The National Physical Laboratory at 125

0:17Hello, welcome to the Naked Scientist podcast, the programme that brings you the biggest breakthroughs and talks to the major movers and shakers in the worlds of science, technology and medicine. I'm Chris Smith and today we're going behind the scenes at the National Physical Laboratory, NPL, as it celebrates 125 years at the forefront of science.

0:46the national physical laboratory or npl was founded in teddington in southwest london in 1900 which means that this year marks its 125th birthday it's a huge milestone for npl which is the uk's national measurement standards laboratory and also the guardian of the physical standards that dictate modern British industry. The world's first atomic clock was built there in 1955, and today its pioneering work in quantum technology as well as artificial intelligence are helping to drive the UK's scientific renaissance. But why was it set up in the first place? Here's JT Janssen, who's Chief Scientist and International Director at NPL.

1:29If you want to trade with a foreign country and you want to sell them something, they need to know what they're buying, and you both need to agree on what it is you're selling. For instance, if you sell a kilo of potatoes to somebody, that other person wants to know that you've really sold a kilo of potatoes. And that means you both have to agree on what a kilo is. And that sort of perpetuates through all measurements which we do. And we do loads of measurements. You do measurements of weight, length, temperature, electrical properties, current, voltage, and all these kinds of measurements of really important units.

2:05need to be agreed internationally. And that means that the UK has a national measurement laboratory, NPL, and also all the other countries around the world have got national measurement laboratories. And it wouldn't surprise you that Germany was one of the first ones to have a national measurement laboratory. Then came the UK with NPL. And very soon afterwards, there was the National Institute of Standards and Technology set up in the US. And basically, most measurement laboratories were set up as support for international trade and economy. It's interesting though, isn't it? Because although we might have had those standards emerging, there were still multiple ways of measuring the same thing.

2:44Let's take the difference between metric and imperial, for example, and the Americans still use imperial. In fact, we've ended up with rocket disasters because people disagreed over the size of a bolt that was imperial in one country and metric in another. So in some respects, it was founded on a sound principle, but it still hasn't solved all the problems, has it? 95 % of the world uses the international system or units, the metric system. And of course, there are still local standards, like we still talk about a pint of beer, but we know that a pint of beer is exactly 500 and something milliliters.

3:20All these other units are still defined in terms of the metric system. And the errors occur when people use the wrong conversion factors between the sort of traditional units, which you used to use, and sort of internationally agreed metric units. Was the whole idea about standardising in this way well received across the world though? So when NPL got set up and said, look, we need to agree on what a kilo is, we need to agree on what a second is, was everyone on the same page? The history of standards and measurements is very long. It started with the building of the pyramids by the Egyptians, and there were many local standards, there were many, many rows about standards, and also people use standards to establish their empire and thereby control the way they taxed and traded.

4:05And when the current system was set up with the Meta Convention 150 years ago, the UK was one of the countries who didn't sign the Meta Convention because it thought it would have an adverse effect on British trade and economy. Very soon, of course, they realized that if the others are grouping together around one metric system, then we really ought to be joining it because otherwise you'll start to lose out because people will not want to trade with you. And is that how the meter came to being and metrication in the UK came into being then? The UK has actually been using metric units for a very long time, even though in popular parlance we still use, of course, all the imperial units.

4:46Yeah, one wonders how stones and pounds and pints have hung on for so long nevertheless. It does seem surprising, doesn't it? What were the main priorities for NPR when it got started, apart from getting the measures right? What were the main scientific priorities and goals? The first priority was to be competitive with countries like Germany and the US and protect UK trade and manufacturing through joining that system. But very soon afterwards, of course, as a country, you want to have your own realisations of the unit. So you want to have your own copy of the kilogram. You want to have your own metre.

5:22say you want your own temperature scale because that gives you a level of autonomy and priority. So that was really in the very first instance to establish those primary realisations. When they actually first began to make those kilograms and then distribute them around the world, so you've got a counterpart in every country which has got its own kilo, did they literally make them all together, weigh them against a standard and then give everyone one to say that's your kilo so that everyone knew they were on the same kilogram scale? The first batch of kilograms was made here in the UK by Johnson Matthew, and they made a whole series of kilograms in one go.

