My Metascience 2022 talk on new scalable ‘technoscience’ laboratory designs, and potential upcoming ‘New National Purpose’ reports

New National Purpose #3 + my talk at Metascience 2022 on the need for a new kind of laboratory structure to unleash creative discovery and invention to complement academia

I previously put out a vision document on a ‘twin for ARIA’, Lovelace laboratories [read here]. These labs were key recommendations of both of the Blair-Hague New National Purpose reports, as ‘Lovelace Disruptive Innovation Labs’. If you read that and were one of the people who asked ‘how could we make this happen?’, this blog and the follow up report might interest you.

This blog is a kind of open email to people who may be interested in giving input to a report on this that we are considering doing as part of the New National Purpose series in October/November.  We are exploring doing a report in this series expanding on the Lovelace labs vision in substantial detail and putting out a plan to enact it, and working to make it reality. It has a low probability of success but, if it works, potentially high impact.

This blog has two parts.

Part 1 is an outline of a talk I gave about it at the Applied Metascience retreat in San Francisco in May 2022. It has a different framing to the one used in the original vision doc in terms of explaining common elements of Xerox PARC/Bell Labs/LMB. 

It can still be improved greatly so any suggestions are welcome. This is a very rough and skeletal draft as a prep for a much more developed report. I’m providing this different framing partly as a setup for part 2 of the blog.

Part 2 is a call for people who may be interested in giving input into the report/plan for trying to do something like this, and for trying to come up with an improved framing. There is an interesting mix of people ranging from former senior royal society figures through to undergraduate historians that we are putting together to flesh this out in a way that could be put to potential philanthropic funders/government, perhaps internationally. 

I’m cutting myself off from the internet for August from the moment I publish this blog, so I’m putting this out and hopefully will come back to some interesting suggestions! My public email is jamesphillipspublic@gmail.com and my twitter is open to DM’s.

PS - I wrote this in a hurry, so apologies that is isn't polished.

‘You will not concede me Philosophical poetry. Invert the order! Will you give me poetical philosophy, poetical science?’ Ada Lovelace

Metascience 2022 Framing of Lovelace Disruptive Innovation Labs

In the early 2010s voices were beginning to emerge saying something had gone significantly wrong in our approach to science and technology, and economic growth more generally. These included figures like Garry Kasparov, Peter Thiel, and Michael Nielsen. You can watch Kasparov’s talk here for a much fuller exploration of this, or read my interview with him after that talk. 

I won’t go into the full details of this there, some of which are explored in notes I wrote in 2019/2020, but examples at early stages of research include:

• A growing ‘oligarchic’ power structure in research, with young people living in a state of servitude to senior ‘scientists’ who were often in reality administrators putting their names on the papers/ideas of junior people.

• Relatedly, an unsustainable ponzi-scheme like career structure, with ‘trainees’ being used as cheap labour without an onward path.

• A systemic aversion to risk and long-term support in research funding.

• A lack of structures to support collaborative team science bringing diverse skills to bear on problems, spanning disciplines. 

• An absence of ‘vision/mission-oriented’ research laboratories, with an emphasis on integrating discovery and invention under one roof. 

• A lack of a role for invention/methods development research at the pre-commercial stage - Brenner said that “Science proceeds through new methods, new discoveries and new ideas, usually in that order”, yet the new methods part is woefully underfunded.

These issues have motivated a search for other ways to organise science and technology research that could result in a better and more creative culture, and a higher longer term return in terms of transformative technologies that improve our lives. In short, supporting better organisation of research is in the enlightened self-interest of all of us.

Some of the above points had been highlighted by Gerry Rubin in a visionary piece in 2006 long before they became mainstream. 

Rubin was a graduate student of Sydney Brenner’s at the Cambridge Laboratory for Molecular Biology. I reposted an excellent interview with Sydney here where he highlights some of the negative changes that have occurred in the organisation of science and technology.

Analysis of the successes and failures of Rubin’s institute, Janelia, are complex, and I’ll blog later on this. Scientifically/Technologically it was a tremendous success.

