Science and Technology Committee

Tuesday 8th September 2026

(began 1 week ago)

Share Debate
Watch Live
Skip to latest contribution
This debate has concluded
10:15
10:15
None Portrait Speaker 1
-

Sure. Thank you very much for the question and the opportunity to meet with you today. I'm Kathleen Fisher, as you said, I'm the CEO of Aria. I've been in post since February when my predecessor, Ellen met with you. Aria was in its infancy. And at this point, Aria is actually starting to deliver on its promise, which is really exciting. At this point, we have more than 300 creators who are working on various projects across 15 different programs, and we are starting to see really exciting results coming out of those programs that I'd be happy to talk to you about in more detail throughout the time that we have together, my time as chair, as CEO, sorry, not chair will be marked by growth and execution.

We will be getting the flywheel of RA spinning at full velocity to deliver on the mission and promise of Aria. The will be spending the full £1.2 billion of the money allocated to Aria in the most recent spending review to deliver on the next cohort of program directors, the next cohort of activation partners, and multiple seed opportunity seeds, as well as seeing the cohort one and cohort two programs to possible extensions and many exciting results.

Of course, Aria's success isn't measured by things like number of creators under contract or amount of money spent, but on results it will, you know, take a while to get the kind of results that Aria is set up to achieve. The kinds of decadal results that DARPA is famous for, the kind of things that Aria will be measured for are, I would say, four different kinds, sorry for three different kinds of impact. The first is scientific breakthroughs, scientific miracles that change the way we think about different kinds of concepts, breakthroughs that change the conversation.

And of course, many of our programs will fall short of the North Star aspirations that we have for the program. By setting those aspirations really, really high, we often are able to drive work further than ever anticipated. But if we fall short, we still can achieve amazing things in falling short no matter what, we will advance the cause of science. And sometimes the fastest pace to a breakthrough is by a well understood failure. The second kind of impact is societal impact or economic impact, and that is one of Aria's most important kinds of work.

It's crossing the valley of death, which is, of course, hard anywhere and at any time. We take that extremely seriously and we plan for it from the very beginning. When we start thinking about a program and thinking about the scientific impact we're hoping for, we think about what happens if we get this scientific impact. How will this translate into, into, into societal impact? And what do we need to do to make that societal impact happen as quickly and as expeditiously as possible, so that we don't have to wait 50 years for that, that impact.

To that end, we just hired a we just created the post and hired a chief translation officer. So there is someone on the executive team whose full time job is focusing on crossing the valley of death. And the third kind of impact is a change in mindset. Aria is a beacon of hope in a time that is somewhat filled with pessimism. It's part of the reason why I wanted to come to Aria was because I saw it as such a beacon of hope.

The creation of Aria represented a commitment by the UK government to commit to not simply reacting to the future, but driving the future and helping to shape it. I spent seven years at DARPA, three years as a program manager and four years as office leadership, spending time doing the kind of high risk, high reward program director, manager driven projects, each of which had a North Star, a very ambitious North Star goal. I've seen how effective that style of research can be, and I was really impressed, blown away by the audacious way that the UK set up Aria to succeed in a very similar style, and I was honored and delighted to be chosen to be Aria's CEO .

And coming here. And part of the reason was because I know I've seen and know how strong the UK Science and Technology research base is. It has the core to be able to drive similar kinds of success. And everything that I've seen at Aria since coming as CEO only reaffirms that base that Aria is on track to deliver that same kind of success for the UK and the world, if left with the same kinds of authority and funding that it has seen to date.

Thank you very much for giving me the opportunity to come here and talk to you today. I look forward to your questions. Well,
10:20
10:21
None Portrait Speaker 1
-

I'd be delighted to. It's hard to choose. I love all my children, but in the course of today, I hope to be able to talk about many of them. I'll start by talking about the Forecasting Tipping Points program, because it allows me to make a couple of different points at the same time, but I'd be happy to follow up and talk about any of the ones that you, you know, any of them. The Forecasting Tipping Points program is exploring using developing sensor technology, low cost distributed technology and models to look at tipping points in climate change, specifically the Amoc current in the North Atlantic and glacial melt in Greenland, to hopefully develop an early warning system about possible climate change that could have substantial impact on the climate of the UK.

Whether or not that North Star comes true, part of what they're developing is low cost sensors, one of which is a small company in Plymouth called ocean that is developing autonomous solar and wind driven robots called Sea Stars, that instead of having one really big ship that monitors the ocean current, a whole fleet of them that could, at very low cost, be deployed. Ocean originally had like one of these sea stars, and now they have deployed five.