6:00And they chose one, which became Le Grand Cay, which was basically the kilogram which was kept in a vault in Paris. And all the other ones were direct copies of this one kilogram. Over the period of the last 150 years, the kilograms which were distributed around the world came back to Paris and were compared with this original kilogram. And of course, over time, they saw that the spread of the kilograms becomes bigger and bigger. That's quite an auspicious track record, isn't it? That was J.T. Janssen, who's Chief Scientist and International Director at NPL in Teddington. The National Physical Laboratory doesn't just deal with standards in weight measurements.

6:41It's also been at the forefront of precise timekeeping too, including pioneering the development of atomic and now optical clocks that are so accurate that they lose less than a second over the entire age of the universe. It is hugely impressive. And here's Rachel Godin, who works on NPL's next generation of optical clocks. NPL has a very proud history of timekeeping. This began with the development of the world's first atomic clock in 1955 by Louis Essen and his team. What they were able to achieve was a clock which was more accurate than any clock of its time. This was a real transformation in the level of accuracy of timekeeping.

7:27How did people actually decide what a second was before we had an atomic clock to keep us on track? For a long time, up until 1960, one second was defined so that the length of a day would be 24 hours of 60 minutes and 60 seconds. So the idea was to take the mean kind of rotation time of the Earth at its own axis and then subdivide that down into the hours, minutes and seconds. And that would then give us a fixed interval for one second. although as we've seen this year the earth doesn't always spin at the same speed so we're potentially therefore introducing some inaccuracy if we go by that physical way of doing it and i guess that must have been one of the motivations to have a system like an atomic clock that's divorced from that that's exactly right and somewhere around the sort of 1930s 1940s people actually started to suspect that the earth didn't rotate at a nice even speed and that maybe it It was a little bit wobbly in its rotational speed.

8:31And it wasn't until having very accurate atomic clocks that people could definitively measure that and see that, yes, indeed, the Earth is a little bit wobbly. And it wasn't, therefore, a very good way to define a second. So between 1960 and 1967, it was recognised that, oh, dear, we've been using the definition of the second based on the Earth's rotation that's a bit wobbly. we should redefine the second to be based on something that's a much more fixed period of time. And the choice was made to define the second to be a fraction of the Earth's orbit around the sun in the year 1900. So now you have the definition based on a fixed length of time.

9:19But this is a really difficult definition to use. This really wasn't very workable. And so in 1967, the second was redefined again to be based on an atomic clock. This is now still the definition that we use today to be basing one second on a cesium atom. And it's got us away from the difficulties of needing astronomical observations to tell the time. How does that actually work using an atom, in this case, cesium, to tell time? Atoms have the property that they have resonant frequencies. You can excite a cesium atom if you shine microwaves at it. And if you tune the microwaves into the frequency that matches the atomic resonance frequency, this now gives you a wave of microwaves that's oscillating up and down.

10:15And so you can use that wave. You can think of the ups and the downs of the waves as the sort of tick-tock of the clock. So you now have a microwave that's been tuned to match the cesium's specific frequency. And the beauty of that is that it means that if you have a cesium clock in one place or a cesium clock somewhere else, anybody who tunes their microwaves to match the cesium's resonance will be tuning their microwaves to the same frequency. It's a bit like when you push someone on a swing, isn't it? When you give them a push at the right time, the swing swings higher and higher until they reach peak swing.

10:52But how do you know what the caesium atom is doing? How is that registered so that you know you have tuned your microwave pulses into the resonant frequency of the caesium to get that timing perfect? Yes, so we need to be able to see when the caesium atom has been excited. And so that comes down to being able to determine what internal energy level the atom is sitting in. And so by trying to excite internal energy levels with laser light, you can determine whether it was in the ground state or the excited state. So the idea is that you prepare the atom in the ground state, you send in some microwaves, you see if it's been excited, this is how you get the information as to whether your frequency was hitting the resonant frequency or not.

11:42Are microwaves and therefore atomic clocks based on cesium sufficiently accurate though? Because there's going to be an inherent frequency limitation with a microwave and there are forms of radiation which are far far higher frequencies and therefore presumably have much better resolution of timekeeping so why don't we go better initially this first atomic clock that was developed at npl in 1955 was based on cesium and so we've stuck with that for a while it actually turned out to be a very good choice the cesium clocks that we have nowadays you can use them to measure one second to 16 decimal places.