Since Gerry Rubin founded Janelia, a number of other labs have arisen that have been consciously experimenting with institutional design. Focussed Research Organisations by Convergent Research, pioneered by Sam Rodriques, Adam Marblestone, Anastasia Gamick and others have also diversified the structural pool. 

I could add to this image DeepMind, OpenAI etc, neither of which resemble conventional public research laboratories/universities. As Demis has said to me, DeepMind has a lot in common with how Janelia was structured in its early days, and Demis studied Bell Labs in coming up with that structure. 

Identifying common features across high impact laboratories

Next I want to talk about a combination of historical analysis and policy. 

Alongside the analysis our network did that motivated creating the Advanced Research and Invention Agency (ARIA), we also closely studied some of the most impactful physical laboratories in history, looking for commonalities that might suggest new scalable models for research labs. This was seeded in part by Gerry Rubin’s observation of similarities between the Cambridge LMB and Bell labs, and through meeting people who had been in those environments like Sydney Brenner from the early Cambridge LMB, Eric Betzig from Bell labs, and others. Some of the conclusions are shared with Gerry’s, other conclusions are different, especially when Xerox PARC is included in the picture. 

When we look at the 20th century, there are a number of technological revolutions that have been particularly transformative. We can debate the exact ranking of the top 10 technology revolutions, but in my judgement three are surely near the top:

• Molecular biology.

• Telecommunications.

• Networked personal computing.

These are now so ubiquitous in their implications for civilization that we barely notice them. 

I start with the assumption that research of this kind is important for improving the world and we should seek to do more of it, and they also generated tremendous wealth and prosperity. In the UK context at least, we have not been the home of any such transformative tech revolution for a long time, unless you count Google-funded DeepMind, which structurally has a great deal in common with what I’m going to discuss. 

Each of these revolutions can be strongly linked to a single physical laboratory, which had they not existed might have substantially delayed the development of that technology (the ‘deletion test’).

Each of these research laboratories had: 

• a transformative impact on the world

• a widely reported unique culture distinct to the mainstream

• minimal obstacles to research

• ability to pursue work that defied consensus

• a highly collaborative environment

There have been studies of PARC and Bell Labs. ‘The Dream Machine’ by Mitchell Waldrop is excellent on the ARPA-PARC vision, and ‘The Idea Factory’ by Jon Gertner is excellent on Bell Labs. There isn’t a single book that overviews the LMB that I know of. Lots of sources on this are in part 2 of my 2019 notes here.

However, most studies look at each of these laboratories in isolation, asking what made them special:

For each of them, you could draw unique features but be unable to know how important it really was. For example:

• For the Cambridge MRC Laboratory for Molecular Biology (LMB) you could say it was crucial that it was founded by a group of people who had just made discoveries that had opened up a whole new field of research (Double Helix). 

• For AT&T Bell Labs, you could say the ‘problem pull’ from the AT&T phone network was crucial.

• For Xerox PARC, you could say the network of brilliant junior mavericks funded by ARPA through the ‘Licklider’ ARPA-IPTO, which essentially moved into PARC, was crucial.

This, however, has some limitations. These are ultimately n of 1 observations. This is not to say these are not important observations, but they are of limited utility in trying to identify a general, scalable model to learn from these environments.

A complementary approach was suggested by talking to people who had worked in these environments, and through further study of it. 

This is to ask what, of the otherwise quite distinctive features of these laboratories, were common between them:

What was striking in discussing these institutions with people who had been there was how, despite major differences in focus area, location, and size, these laboratories had remarkable commonalities.

What was also striking was how different these commonalities were to be the basic assumptions of how much R&D is organised today, and how little interest there seemed to be in this. Further, the model of research they seemed to imply appeared to dovetail very well with a lot of the challenges we see in our current research systems, suggesting at least part of a solution.

I’ll organise this into 3 sections, which differ from the vision doc, and may be an improvement or not.