And the plan is in a year, they would have 50 of these to be able to be deployed and monitor the the , the subpolar gyre. They are now deployed and being able to receive on a, a minute basis, instructions that are being derived by the models to where to go and what sense, what measurements to retrieve, to send back to inform those models. That company was subsequently contracted by NOAA, the National Oceanic and Atmospheric Agency of the United States, and asked to.

Would they be willing to try to monitor the. A hurricane to get the. The conditions of the eye wall of the hurricane and the CEO of that company, Anita Lavarack, thought about like, do we want to do this? Like, that's super high risk, like trying to deploy these meter big vehicles during a hurricane. Like the odds of failure are pretty high. But she's amazing and was like, let's go for it.

Like, let's worth the risk. So they, in fact, deployed their sea stars during. And it wasn't just any hurricane. It turned out to be a category five Hurricane Humberto. And they succeeded at monitoring and transmitting back the conditions during that category five hurricane. And this is. And they subsequently, they now have contracts with like five different governments to monitor various kinds of sea state conditions. This is an example of the way Aria works and the kind of benefits that Aria can deliver.

So even if the the Forecasting Tipping Points program falls short of its North Star and we we're still on track to potentially deliver the North star of the program. This kind of small company could grow and deliver enormous technology for monitoring ocean conditions for all sorts of different kinds of applications. So it's an ancillary benefit. And then you asked about the ecosystem. One of the ecosystems that we have are activation partners. So Aria doesn't give funding to our creators and then sit back and watch.

We actively monitor the progress and work to remove friction throughout that process of scientific discovery through to hopefully massive economic benefit. And yeah, one of the kinds of activation partners we have is Emoto Designs, which is a small company in Sheffield. They came to us and said, you don't know this yet, but one of the kinds of friction your scientists are going to run into is they need bespoke engineering. And they were really right that we need bespoke engineering.

They're providing, at this point, bespoke engineering to more than 35 of our creators. And one of the kinds of creators was, again, on the forecasting tipping points. It turns out they needed batteries that work in Greenland. Greenland has temperatures that go down to -50 degrees, and normal battery technology doesn't work at -50 degrees. Those creators talked about the normal process that they would use is applying to a funding agency for a grant to pay for bespoke battery technology.

That process would take six months. Then they would. If they got that grant, they would then go and put out an RFP to find a company that could do engineering for them, and then they would iterate with that company to create a battery. That whole process they estimated at best would take 18 months. Instead, they had a conversation with Emoto designs. They iterated three times, and they produced a battery, a modular battery pack that would work in -50°C in three months.

So that removed a massive amount of friction from the their effort. But also it works for all the other teams that are working on low cost sensors for Greenland. So there are potentially 11 other teams that benefited from that same effort. So that's an example of an activation partner and the synergies that we are seeing. And I've already taken up too much of my time. So I will defer talking about other children that I love in Aria to other questions.
10:26
10:26
10:26
None Portrait Speaker 1
-

Yeah. So we track progress in lots of different ways actually. So the first one starts with the program directors themselves and the program creation process. When we design the program, we design in what measures of success, what we think they look like milestones or metrics and kind of what we think success should look like. And then and this is of course on a program by program basis, because like what success looks like in forecasting tipping points and what it looks like in an AI safety program and what it looks like in a, in a precision neurotechnologies program are all completely different.

So it's program by program basis. So when we design the program, we set up milestones and metrics of like what we think it's going to look like. And then the program directors meet with the creators on a quarterly basis. So they're talking to them all the time and like, how are you doing on the, on the science? How are you doing on, on translation kinds of things, although it's kind of stacked. So you first do the science and then you do the translation, but you're kind of thinking about both all the time.

The we track red, Amber green on how we're, how we're doing towards those. Sometimes the science might not come out exactly the way you expect, but you can pivot the science towards, you can change the science a little bit and change red to amber or amber to green as you're going. So that we don't say like, if the science isn't doing exactly what they set out in the proposal, then that's a failure. Like we're well aware, like we're doing very, very hard science and you can't predict exactly what's going to happen.

So there is room for manoeuvre. So the program director is constantly monitoring what's happening and, and working with the creators to shift as appropriate, while keeping the North Star goal of the program in in mind . At a annual cadence, the program director reports out to Aria, where the program is on on, on track for its metrics and milestones. At that point, we bring in outside experts who, both on the science and on the translation side, who look over the program materials and provide guidance to to the agency, to me and the chief science officer and the chief translation officer.