12:19So these are the clocks that underpin national and international timescales. So they're reassuringly accurate. But as you've raised higher frequencies, you enable you to reach better accuracies. And so this is a general sort of principle. If your clock ticks faster, it's easier to subdivide an interval of time into smaller and smaller time intervals. So if we could have an atomic clock that's based not on a microwave frequency that's ticking at about 9 billion ticks per second, but if we could have, for example, an optical frequency, so using a wave of laser light, this can tick about 100 ,000 times faster, and this is going to then unlock a new level of performance.

13:08And those will then tick at those correspondingly higher frequencies so you would get that extra precision, that higher resolution of timekeeping. That's right and these so-called now optical atomic clocks are able to measure one second to 18 decimal places so a hundred times better performance than the cesium clocks that are currently used for the definition of the second. Now we started talking about how we first define the second will we see the second change again because you've mentioned that we used to do it by how fast the earth was spinning then we were doing it by how fast the earth's going around the sun then we're doing it by the atomic clock is there going to be a fourth definition of the second or are we done no we are very much anticipating right now a new definition of the second these optical atomic clocks having an accuracy that's a hundred times better than the cesium clocks, this is pausing us to say, hang on a minute, we've got clocks that are already 100 times more accurate than the cesium clocks.

14:11So we should redefine again, the international metrology community is currently targeting 2030 as the date for a new definition of the second based on an optical atomic clock. And that would then unlock 100 times improvement in the accuracy with which people can tell the time. Whoever it was who said time doesn't stand still, they were absolutely right, weren't they? Certainly, yes. And there are all sorts of interesting problems that you encounter when you start to get to this 10 to the minus 18 level of accuracy. And I think a very kind of interesting problem is that of general relativity. So general relativity tells us that time runs at different rates in different gravity potentials.

15:03So that means at different heights above the Earth's surface, time would actually run at a different rate. But the scale of the effect is that for every metre change in height, time itself and therefore your clocks would tick differently by a part in 10 to the 16. So if we now have clocks that are able to be accurate at the part in 10 to the 18 level, it means that they can resolve a change in height of just one centimetre. So we've really got to be not only making our clocks accurate, but also being very aware of the environment that they are. If we're trying to build an international timescale where we're comparing lots of clocks together, we need to have knowledge of their height differences at a centimetre level it's a huge challenge because of course my feet are aging at a different rate than my head aren't they for the very reason you've just outlined that's right and and the the scale of that effect is um it's about a couple of nanoseconds per year that your head would be aging faster than your feet so yeah you can decide if that's a if that's a worrying amount of difference or not it's certainly a very odd concept it is indeed but that was so interesting I hardly noticed any time passing.

16:20Did you? That was Rachel Godin. The Naked Scientist podcast is produced in association with Spitfire, cost-effective voice, internet and IP engineering services for UK businesses. Find out how Spitfire can empower your company at spitfire.co.uk.

16:41Music in the programme is sponsored by Epidemic Sound, perfect music for audio and video productions. And this is the Naked Scientist podcast with me, Chris Smith. And today we are going behind the scenes at NPL, the National Physical Laboratory, which is celebrating its 125th birthday. In a minute, we'll find out about NPL's role in harnessing and safeguarding emerging AI technologies. But first, we're going to delve into the quantum world, a field that seeks to use the weird rules of quantum physics to create powerful new computers, sensors and communication systems. The National Physical Laboratory has played a major role in the development of quantum science throughout its history.

17:22Here's Chris Coggs. A hundred years ago, a bunch of really great scientists and really world-leading really laid the foundations for quantum mechanics and quantum technology as we know it today. And then ever since that time, you've had a wealth of scientists throughout the decades really try and delve in on those theories and try and understand kind of the fundamental building blocks of life. So all the changes that happen at the atomic level, how interaction from atoms to atoms work, and ultimately how life functions. A lot of those scientists you cite, they actually famously, or at least two of them, have said if you weren't baffled by quantum mechanics, then you just didn't understand it.