1. Physical laboratories with long-term, single source funding

The first observation here, which is circular to the definition used but nonetheless very interesting to me, is that in three of the most important technological revolutions of the past century, a single laboratory can be identified which had a markedly disproportionate influence on the field. We can see this phenomenon repeating with AI today. 

Secondly, the basic funding setup of these laboratories is also distinctive. Each of these labs were supported by a single major funding source, with a minimum-to-no other sources of funding. 

I suspect this is important for a range of reasons including that the addition of multiple different funders prevent cohesion of vision and approach, and the need to continually please multiple different external organisations. As you add more and more funding sources, you become more and more generic in your approach, more tied into competing oversight/evaluation methods, and more bureaucratic. I could give many examples of this in the UK context today. These issues are especially pronounced when what you are doing is far from the mainstream, making it hard to convince one, let alone several, funders to give continued long-term support. Put simply: with each additional funder, more compromises have to occur.

Thirdly, this was ‘hands off’ monopolistic or public funding. A common objection is that the private sector will do this kind of research. But in each case we see either public funding, or an organisation with an effective monopoly on a product/service (AT&T and Xerox), providing the funding. Again, we see this repeating with AI today with Google DeepMind as well as OpenAI to a lesser extent. 

Almost all of the major advances in AI in the past decade have come from monopoly-like funding sources like Google.  Yet there is very little consideration of how they approach R&D organisation differently - it isn’t just money. People wonder at the amazing results DeepMind produces, but almost nobody asks how DeepMind is structured internally to achieve those results.

Alan Kay describes this kind of funding as ‘mad money’, Brenner the ‘casino fund’: these organisations could afford to fund very speculative, long-term work in a way not clearly tied to short-term profit considerations. 

Fourthly, these labs were aiming over a multi-decade time horizon, though each transformed their relative fields at much shorter time horizons as this long-term perspective let them explore different areas of the technological frontier others could not do. 

These properties allowed the creation of what I termed ‘connected bubbles’, free to collaboratively defy the consensus of the time, and avoid the perils of needing to ‘please the gallery’ of the status quo, yet also free to interact with the existing mainstream system when appropriate. 

2. Cycles of discovery and invention under one roof - ‘technoscience’

Each of these labs were pursuing a broad vision of the future through science and technology research, which lay outside of the mainstream.

• ‘One policy, one system, universal service’ - Bell Labs. The vision of a single telecommunications network for the United States.

• Computers are destined to become interactive intellectual amplifiers in the world universally network worldwide - Licklider’s ARPA-PARC vision. Explored in more detail in my blog here.

• The understanding of cellular life would come from understanding interactions between molecular structures - early MRC LMB. This was the break with conventional biochemistry that occurred in the 1940s and 1950s. 

I’ve written more about visions in the context of these organisations and the early ARPA here and here. Michael Nielsen has a great short paper on it here that is well worth analysing. 

Alan Kay describes the effect such a vision has on a group of people:

“A great vision acts like a magnetic field from the future that aligns all the little iron particle artists to point to “North” without having to see it... The pursuit of Art always sets off plans and goals, but plans and goals don’t always give rise to Art. If “visions not goals” opens the heavens, it is important to find artistic people to conceive the projects’’

There is important analysis to do here about what kind of things constitute a good vision for this kind of laboratory. ‘Curing cancer’, ‘netZero’, and ‘solving antibiotic resistance’ are all very laudable goals, but these are not quite visions in the way the 3 above are. In any report we do on this, the details of what made these such powerful visions will be very important. Otherwise I suspect you will end up with very generic visions that don’t ‘align all the little iron particle artists’ as Alan describes it. Something around ‘converging opportunities to develop both inventions, discoveries, and theories in the same technological space’ might work but it’s off the top of my head and might be duff/need improvement. 

The laboratories then brought a diversity of skills and approaches to bear on pursuing these visions. This is a point Gerry Rubin made in 2005. But it also extends beyond the examples he gave. For example, Xerox PARC had psychologists, programmers, engineers, biologists, and a range of other skill sets working on Licklider’s vision. 