Like, is this still worth doing or should we be pivoting? The program directors have been quite proactive, and I've been really pleased to see, coming in from DARPA that program directors have been terminating projects that are not on track, and the agency has been either adding resources or taking repurposing resources that are not on track. The safeguarded AI program, for example, determined that one of the tracks of the program, which was about verifying AI itself, the.

Statistical AI, was, was not going to be possible given how fast AI is moving and repurposed that part of the funds towards using a combination of AI and formal methods to verify infrastructure instead, which is a good repurposing of those funds. So we're basically regularly looking at the work that we're doing to make sure that the funding is is appropriately allocated.
10:30
10:30
10:31
10:31
10:32
10:33
None Portrait Speaker 1
-

Yeah. So it's it's 3 to 5 years. So it's not always three years. And I think what's really useful about that is that it gives you it doesn't continue by default. It stops by default and then you can continue it. So I think, you know, illustrative example, again from DARPA is DARPA invested in artificial intelligence for 60 years, but it didn't set out to invest in AI for 60 years.

It invested in AI for 60 years because over and over and over again, different people continued to decide that investing in AI was a really good choice, even though for most of that time there weren't massive returns on AI. So Aria does have mechanisms to decide to double down, to add investment, to continue to, to reinvest in areas. But those reinvestments would be a decision that this was still a good area for Aria to invest in, that it was still something that wouldn't happen without Aria's investment, that other funding agencies, that other sources of funding would not pick up the ball and run with it.

So I think what's really important is that the funding doesn't continue by default. It stops by default, and Aria has to make a proactive decision to continue the funding. And one of the things I think that's useful is last two weeks ago, three weeks ago, I was at New Castle talking with a researcher there who was again working on the forecasting tipping point. He had a a sensor that could detect moisture, which he originally was using to detect moisture for farms, but was repurposing to detect glacier melt in Greenland.

And he originally wrote the proposal for Aria, thinking it would take three years. And he's accomplished in a year for Aria, what he had planned to do for for three years. And part of that was because he originally planned to work 20% time. But Aria asked him, can you only work 20% time? Or if we could fund you, can you work 80% time? And he's like, sure, I'd love to work 80% time.

And we talked to his university, and his university was fine with him working 80% time. So working for Aria can compress the amount of time and things like the example that I talked about with the emoto and the bespoke engineering and the pooling, the resources, you can go faster. Like part of it is like we're working at a, at a sprint. And so.
10:35
10:35
10:35
10:35
10:36
10:36
10:36
10:37
10:38
10:38
10:38
10:38
10:38
10:39
None Portrait Speaker 1
-

Thank you for the question. Right. We are in the middle of recruiting cohort three, and it's important that we recruit program directors regularly because program directors are term limited, so they will leave, right? So Aria will be always recruiting every two years is the vision that us we will be recruiting cohorts. I think one of the important lessons is that we recruit an entire cohort, because that will give us balance across our opportunity spaces.

I think one of the lessons of or one of the really exciting things about how Aria recruits program directors again, is as a cohort so that we can view a kind of whole class of candidates at the same time, which is different than how DARPA does it. DARPA recruits one at a time, which means you're making a decision. Is this person above the bar? Is this person above the bar? Which means you have a very small pool at any one time to be considering for a particular post.

Whereas Aria, I believe we currently have, we're the current class deadline is September 28th, and we have more than 1200 applicants right now. The last round we had 400 applicants. So I think that this is a testimony to how well Aria is perceived, both here in the UK and abroad, which is really, really exciting because the quality of the program director is really, really important for the success of the agency and the and the programs.

So in terms of your question about what are the qualities we're looking for in a program director, there are multiple. One of them is technical depth. We need. Because what they're doing is managing science breakthroughs. They need to understand really good science and what really good science looks like. They also need to be technically broad because all of our programs are interdisciplinary. And so you need to have an appreciation not just of one particular discipline, but be able to talk across disciplines.

You also need to have good technical judgment, because people will come to you and try to sell you stuff. And so you need to be able to appreciate like, is that sales pitch well grounded or is it snake oil? So technical depth, technical breadth and technical judgment. Then you need to also have an entrepreneurial spirit because Aria is not in science for science sake, right? Just getting a scientific breakthrough is not enough, right? Aria's mission is to make the UK and the world a better place through scientific breakthroughs.

So like making the world a better place requires getting across the valley of death. So our program directors have to have that entrepreneurial spark. And so we're looking for that in some way. And then you need to have really good communication skills because as the program director, you don't actually do the technical work yourself. You explain and convince and cajole other people to do the technical work. And that requires really good communication skills.

