One day we will have cured all diseases, cut transcontinental travel times by more than half, and enjoy cheap, sustainable, plentiful energy. That day is not tomorrow. But why not? The act of posing that question may seem preposterous and impatient: hard feats demand hard work and long time horizons. But asking it is completely reasonable if it helps us find reasons for those long timelines and ways to shorten them.
Sometimes we know we want to solve a problem but the way to get there is unclear, as the multi-decade failure to develop anti-Alzheimer’s drugs show: even with unlimited resources we wouldn’t be able to swiftly produce any such drug in a few months: we need to continue exploring and researching into the fundamentals of a given field and let scientists freely survey the domain.
But often that is not the case: sometimes we know roughly how to get there, if only we had the will, the coordination capability, and the resources. Like going to the Moon or many megaprojects. We argue that these problems are present across science and technology, and solving them could unlock great progress for an entire field.
We refer to these as bottlenecks.
In the past few years, we’ve heard narratives of generalized stagnation throughout the economy, and a number of vivid examples of a future that was seemingly promised but never delivered have entered the popular discourse: Notoriously, we still don’t have flying cars.
But whereas landmark books such as Tyler Cowen’s The Great Stagnation was published in 2011 or Gordon’s Rise and Fall of American Growth in 2016, one can also find books making the case for the exact opposite argument, from the unabashedly optimistic Abundance (2012) to Kurzweil’s classic The Singularity is Near (2005). There are better things to do than arguing who is right. The Great Stagnation is here but not evenly distributed: Total Factor Productivity—and with it, GDP—growth have indeed stagnated since 1970. Airplanes are not flying any faster. But at the same time, we have made miraculous advances elsewhere—perhaps more in the life sciences, electronics, and software sectors than elsewhere—from mRNA vaccines to smartphones to viable electric vehicles. But rather than asking the high-level question of “who is right”, we could be asking what concrete steps we should take from here on, field by field. That is, for each given field of technology and science, what would it take to speed it up, regardless of its past progress. This framing does away with the past rates of progress and rather looks forward at the future that unblocking these bottlenecks could bring: if a field is proceeding slowly, how do we accelerate it? If a field is proceeding at breakneck speed, how do we further speed it up?
Bottlenecks are not necessarily the lack of a particular enabling technology: A bottleneck can be cultural, institutional, or technical. Often bottlenecks are entangled but can be unlocked iteratively. For example, initially mRNA vaccine production was bottlenecked by the lack of manufacturing capability of lipid nanoparticles (a scaling issue). Supersonic aviation remains banned overland in the US due to FAA regulation, making it more difficult for companies to start in that space (a policy issue). Nuclear energy remains a fraction of the energy pool because of a combination of cultural perceptions of the technology (scary mushroom clouds) and regulation that makes it onerous to certify new designs (an entangled cultural-political issue) but that may be partially overcome via new reactor designs (a technical issue).
Bottlenecks eventually get solved, it is a matter of how and when. Boom Supersonic did not have to wait for flights over the US to be permitted nor wait for so-called “low-boom” technology to be developed. Rather, it was the realization that developing supersonic airliners—even if only operable in transoceanic routes—would bring a profit. Those profits can then be reinvested in larger, less noisy, faster, airplanes. It is the same approach that Tesla followed: instead of waiting for better batteries and networks of chargers, they took it upon themselves to build the charging infrastructure and make electric cars appealing in their own right rather than a declaration of ideological affiliation for environmentally-minded consumers.
Last June, together with Adam Marblestone and Leverage Research we organized the 2021 Bottlenecks in Science and Technology Workshop, a small private gathering bringing together technologists, entrepreneurs, and academics for a weekend of presentations and unstructured discussions on what is blocking progress in a range of fields, from aging and energy production to psychology and economics.
We held the workshop at the headquarters of Boom Supersonic in Denver, Colorado. This was a perfect setting, as stagnation in aircraft speeds is one of the most iconic and recurring examples of technological stagnation—with the Boom crew at work on their XB-1 demonstrator, this gave us all a constant reminder that dedicated teams can make real progress on seemingly insurmountable problems. Facile objections like "the project is capital intensive," "the founder lacks aerospace experience," or "this could only be done by large, established players" aren't stopping Boom.
We had a total of ten of these presentations. We wanted to not only find out what each field is bottlenecked by, but more broadly learn about the act of finding bottlenecks itself. Each presentation was unique: Some went broad and others deep, some highlighted current efforts in the right direction, others proposed projects with a concrete price tag that could unlock further progress.
Finding these bottlenecks is not easy. First, because one can be too vague about it, just saying that “it’s because regulation” doesn’t cut it. Which regulation is it and which Senate Subcommittee does one have to go through to change it? Second, because one can often find a clear cause for any one given bottleneck, making one wonder if this bottleneck is the real bottleneck or if it’s rather something upstream. If it’s a regulation, why is it there? There is probably a coalition of interests that put it where it is that may oppose changing it. What to do about that then? Heuristically, we came to think that something counts as a distinct bottleneck (and that the analysis could stop there) if it could be reasonably solved by a single project that includes all its “upstream” bottlenecks.
Finding bottlenecks is one thing, but doing something about them is a completely different issue. Pointing out what’s hard to accomplish is orders of magnitude easier than accomplishing it. So for the workshop, we made sure to showcase actual efforts that go beyond problem finding and move into problem-solving.
Patrick Collison presented a keynote on Fast Grants, a funding mechanism that supported scientists doing urgently needed research to address COVID-19, from clinical trials of repurposed drugs to the development of rapid tests. The bottleneck there is not so much a lack of money (If there’s a well-funded area of science that’s biomedical research, and the US’ NIH funds 60% of the world’s non-private biomedical R&D), but rather an insufficient sense of urgency. Fast Grants was launched in 10 days and was set up to decide on grants within 48 hours (at least in the early days). 64% of scientists that got the award said their work wouldn’t have taken place otherwise.
Adam Marblestone presented on the concept of Focused Research Organizations, with two working examples focusing on connectomics and the study of non-model microbes. Focused Research Organizations are non-profit entities that are organized to pursue a given goal rather than blue skies research, akin to “mini Manhattan projects”.
I presented on a project I have been working on for the last year, the “Rejuvenome”, a project that seeks to produce a freely available, comprehensive dataset of how multiple interventions and combinations thereof affect aging in genetically diverse aged mice. The project recently got funded and is being hosted at Jed McCaleb's Astera Institute. This effort was conceived as part of a bottleneck analysis informed by conversations with experts in the field. I discussed not only how the project came to be but also how we could have gone faster: what the bottlenecks in bottleneck solving are.
Between the presentations, there was plenty of unstructured time for discussion—the comparative advantage of in-person events—and so we had a session comprising dozens of ad-hoc brief “lightning talks” with sessions covering a variety of topics from DeFi and how blockchains could support more diverse forms of science funding to a 5-minute long ideation session about new institutions that should exist.
With this workshop, we hope to help catalyze a network of like-minded individuals around the idea of bottleneck analysis. The reason bottlenecks exist is to an extent an information problem: if only the relevant parties were fully aware of the benefits of bottleneck solving and the specific resources required we would see more work on this class of problems. That’s why we are interested in talking to you: if you see something that’s broken that is unlikely to be fixed soon enough by existing mechanisms, reach out to us! We may be able to help. Even if you don’t see yourself working on the problem directly, connecting the analysis with the right people—making it visible to them—could lead to some progress actually happening!