These laboratories brought cycles of discovery and invention under a single roof - ‘technoscience’.

Going back to the origins of what we now call science, for example in the early years of the Royal Society, we find figures like Newton, Wren, and Hooke, who spanned differing fields of discovery, engineering, architecture and other areas, without a clear dividing line between these pursuits. The same of course is true of figures like Leonardo. Today’s UK system is predicated on discrete categories of discovery, invention, and innovation occurring in a siloed, linear way. 

In each of these laboratories, a range of discoveries, inventions, and theoretical advances occurred, with the relevant skillsets working side by side, having lunch/coffee together, without clear boundaries. ‘Cycles of Invention and Discovery’ by Narayanamurti and Odumosu, and ‘How the US lost its way on Innovation’ by now-ARIA CEO Ilan Gur, make similar points about how the combination of discovery and invention under one roof has been deprioritised, and how we need new approaches to seize the opportunities this presents.

The phrase technoscience, used earlier, comes from the title of the follow-up to the Narayanamurti book, titled ‘The Genesis of Technoscientific Revolutions: Rethinking the Nature and Nurture of Research’. Both are worth reading in my opinion.

The phrasing of ‘cycles of discovery and invention’ under one roof comes via the brilliant Cambridge academic Eoin O’Sullivan (@DrEoinOSullivan, who suggested it having read the original version, and recommended the book above. H/T to him and Rob Miller (see below).

(There is an excellent ‘roots of progress’ piece by Jason Crawford on how we need a career path for invention worth considering for the report, highlighting how Invention has fallen between the cracks of academia and business in our organisational/incentive/career structures - highly recommend it! He presented it first at the Metascience 2022 workshop this framing was trailed at)

3. A non-academic research structure

The third key element is to me where the most important elements of this story are, but they are critically dependent on the first two main points regarding funding setup and organisation, for reasons that can be explored in the fuller report. You can’t have the setup that follows without the earlier features. 

The way in which researchers were organised, incentivised, and evaluated in these laboratories was very different to a conventional research environment today, and I’m not aware of anything outside DeepMind in the UK where these features are present clearly (DeepMind is not exactly like this, but it has many features). Time and time again the memoirs and accounts of these labs describe the culture they had, and the incentives researchers had, as absolutely critical. 

The first feature here is that all researchers were evaluated internally, based on a ‘coal face’ view of their contributions. 

Researchers contribute in many ways. Some produce lots of papers. Some produce none, but are vital sounding boards. Some produce a paper every 10 years, but change the world. Some produce dozens of papers that collectively make a difference. Some are brilliant at spotting the latest hot topic and enacting the right plan quickly, some, like Sanger, are long-term meditators, with decadal periods of ‘nothing to show’ then a nobel prize.

As Brenner writes: 

“I think one of the big things we had in the old LMB, which I don’t think is the case now, was that we never let the committee assess individuals. We never let them; the individuals were our responsibility. We asked them to review the work of the group as a whole. Because if they went down to individuals, they would say, this man is unproductive. He hasn’t published anything for the last five years. So you’ve got to have institutions that can not only allow this, but also protect the people that are engaged on very long term, and to the funders, extremely risky work.” Sydney Brenner

In these labs, researchers’ funding was tied to their presence at the organisation. If you were there, you had support. This is very different to most situations today where people get positions, then have to get different sources of funding. A given research ‘lab’ in a department could have well over a dozen different funding sources.

This internal research evaluation was possible due to full internal funding the institutions had, with the external evaluation occurring of the group as a whole. Not only did this allow researchers to focus more on research, it allowed risk to be spread across the group in a portfolio fashion so people could engage in riskier work, and allowed for them to be evaluated by people who knew their work best. This is shown on the left in the diagram below.

On the right is a more typical approach, with each researcher being pulled in a different direction by different external funding criteria, in an atomising and soloist manner preventing the emergence of a vibrant community of people pulling in the same direction. Each line represents an individual source of funding, illustrating the many different pulls and incentives inside a given department.