Similarly, leadership skills, because you have to have those hard conversations with the scientific teams that like your work isn't, you're not doing what you need to be doing. You have to be able to fire people, right? You have to be able to motivate people to, to do the work. So like that requires really good leadership skills. You also have to have really good resilience because you have to go out around the UK and the world and explain what you're trying to do and get told by people like, that's stupid or that's really exciting, right? It's kind of like doing a PhD.

Like the first time you put together a program. That process, even if people tell you you're doing a good job, you don't believe it. Kind of like doing a PhD. You don't believe it until you're actually kind of all the way through the process. And so you have to be resilient. And then nobody comes to Aria or to DARPA with all of those skills, which is probably not surprising given the list of things I've just told you.

And so you also have to be able to take feedback. Well, because if you don't take feedback, the time it takes you to learn all those skills is too long because you only have like a three year appointment plus 1 or 2 years of renewal. And so people who are able to learn on the job faster by listening to people who are working really hard to coach you, to get up to that set faster will make a bigger difference faster.

So those are the set of skills that we are, we are looking for, and we found them in the first two cohorts. And I'm convinced that we will find them in this cohort as well. Some of the learnings first cohort was heavily academic. Second cohort was heavily entrepreneurial. We're looking in the third cohort for a balance of both academic and entrepreneurial. The first cohort, we allowed people to be 80% time to accommodate academics.

The second cohort, we made 100% time at Aria. Third cohort will go back to letting people be 80% time to allow, again, people to be academics. That is partly because academics often have PhD students that they don't want to abandon. And so it is definitely I've seen this at DARPA as well. It's it's definitely doable to be an academic on 80% time and be effective as an Aria or a DARPA program manager.
10:44
10:44
10:44
10:44
10:44
10:44
10:44
10:44
10:45
10:45
10:45
10:46
10:46
10:46
10:46
10:46
10:46
None Portrait Speaker 2
-

Have him try again.
10:46
10:47
10:47
10:47
10:47
10:47
10:47
10:47
10:48
None Portrait Speaker 1
-

Yeah, so there are actually lots of different kinds of program structures that can, can work. And the right program structure depends again on the what the program is trying to accomplish. So, you know, one structure is you allocate all of the money kind of up front and you then actively manage as the program goes along, cutting the teams that are not effective. Sometimes you allocate at the beginning and because you assume you're going to cut certain teams.

So the programmable plants program, for example, took that took the first strategy there because plants, it turns out, grow really, really slowly. And so she anticipated that it was going to take no matter what. It was going to take a long time for the plants to grow. And so doing something faster was not going to be helpful in the enduring atmospheric platforms program, where the teams are going to have a relatively expensive. Costs to send various kinds of vehicles up into the stratosphere to set up communications.

The program director wanted to take approach where we invest in a whole bunch of different approaches, kind of overcommitting at the beginning, and then cutting the teams, the ones that show the least promise, which allows you to put more shots on goal. And that's an advantage of that approach. But it means that the teams know that they are actively competing against each other for a smaller pot, and so they won't be working with each other because they know that they're going to get cut.

Some of them, they have to get cut because the team doesn't. The program doesn't have enough budget to continue all of them forward. The Forecasting Tipping Points program set up a structure where they weren't sure which approaches were going to work. And so they set up a two phase program where the they ran the first phase for 18 months, and they're going to send out a second solicitation in the next month or two to set up the second phase of the program.

And that was appropriate because they they knew that they would be able to iterate and get information in the first 18 months that would dramatically affect what they wanted to do. In the second part of the program, the the Trust Everything Everywhere program that Alex is running is running a competition because he wanted to try out a whole like, see what was out there to, to give because that space is moving so fast and there's so many players, it didn't seem that he would be able to get all of the people on contract that he wanted to be able to try out.

And so by running a competition, you can throw it out there and see who all will come to play. The challenge of a competition is it's extremely challenging to get the rules exactly right so that you are rewarding what you want to be measuring and not inadvertently rewarding something else. And you need to be very careful about setting it up so that you can't. The the players can't cheat in a way that you don't want them to be cheating.

So those are all interesting, different structures. We recently hired a chief science and technology officer who comes from spending significant time at DARPA, who has had like a large amount of his career working on different program structures to help, particularly cohort three, be very sophisticated in program structures. I think one thing that he'll be particularly looking at is setting up some programs. I talked about kind of phase structure and funding structure. Another really interesting structure is the technical area structure.

And one thing that DARPA Aria has done quite a bit on is. Social impact kind of structure and an adoption kind of structure of, of how to, how to, how will this technology be received, which I think has been really interesting. Another technical area that is very common at DARPA is an evaluation team, because often how you evaluate cutting edge science is itself a research area, and it provides program directors with a really interesting lever arm.

