new, semi-autonomous offices within existing federal agencies. 1) Use new offices with distinct operational niches to spark a bit of internal competition, similar to the way that Netflix was able to transform its business model from mailing DVDs to streaming video.

What are key, policy-relevant questions where we could use more academic research?

I spoke recently to econ PhD students at the NBER Innovation Bootcamp about holes that I think are not well-answered by the existing literature.

Here's my list of 11 in no particular order:🧵 1) Should optimal intellectual property protections vary across industries or technology areas based on the underlying structure of knowledge production?

(weak protections in software vs strong protections in pharma as an archetypical example)

If we built Jurassic Park and the dinos escaped, it would be extremely easy to contain or eliminate them with modern weaponry If the U.S. military decided it was a top national security objective to eliminate feral hogs, I assure you we could wipe them from the face of the Earth.

https://twitter.com/JerryDunleavy/status/1528440788316655616

Wow. The UK is really getting aggressive about recruiting high-skill immigrants.

Can you imagine if the US created something like this?

https://twitter.com/Sam_Dumitriu/status/1418559832949460997

This has actually been an ongoing theme for the Johnson administration, here's a piece I wrote at the beginning of last year:
nationalreview.com/2020/02/us-imm…

Moderna has started human trials for an mRNA-based flu shot!
theverge.com/2021/7/7/22566… B/c normal flu shots take a while to grow, we have to sort of guess what variation will work best before we know what we're actually facing that season. Consequently, we see pretty low effectiveness rates (40-60% usually)

With mRNA, we could wait until we have a clearer target.

I'm all on board with more applied R&D spending, but I think this is misguided from @RobAtkinsonITIF:
1) Basic research is fuzzy. It obviously has more spillovers than applied, but there is a lot about scientific knowledge generation that is tacit.
itif.org/publications/2… That the best mRNA production facilities are in the same country that invested most in mRNA basic science is not a coincidence. Katalin Karikó is now a VP at BioNTech because her basic science background was helpful for translation to applied work.

lol, I missed that a TikTok prohibition made it into USICA Just strike the last three words here, imo

A vaccine buy-out is preferable to this on several levels:
1) IP suspension through the WTO requires unanimous approval. A buyout can be unilateral.
2) A buyout helps get at the underlying manufacturing know-how that is actually more important than the IP.

https://twitter.com/AmbassadorTai/status/1390021205974003720

3) Incentives. Yes, public R&D helped fund this vaccine, but we want to send a very loud, very credible signal that it's worth spending billions and billions of dollars on speculative treatments that have a good chance of failing.

🚨 New post 🚨

I wrote 4400 words on various proposals to scale vaccine manufacturing capacity.

Should we...
- Suspend IP protections?
- Offer incentives to manufacturers?
- Use this as an opportunity to bootstrap new vaccine capacity around the world?
agglomerations.tech/how-the-us-can… I think it's underrated the extent to which ~Most Of The World~ is not getting vaccinated right now.

That's problematic both morally (it's bad for people to die) and as a public health crisis (it increases the odds of a significant Covid mutation).

I don't think people have quite internalized *how* the US became the global leader in science and technology. It's partially a story of massive global talent migration.

And it's important to get this story right if we want to maintain leadership.
thenewatlantis.com/publications/t… Around 1900, the US was a bit of a scientific backwater when compared to Europe. Most of the exciting work was being done in Germany and Austria, especially.

Then came three enormous waves of academic and scientific talent to the US.

definitely no way this will be gamed ahahah

It is true in some sense that "we" (policymakers, posters, etc.) don't know how much debt is too much, but we have a really underused crystal ball that we can look into called "long-term market expectations"
noahpinion.substack.com/p/no-one-knows… The main risk of debt is much higher inflation, and thankfully we have a variety of ways of asking the market how much inflation it expects over long time horizons.
fred.stlouisfed.org/series/DFII30

My article yesterday was pretty optimistic about the future.

So let me add a short thread today with some possible reasons for pessimism, even if you agree with the piece.
agglomerations.tech/cracks-in-the-… 1) One interpretation of recent events is that we've been making expensive investments for decades and now are *finally* starting to see results.

But if these are one-time payoffs, then it's still taking longer and longer to see results and that's a story of stagnation.

Really thoughtful paper from @timhwang on ways of shaping the "terrain of AI competition".

Essentially, we can strategically prioritize investment in specific AI techniques/subfields that we think will help democracies relative to authoritarian regimes.
cset.georgetown.edu/research/shapi… For example, authoritarian regimes are perceived to have a data advantage over democracies because they are more willing/able to mass-collect data from their citizens without privacy concerns.

1/ The new 2019 AI Index report is out! It's a really great and ambitious project. Here are a few graphs that stick out...
hai.stanford.edu/sites/g/files/… 2/ First, here's where the growth in AI papers in coming across China/US/Europe:

I like @elidourado's list of things we should do if we are ethically compelled to max long-run sustainable economic growth. But I think this part is wrong. cato-unbound.org/2019/01/28/eli… I suspect true growth maximation would do _more_ social engineering than present, but in ways that actually are growth enhancing. There's no reason to think individuals are fully internalizing the positive externalities of facilitating growth in their own consumption/investment.