The second feature is the use of shared resources and technical support. 

Resources were shared between researchers in the laboratory building (left in diagram below), rather than existing in siloed labs as in most departments today. This had many positive functions including ensuring that anyone, however junior or new to the lab, could access the requirement support and prevent the need to build personal ‘empires’. This was a crucial ingredient of the ‘empowerment of the young’ explored below. It also made the very small group sizes (see below) possible.

This contrasts with the situation in research today (right side), where siloed labs accumulate equipment and resources, empowering seniority and the building of research ‘empires’, promoting a winner-takes-all mentality and also promoting ‘CEO’ scientists, who can navigate the system well, over ‘Chief Scientific officer’ scientists who are creative and motivated by doing experiments and analysis with their own hands. 

The third feature here is that researchers did research, at the bench. 

Principal Investigators today, who are nominally the ‘scientists’, have a wide range of duties such as writing grants, giving research lectures, teaching, administration, managing research, editing papers, reviewing papers etc etc, whilst the graduate students/postdocs are the actual researchers.

In these labs, those doing the research got the credit for the research they did with their own hands. At the LMB, Fred Sanger did the research for his second Nobel Prize with his own hands, whilst fellow Nobel Laureate Max Perutz continued doing his own research, again with his own hands, well into his eighties. 

This contrasts strongly with the modern situation in which principal investigators are in practice administrators. Several recent nobel prizes have been given to people who in reality had little to do with the research itself. This current system therefore  strongly incentivises people to build empires for themselves, employing dozens of people working under them, taking credit for their work, to maximise their chances of winning the big prizes and their prestige. Ie, the optimal path today to being a ‘great scientist’ is to become an administrator.

Brenner writes of this scathingly:

“I’ve never believed in these group meetings, which seems to be the bane of American life; the head of the lab trying to find out what’s going on in his lab…… Nowadays, most of the people who say they’re in science aren’t really in science. They’re in something else. They’re in the management of science. There is something I call 3M science - money, machines, management...And these people do believe that everything can be solved by what the Americans call process. That is you simply have to have a way of doing this, then apply the resources, and you will get the answer. They’re only challenge is: will I be awarded good points? Will I get promoted? Will I be able to survive in the economy of science? In a funny way, like that Kafka question, one might ask does science create the scientists, or do the scientists create science?" here and here.

Behind the scenes a lot of the leadership of science are critical of this too, but of course few will speak out because their position within the prestige structures would collapse. So they all nod along as a subset of professors employ twenty or more ‘trainees’, few of whom they see for more than a few hours a year, whilst picking up the prizes for the work those ‘trainees’ do. I think this is an exploitative system.

Fourth, the notion of a research group was extremely nebulous, with these environments resembling a ‘community of junior fellows’

Today a department at MIT or Harvard will have some labs with 50-100 trainees in it, working under a single ‘principal investigator’. 

Contrastingly, In PARC/LMB/Bell, the notion of a group was very nebulous, where it was present at all. There was, at the LMB, the notion of a ‘lab head’, but even that was very informal and largely they served as a recruiter for what would now be called ‘junior fellows’. Labs at the LMB maxed out at 6 people, as the upper limit, usually smaller. At Bell/PARC, labs were essentially absent. Rather, groups emerged in a self-organising way, as opportunities on the ground arose. Bell labs technical staff had a technical assistant, making a group size of 2, whereas this was totally absent at PARC where there was not even a pretence of a group. 

These two approaches are compared in the slide below:

This contrasts with research today, where it is very difficult to compete with a very small group in a conventional research environment if the field you are working in requires cutting edge and expensive technology. Lab size has become a marker of prestige and distinction.

This is the most important point to me of the whole thing. My own conviction is that the principal investigator model of organising research is a core root cause of a wide range of dysfunction in modern research, and is an anachronism not well motivated by the history of science. The recent trends toward the Principal Investigator becoming a bureaucratic position focussed on accumulating grants and power simply reinforces this belief. 