Like if you're creating challenge problems, you can make challenge problems more or less difficult, and you can choose challenge problems to facilitate translation or adoption. And having a separate technical area that is responsible for managing evaluation gives the program director another lever arm to to move. And so the new Chief Science and technology officer, who will be starting on October 1st, has deep expertise in this kind of structure choices. And he will be working with cohort three in those choices. Thank you.
10:52
10:52
10:53
None Portrait Speaker 1
-

Thank you for the question. I appreciate that Aria has indeed a a mandate to bring economic benefit to the UK. And I, I think that Aria is is indeed a value creation agency primarily, but I think it also has a value capture role. I think that Aria's. Aria's value creation mechanism by itself will bring tremendous economic benefit to the UK. For example, the precision neurotechnologies effort that Jack Caroline has is bringing is working on creating technology to address neuro neurological disorders and neuropsychiatric disorders that currently afflict 1 in 6.

Citizens of the UK, people in the UK and cost the NIH, NIH, National Health Service sorry, NHS £4.1 billion a year. So if that. If that program delivers technology that could, in a much more economic way, treat those patients that will deliver enormous economic benefit to the UK without having any kind of profit associated with it. So value creation can be economically impactful, both by not costing the NHS an enormous amount of money, and by returning the people who are currently suffering from those kinds of ailments to productive workforce.

That said, I agree that value capture is also an important part of what we are doing, and that's part of the reason why we created the Chief Translation officer, whose job it is to work with program directors, to work on well, to identify what the impact path is like. Every time we create a program, we also think about what is the path for translation of this program, what is the path for for social or economic impact? You know, some of them is, are things like climate change, but some of them are creating companies and driving those companies to across the valley of death and commercialization.

And how can we deliver that benefit to the UK? Our IP terms are designed to maximize that. We've chosen not to take intellectual property, because we think the best way to drive that economic benefit for the UK is to have the creator themselves maintain the the IP rights so that when the creator goes to start a company, their cap table is clean and they have the best terms to go to the venture capital companies and get that funding because they will be maximally incentivized to, to go and do that process.

We have worth multiple activation partners worked to create a climate to incentivize that entrepreneurial spirit to go and drive the commercialization through those mechanisms. Yeah, I can say more, but I will stop there for now.
10:56
10:57
None Portrait Speaker 1
-

I think that crossing the valley of death is really, really hard, and we need to make sure, like crossing the valley of Death successfully needs to be the first priority. If we can successfully cross the valley of Death, then we need to think about how to make sure we cross the valley of death in a way that maximizes benefit to the UK. And I think that one of the things that Aria is doing is working to create an environment that makes crossing the Valley of Death in the UK as maximally attractive as possible.

That's why we're doing things like the Venture Cafe with the activation partner in London and Manchester and Edinburgh to create a climate where entrepreneurs know that they're not alone, that in fact, there's tons of entrepreneurs in the UK. And that like one of the things that I've been asked by creators who are thinking about becoming an entrepreneur is like, well, what if I create a startup and it fails with the kind of the implicit notion that having a failed startup is like a kiss of death and your career is over, versus like the mindset in Silicon Valley is like a failed startup is, is a perfectly good thing to have on your resume.

It gives you credibility to start another startup. And I think that like those creator communities, the venture cafe of like creating that mindset of, if you don't try, then that's a failure, right? Like trying and failing is a perfectly reasonable thing to do. It's why we have the activation partner 50 years, who is a one of the world's best deep tech venture capital companies, working with us to run the 5050 partner, to work with scientists on how to create a pitch deck and how to be an entrepreneur again, to create a population who will become entrepreneurs, who who will decide they want to create a great company and want to stay in the UK and do value capture here.
10:59
10:59
11:00
11:00
11:01
11:02
11:03
11:04
11:04
11:04
None Portrait Speaker 1
-

To be true.
11:04
11:04
11:05
11:06
11:06
11:06
11:06
Lord Ranger of Northwood Portrait Lord Ranger of Northwood
-
-
Thank you for your time today. And just following on on Lord Grayson's sort of questioning about economic benefit to the UK. And look, as someone who spent most of my time in the private sector, I'm very supportive of collaboration between industry and government. When we're at the cutting edge of science and technology. But the announcement last week where the chair of Aria now who has resigned, was moving on to a global US firm, anthropic, but was going to retain his position as chair of Aria.