18:02I think Richard Feynman went a bit further and said if you think you understand quantum mechanics, you didn't understand it. So it's clearly a very hard to access thing. Just in a nutshell, when we're talking about the quantum world, what do we actually mean? At a very fundamental point, you essentially have how the energy of each of our electrons and protons and neutrons, how those energies all interact and relate to each other. And so what you can do then is you can build up technologies based upon these fundamental building blocks. And you can do some really great science that you then do with interacting them or entangling them, as we call it.

18:40And that's where really the fundamental parts of quantum computing can come from. is using building blocks of physics to do computation. Why is a quantum computer regarded as the holy grail? Why is it potentially so much more powerful than the best supercomputers that we can build today? So you can kind of think about quantum computers that they work very differently to today's supercomputers. So some of the ways that classical supercomputers work. The anticipation and expectation is that quantum computers will eventually be able to solve problems much faster than today's supercomputers, but also problems that are so hard that even running them on the most powerful supercomputers that we can think of today or hope to think of in the future, they just wouldn't be possible.

19:31And so you start to bring in problems, really fundamental chemistry problems or material science problems that would be out of reach for most computation. You also have aspects of slightly more real world applications in finance or transport or medicine where quantum computers, if they actually arrive and do what they're expected to do, should be able to unlock options for moving items around the globe or for making sure that information is secured to transfer between different people and organisations. where are we along the path towards realizing this though that's the aspiration sounds fantastic but how far are we away is it like nuclear fusion where people say it's 10 years away and it always will be or is this about to be realized or has it already been realized so i've been working in the field not too long really in comparison to to many of our other colleagues but actually in that in that time i've seen a huge acceleration of what people are claiming quantum computing quantum sensors and quantum networks and what they can do.

20:39So I would say that we are really rapidly accelerating. And because more focus is now put on quantum computing, there's a real push from governments, from industry, from academia to really kind of drive this innovation. So progress that would have been made in the last hundred years is actually really starting to condense a lot more down and you see step changes occurring a lot quicker than we have in the past. I thought you were going to say there's going to be a quantum leap. I thought you were going to go down that path for a minute. But what's NPL's role going to be in this then? Because this is very much the preserve of industry and technology to invest in order to make the breakthroughs that are going to make this realisable.

21:25Where does NPL come in? MPL were working on quantum long before it ever became quantum. If you think about some of the clockwork that we do, we built the first atomic clock in the world, and that's essentially on fundamental quantum principles. And then from there, we've just been building upon that. So we've been building technologies, we've been supporting industry and building their technologies as well. And we work throughout. So we work with academia universities, we work with industry, We work with government to help identify problems, issues and challenges that they might face when dealing with quantum technologies and help them to overcome those barriers.

22:07So one of the main problems we have right now is how to scale some of these technologies. And when we mean scale, we don't just mean, you know, make them bigger. What we actually mean is making them have the same properties and performance when you start to increase things like the number of qubits. or you might look at how you can make country-wide quantum networks, taking it out of the lab from a lab bench and really putting it into a critical national infrastructure like a communications network. Is that what you're working on right now? Yes, so we actually work with companies to help them develop their qubits.

22:48We help them to develop their sensors that will essentially be used to map things like gravity or brain imaging and brain scanning. We can work with companies to help develop their really secure communication technologies. And all of these wrapped up is really where kind of NPL has its bread and butter in measurements and standards. Chris Cox, who works on quantum technology at NPL. Well, now we're going to venture into the realm of artificial intelligence, an emerging technology that's made its presence felt in what feels like record time in almost every aspect of our lives. Sandeep Bhandari is NPL's Head of Digital Innovation, and AI is very much his domain.

23:32So measurements are often seen as this invisible utility without which things actually wouldn't move anything from basic blue skies research and development done in laboratories all the way through to large commercial entrepreneurship and those kind of activities. all of those things rely upon the ability to go back to a set of measurements to have that traceability and that immutable proof. Now it's one thing if NPL does that just for the UK but we need to trade, we need to travel, we need to innovate, invest, make, manufacture all across borders. So how do we then make sure that this system of units works across these borders and that kind of then brings us towards AI.