Here is a visualisation I quickly made for my talk of how a community of individual fellows (circles) formed groups over time in a self-organising way, with collaborations (lines) forming dynamically over time based on local information, without the need for apply externally for grants: 

This of course crucially relies on the features described above, with that group of fellows having a vision in common, having shared resources to work with, having internal funding so they didn’t need to each apply to separate external grant bodies, and having this be factored into their evaluation via local, ‘on the ground’, review.

If these ‘fellows’ were all pursuing completely independent research agendas, this self-organising ‘living organism’ [Kelly, Bell Labs] of research would not be possible. 

Hence all the points above feed into this image of science as a bottom-up, collaborative process of junior people working together to fulfil a vision.

This ‘small groups’ mindset has subsequently been vindicated by research suggesting that small groups are particularly found at the origin of truly groundbreaking research:

There was extreme focus on getting the very young people to work in these laboratories as these ‘junior fellows’. The oldest researcher at PARC was 30 years old. Brenner has highlighted how the continual arrival of the world’s most promising molecular biology talent was a, maybe the, key driver of the LMBs success. At Bell, the average age of a member of technical staff was 37. 

Note that neither PARC nor Bell labs had anything resembling postdocs (this changed in the 1990s at Bell). These were researchers, not ‘trainees’.

Fifth, the role of senior people in these environments is very different to modern research environments.

The role of senior people was as ‘managers who didn’t manage’ or ‘research impresarios’, doing a ‘service to the community’ (Alan Kay). Today, many senior people are trapped (or choose to be trapped) in a system where employing as many junior people as possible working for them is profitable to their own objectives, especially in the US system but also in the UK. 

There was no route in any of these institutions to build a large group working for 
you whose papers you put your name on. To do that you had to leave. Some of the most impactful researchers, such as Information Theory pioneer Claude Shannon, never became managers at Bell Labs. Shannon’s job title whilst at Bell Labs was the same as the day he joined - ‘member of technical staff’.

Summary

This is loosely summarised below:

Below is a summary in the form of a contrast with the ‘postwar mainstream’:

Some of the reasons for choices of what features were key were also informed by studying what went well and not well at Gerry Rubin’s Janelia, which I’ll blog about later.

Note that for each of these features above, there are some edge cases and exceptions. For example, Bell Labs also had significant top-down directed work alongside the more open ended research. This is something to explore in more detail in a fuller report.

For each of these points, there is much more substance in the 2019 notes I wrote on this from studying the environments, with input from people who had lived and worked in them. This is brought to life much more with stories and anecdotes from the labs - many of these are in the footnotes of my blog here.

Part 2 - Next steps

There is a possibility far enough above P=0 that some of this could be made to happen that I’m going to invest some more time into it. Its low probability, but high impact if it works. To achieve this, first this and the earlier two documents (vision doc here and research notes here) need to be integrated together, improved a lot, put into concrete policy steps with a plan etc. There is also Rob Miller and Eoin O’Sullivan’s related ‘disruptive innovation labs’ pitch, and some other unpublished proposals floating around.

Over the next few months we are working to develop a larger report/series of reports focussed specifically on the need for a new network of laboratories, combining historical analysis, policy, and practical politics of how you might make something like this happen.

We’re growing a network of people who might be interested in how we could set up a network of laboratories like this to give input to the report. Hopefully by the end of the year/beginning of next we will have a solid range of documents and networks to pursue this through. There are some promising signs opportunities may be opening up - there are various things I can’t share publicly yet.

There are rumblings of support for a UK national lab network in various areas, and the Nurse review softly called for a rebalancing of our institutional landscape. But the status quo default is always to set up a ‘university department without teaching’, not learning the lessons from the likes of these labs and also places like DeepMind. To shift this will require a polished report laying out the case for doing things differently in substantial detail as a starting move.

I’ll be working on this october/november.

Some questions to ponder:

• What alternative framings could be used to explain this? 

• What graphics and illustrations could convey the social dynamics in these kinds of labs? I have a strong intuition that diagrams/animations are the best way to convey the differences with conventional research organisations, but I am no artist despite my wishes…….