And I think that actually led to quite a bit of dismay in the UK, as we're talking about UK economic benefit, UK businesses and at this point in time, particularly in the AI industry, where growth, development, investment, it's such a hotbed of, of development that we're trying we're competing in that global race. That's somebody who had been so close to the UK's AI strategy, not only the AI Safety Institute, the AI action plan, the architect of the AI action plan, and then as chair of Aria, would remain in that position and take up such a critical role.

And that dismay, that confusion. And I say this as someone I should declare, as a co-founder of the business AI Alliance, which has several hundred UK AI startups in it, as vice chair of the AI, PPG and an advisor to some UK AI businesses, I heard this and it was felt and it was in the press and there was a genuine anger, and I think it united the political spectrum as well as industry as questioning how this decision had come about.

Now, I know yesterday McAuliffe had announced that he was resigning as chair. But what I would like to understand is what was the thinking within Aria before that resignation in terms of allowing that role to continue?
11:08
11:08
11:08
11:09
None Portrait Speaker 3
-

Did you receive any.
11:09
11:09
None Portrait Speaker 1
-

Yes.
11:09
11:09
11:09
11:09
11:09
11:10
11:10
11:10
11:10
11:12
11:12
None Portrait Speaker 1
-

Indeed.
11:12
11:12
11:13
None Portrait Speaker 1
-

Yeah, so I've talked about how Aria's mission is to make the UK and the world a better place through scientific breakthroughs. That mission on a per program basis requires a science miracle and then a crossing the valley of death miracle. So that's a really daunting mission. And it requires working with the absolute best scientists. We know that the best scientists are spread out across the entire UK. Like I've talked about Austin, which is in Plymouth, I've talked about Emoto designs, which is not a creator but an activation partner that's in Sheffield.

I haven't mentioned Touch Point, which is a small company that is on our robot dexterity programme that is in Sheffield. I mentioned the creator that I talked to in Newcastle the other day. Right. So we have worked with fantastic creators across. Those are just four that I've mentioned across, not the Golden Triangle, but we are not where we want to be. The . The Golden Triangle does have a dense concentration of world class scientists, so it's not surprising that they are well represented in our portfolio.

But as I said, we are not where we want to be, and that is why we've put in place the requirement that our program directors must do their discovery process. When a programme director sets up a programme, they kick the tires on their thesis by going and talking to scientists all over not just the UK but the world. We now have a a process by taking their programme thesis and identifying the science and technology that underlies that and cross-referencing that with the scientists across the countries and regions of the UK, and figuring out kind of where they are and requiring the programme directors to talk to a representative sample across all the countries and the regions of the UK.

I just got back from meeting with the the Dxa of Scotland and the Dxa of Northern Ireland, and I'm scheduled to go meet with the same in Wales and the Dxa of Scotland and the Dxa of Northern Ireland agreed to work with Aria to make sure that. In that process they would help us identify the. The scientists in Northern Ireland and Scotland that we should be talking to. For example, part of the challenge is that the way Aria works is very distinct from the way the traditional funding agencies work.

For example, we don't have a call that comes out on a fixed deadline like it's November. The deadline comes out now, right? Our calls come out when the programme director has finished the discovery process. And when that call comes out, there's a relatively short period from that call to when the proposals are due. So the people who are in the know have been following the the the thesis publication or the discovery process to know that the proposal is coming and have already been talking to the programme director.

So that's why I went to Newcastle a few weeks ago, Scotland last week, two weeks ago, Northern Ireland last week to meet with faculty at University of Strathclyde, University of Glasgow, Edinburgh University, Ulster University. It's. ET cetera to communicate this so that people across all of the countries and regions of of the UK know how to work with us so that we can pick the best scientists across all the countries and regions of the UK and at the same time distribute the funding more equitably across the countries and regions.

The program directors are required to do this discovery across the UK and in proportion to the scientists in those in the areas of their programme, and they can't get their program approved until they have demonstrated that they have done so. We had one programme director a couple of months, months ago say, well, I got all but one, and I'm like, okay, well, we'll delay your programme approval meeting until you're done. And then like, oh, okay.

And then they had got that one scheduled. So we're taking that seriously. It's really important for Aria to deliver on our mission. And the best way of doing that is reaching the best talent throughout the entire UK. And we are committed to doing so. And the same thing with gender applies as well.
11:18
11:18
11:18
11:19
None Portrait Speaker 1
-

I mean, that is what the engagement is set out to do is to find those those people. Another thing we are doing is refining our our data collection. Right now, our data collection is only at the prime level, which is the primes are the people who have contracts directly with Aria. But most of our contracts are in fact. The primes that have contracts with subs that are spread out. So prime contracts don't do a very good job of tracking, of tracking where our funding is going.