24:09So the underpinning quality infrastructure, the ability to measure and have confidence in the data going into these systems and then the learning outputs coming from the AI systems doesn't yet exist. So if we want to leverage all of the benefits and opportunities that AI could bring us in terms of saving lives, being more efficient, helping with the climate change challenges, we really do need to have a way of being able to have confidence in what the AI is doing. So is it that you're trying to anticipate, well, what will we need to measure? What will we need to have policies around? What will we need to have a sort of framework in order to do and you anticipate where the technology and those sorts of things are going to need to intersect is that your role i would say that's part of our role one of the key things that npl is tasked with apart from being one of the core components of the uk's national measurement system is that we are specifically asked to look at supporting industrial innovation and helping accelerate the smes the small and medium-sized enterprises within the UK as well as the large military nationals.

25:09So helping to develop industries of the future, but giving them the confidence through measurement that they can innovate, they can develop new technologies, they can do the research. So part of our role is essentially to help others accelerate and enable others to deliver impact, be it through business or by saving lives or by developing the right policies that enable these technologies to flourish in a positive way. And when it comes to things like the kilogram, you've got one of those. When it comes to the second, as we heard from Rachel earlier, well, you've got an atomic clock or two, and you can deal with that.

25:39When it comes to AI, what's going to be your standard? Or what could you show us to say, well, there's the AI equivalent of the mass that is the kilogram that you set the record by? So that's a great question. And that's exactly what we're starting to work on right now, is to develop the approaches. So some of our colleagues within the data science and AI department have already developed the first version of the trustworthy and safe AI lifecycle, which enables us to go through methodically how we would approach these metrics. We have some measurement capabilities already for AI. So recently, most recent example was helping some of the wearable device companies understand how much confidence they could have in using your smartwatch to measure atrial fibrillation, for example.

26:23So for that kind of work, not only did we get the wearable manufacturers in the room we got clinicians we also got people from the medical regulations authorities as well as then potential patients and really getting a holistic view on the types of questions and the kinds of context we need to be thinking about to explain how much confidence you can have in that device so i would say we're on the journey we've definitely started to do a lot of work but there's a lot more to be done and it can't happen in isolation through NPR. A lot of people though have said there's disquiet around the fact that this is moving very very quickly and the technology is way outpacing regulation or our ability to make guesses about where the risks and threats may lie.

27:06Where do you sit on that argument? Do you think that it is going too fast and we need a bit of breathing space or do you think actually it'll all come good in the end? So I definitely agree with it moving fast. Often when I stand up and present my last line at the presentation is by the time we finish this meeting here today there'll be two more algorithms or two more models that I need to go off and now worry about or think about. I don't think we need to take a wholesale pause. Anybody even with the best of intentions that did try to cause that pause wouldn't be able to actually stop what goes on out in the outside world be it in in the UK or further afield.

27:40So what I would think is our take on this is actually let's make sure we've got the fundamentals and the foundations correct. So what are those sector or application agnostic things that we can measure and building blocks that we can then use to apply to other ways of evaluating AI. We're very much thinking about what we need to do about the large language model side of things because that is evolving very rapidly. But I think it really, for us, comes down to breaking it down into understandable bite-sized chunks and then starting to stitch those blocks together to then get even more ability and capability to understand and evaluate AI with confidence.

28:14If we take a climate change analogy though, we're now grappling in the here and now without enhanced understanding of the impact we have on the planet measured in the multiple billions to potentially a trillion tons of carbon dioxide up there is there not a risk that if we just allow these runaway trains of large language models and so on to do their thing that by the time you catch up and work out how to gauge the risk and mitigate it the knowledge space will have been heavily polluted by the deleterious aspects of some of these models, the things they confabulate and so on. And we may then have quite a clean up on our hands.

28:53I think there's a big skills and an education piece that needs to happen. These very large language models that we're now seeing advertised on television and constantly on our digital feats are very much generalist type of applications. What we're seeing more and more of is to really leverage the power of these tools and technologies. and they are very powerful and could be very useful. In fact, they already are very useful. But the way to really harness that is make them a bit more focused. At the minute, we call them large language models. The approach we're looking at is what do small, narrow language models look like, for example.