• In a scaled network, what features of such labs should be kept constant, and which others could vary or would need to vary field by field? For example, there is an argument, right or wrong, that biology requires slightly bigger group sizes than the 1+1 Bell Labs group size.

• What's the ‘minimum viable product’ worth pursuing, below which it won’t get sufficient shared brand recognition? Is this even a sensible question/criteria? Ie, if you only make one of them, you can’t take much risk, whereas if you set up a selection of them you can deal with risk at the portfolio level.

• Which senior figures could be trusted to be part of this and not turn into their own personal fiefdom doing their own projects? ‘Mentors not micromanagers’. 

• What is wrong with the analysis in this and the other docs? What are the best refutations/counterarguments to this kind of approach? What would falsify the arguments made in this set of documents? Constructive criticism is very welcome.

• What name to give to such a network? ‘Disruptive Innovation Labs’ is good but unfortunately ‘innovation’ is not what these organisations did: they invented. The name Lovelace was chosen as it enshrines the constant search for creative promise, not prestige, and her vision-oriented thinking re-computing. Can we come up with a short name that captures the essence of these kind of environments?

• Who could be the ‘CEO’ of such an effort? I think there needs to be a single person in charge or it will die through committee/consensus etc. Maybe I’m wrong, but the pull of the status quo is so strong you need an arbiter who can continually ensure it isn’t drifting back into a fancy version of simply replicating the strengths and weaknesses of the status quo, instead of structurally diversifying.

• How can we help people escape ‘plato’s cave’ - most people want an improved version of what they deal with everyday. How can we make real to people what it was like to work in an environment radically different?

• What other questions need developing/exploring/answering as part of a report on this?

A cautionary tale… when I tried to bring together a group of senior figures in a research field to work on a lab pitch like this in 2020, two of whom had worked at Bell Labs in the 90s, what came back was a group of principal investigators each having 12 postdocs working for them - ie, the opposite of the mechanics of these labs actually worked. This was both unsurprising but also depressing. Such an endeavour would need a ‘trustee board’ of people from these environments who, throughout their post Bell/PARC/LMB careers, stayed true to the vision those labs embodied, and didn’t, as someone from those labs said to me, ‘sell out’. These would be ‘guardians’ of the vision, lending their credibility to the endeavour and pushing back on this kind of behaviour.

One of the challenges here is that so many of the assumptions of how research is organised today is broken by these examples - so the language policy debates are framed in ways that don’t fit. ‘Discovery vs Invention’, ‘Basic vs Applied’, ‘Trainee vs Principal Investigator’, ‘Private vs Public’, in key ways each of these breaks down when you look at these models. This makes it hard, at least for a dunce like me, to find succinct ways to convey what are really the common elements of a range of stories and anecdotes about how these great labs worked.

Again, setting up something decisively different to and complementary to the existing system is extremely difficult. ARIA was extremely difficult to get through intact. Physical labs are harder than ARIA. But hopefully that was a useful training  exercise. And it's also shifted the debate a lot and put applied metascience on the map in the UK. ARIA has brought some like minded people interested in new science models into the orbit of real institutional power in a way that wasn’t the case before.

My public email is jamesphillipspublic@gmail.com - I might take a while to reply, apologies if so. 

I have found some people who have interesting takes on what kinds of things such institutes could work on, but this remains a scarce resource. If you have ideas, let me know :) 

Same credits as the previous Lovelace blog, + Sam Rodriques who has been thinking a lot about these kinds of questions: “Credit to the brilliant artist Evie Kitt for the logo design of Lovelace. Credit also to Adam Marblestone, Chiara Marletto, Alexey Guzey, Virginia Rutten, Gaurav Venkataraman, and my brother Matt who gave significant time in 2019/2020 discussing and helping with the thinking around this and reviewing and improving the doc. It was also kindly reviewed by a number of people who had worked in the institutes we wanted to learn from, such as Eric Betzig and Alan Kay.”