And so we are working on refining our data collection so that we have a much better sense of, of where Aria's funding is actually going and tracking not just the people who receive Aria funding, Receive Aria funding, but the people who apply for Aria funding, both at the Prime and at the sublevel, so that we can get a better sense of is the problem people not applying or is the problem people not getting funding? When I went to the meeting at.

. One of the comments that I heard in the audience was like, they are serious about wanting to work with us. So I think part of the challenge is people not believing that Aria actually does want to work with people outside of the Golden Triangle. So I think some of the. engagement really is important from the point of view of like, we don't just care about the usual suspects. We really do care about working with the best people wherever it is.

I also think Aria's funding model of, we cover. 100% of the costs and not just. 80% means that we universities don't have to to pony up some of their own money, which means the universities that are not so well funded are not at a disadvantage in applying for funding. And I think getting messages like. So I think the messaging is actually a really important part of delivering in the long term. I want to try those.

mechanisms as the as the next step, which is a significant step up from what we had been doing . Because I, I, I don't want a sacrifice on Aria's prime . Mission,
11:21
11:21
11:21
11:22
11:22
11:22
11:22
None Portrait Speaker 1
-

Then, you know, is it something that only like if Aria doesn't do this, it's unlikely to happen? Like those are some of the questions. Another key question that we ask is, does the UK have. enormous scientific potential in this area? So if we put out the funding call, will we get really, really strong proposals so that we can run the majority of the science and research with UK technology? But it is.

Unlikely that all of the science and technology. capability to have the program have maximum chance of succeeding exists in the UK, so we. reserve the potential to fund up to. 20% of the research outside of the UK. When we get the proposals in, every proposal that gets funded that is outside of the UK has to cross a really high bar, that this is a. capability that is not in the UK, but that.

enormously contributes or adds to the likelihood that the UK part of the program will succeed and cross the threshold across that scientific barrier across the valley of death, and do that value. creation and or cross the valley of death and produce the value, the promise of the North Star of the program. So all of the non-UK funding has a much, much higher bar that it has to cross and has a case that it will add enormous value.

Some of the times those non UK. contracts cause the non UK companies to relocate to the UK. So like normal computing is one of the creators on the Nature Computes better program. They moved to London and have. subsequently invested like. 150% of their Aria funding , because the talent that they've been finding in the UK in London has been so strong. So there .

Are sometimes
11:24
11:24
11:24
11:25
11:25
11:25
11:26
None Portrait Speaker 2
-

. When you
11:26
11:26
11:26
11:26
11:26
11:27
11:27
11:28
11:28
11:28
11:28
11:29
11:29
11:29
None Portrait Speaker 2
-

Definitely world .
11:29
11:30
11:30
11:30
11:30
None Portrait Speaker 1
-

Project based agencies. The mission is different. DARPA's mission is national security. Aria's mission is making the UK and the world a better place through breakthrough Science. The DARPA does a lot of classified work, which adds a ton of friction . That Aria doesn't have to have to deal with. DARPA budget is a lot bigger. DARPA is a lot older and has a proven track record that gives it.

Yeah, it doesn't have to prove its existence yet. Aria still has to prove its existence. The DARPA has the Defense Department as a a customer, which I think is actually overblown. I read a lot about or people talk about how well DARPA has a built in customer. I mean, having been the director of the Information Innovation Office, DARPA Computer Science Office that just got renamed Ipto, there was nobody in the Pentagon who was like eagerly awaiting the products of Ipto or I to O.

And so I spent a lot of my time as the director of I. Two o. strategizing about how to take the results of our successful program and subversively getting them to be translated into the Pentagon. So I think that translates very well into Aria, where we have to figure out how to engineer translation. I think the I mentioned earlier about the hiring process, how DARPA does hiring kind of on one off and, and Aria does hiring more in cohorts.

Aria will still have to hire sometimes in individual cases. And so the experience of like that, I had a lot of experience of hiring. individual program managers at DARPA will be helpful in the cases where we have to hire program director, program directors at Aria. individually will be helpful. All right. DARPA has another level of hierarchy that Aria doesn't have, because the scale of Aria is smaller, like the scale of Aria is roughly the size of the office that I managed at DARPA.

So like Aria's scope is like all of science, roughly the way DARPA scale is all of.
11:32
11:32
11:32
None Portrait Speaker 2
-

DARPA.
11:32
11:32
11:32
11:33
None Portrait Speaker 2
-

.
11:33
None Portrait Speaker 1
-

Aria does? And if you do, how could that happen? Yeah, I mean, I think that there definitely is and this is something that I've talked to our chief translation officer about, right? Every program. director has what we call an impact maximization plan, which is you assume your science pans out. What's your strategy for crossing the valley of death? Or like, not every program is doing. commercialization as their strategy, but like, what's your strategy for impact? And some of the programs, like they have a strategy in place, but the science is very early.

And so mostly what their strategy is just, they're just thinking about what they would do if the science pans out, but others are further along and they're like, maybe have an entrepreneur in residence who is working with the, the creators to think about putting together a business plan cases and teaching their creators that a pitch deck for a VC is not 17 slides about how cool the science is and one slide on give me money. It's much more 17 slides on what the market is and one slide on give me money.

And then if you get further than that, it's like all hands on deck. Like how do we actually like some of the cases? It might be we have to create a market like, and how do you go about creating a market? And how do we coordinate like with government to like maybe do that? And things like advanced market commitments could be very helpful in creating a market. I think things that like with the, the scaling inference lab, where the, the, that's a program that is in the AI hardware space.

And one of the challenges like , so that program started with, we are going to create novel hardware because the end of Moore's Law and Dennard scaling. We. We are. End of Moore's Law is maybe overblown, but like we are now having to create more and more unusual kinds of hardware and coordinate different kinds of hardware to be able to continue to produce more, faster and more kinds of hardware. And that program funded creation of different kinds of hardware.

And very successful so far from a pure like this piece of hardware is doing much faster, but the end user doesn't care about funky pieces of hardware. You have to put the funky pieces of hardware together into a system that actually. delivers overall system performance. And so we stood up the scaling inference lab, which is actual rack. compute that can deliver. performance that an end user could measure. And there we are partnering with the government to like where they are providing commitment that they will buy it.

So that's an example of Aria doing exactly what you were talking about. And that's a thing that our chief translation officer would be working on with program directors when they get to that point to, like all hands on deck, all parts of the .
11:36
11:36
None Portrait Speaker 2
-

. Was.
11:36
11:37
11:37
11:37
None Portrait Speaker 2
-

. Parts of the
11:37
11:38
None Portrait Speaker 2
-

. To
11:38
11:38
11:38
11:39
11:39
11:40
11:40
11:40
None Portrait Speaker 2
-

And I,
11:40
11:41
None Portrait Speaker 2
-

. Added in
11:41
11:41
11:41
11:41
11:42
11:42
11:42
11:42
11:42
11:43
None Portrait Speaker 2
-

by . Other
11:43
11:45
11:45
11:45
11:45
11:46
11:46
11:46
11:46
11:46
11:46
11:47
11:47
11:47
11:47
None Portrait Speaker 2
-

Had an . Almost
11:47
11:48
11:48
None Portrait Speaker 2
-

all of those .
11:48
11:48
None Portrait Speaker 2
-

. Does
11:48
11:48
11:48
11:48
11:49
11:49
11:49
11:49
11:49
11:50
11:50
11:50
11:50
11:50
11:50
11:50
None Portrait Speaker 1
-

And then a program on precision mitochondria. So that also would amount to roughly. £100 million investment. So Aria does do things in the life sciences. That said, Aria has, what. $1.2 billion to invest across all of science in the spending review period. And with the like. That means that although. 1.2 billion sounds like a huge amount of money when you spread it across all of science, it's not very much money.

And Aria can only do things where we have program directors. So kind of the unit of energy or deployment is program directors. So I experienced this viscerally at DARPA , where I really, really wanted to do a program on shutting down rogue AI, which seems very relevant these days, but I couldn't find a program manager to hire who wanted to, who had the expertise and wanted to work in that space. It took me like two years to recruit a program manager in that space.

And so for Aria to do work in a particular space, you have to hire a program director in that space. The question of of data and the protections around data is always really important. And you have to be very careful. about who has the data, who has access to the data. And that is something that we would, you know, always take very seriously. We have always the ethics and social.

responsibility and the, the, the normal procedures and practices about how you ethically and responsibly do work in a particular area that if we stood up a program that was dealing with big data in a medical area, we would be following those practices and potentially working on research on how to do that work even better.
11:52
11:53
11:54
11:54
11:54
11:54
11:55
11:55
11:56
11:56
11:56
11:56
11:56
11:56
11:57
11:57
11:57
11:57
11:57
11:57
11:57
11:57
11:57
11:58
None Portrait Speaker 1
-

morning.
11:58
This debate has concluded
Return to start of debate - Return to top of page