[Music] aaar is an author futurist and investor he is the founder of the exponential view newsletter and podcast which explains how Society is changing under the force of Technology aima's first book exponential looked at how Technology's rapid pace of development
is impacting the world in this conversation we discuss where AI is going next how it will change companies the impact of AI on energy and the future of Technology's impact on climate lots more as well we do talk about investing but obviously we're not here
to give you investing advice azim is one of the most influential thinkers on how technology shapes our future welcome azim it's wonderful to be here with you two giants thank you you wrote the book in 2021 what has come true and what was wrong I do reflect on the book regularly
uh it was a quite a fool Hardy errand I think to write a book about the speed of change of Technology given the speed of the publishing industry in honesty most of it has held up pretty well because we what I was doing was identifying underlying processes the underlying process by
which Technologies get cheaper uh and as they get cheaper we start to put more money into them and so we get them cheaper more quickly that's a fundamental process and that held up so there are things in there like the arrival of drones on the battlefield and I even have a set section on the Turkish
bayor drone which was the Drone the ukrainians relied on the most in the first 3 or 4 months of the war so a lot of it held up pretty well I think there are two areas where uh where it didn't hold up as well the first was one of of emphasis so I wrote about gpt3 I wrote
about the potential but I didn't finish drawing that that line uh as far as I could I did have 30,000 words on AI which I had to cut uh and the second thing that I didn't quite get right was when I to looked at the issues of trust and polarization within Society the emerging early
evidence was that you were seeing filter Bubbles and uh you polarization and homogen homogenization um on social networks and we we drew a connection to say that that was really having an impact on people's political Behavior but within a year of the book
coming out I started to see more evidence and persuade myself that the impact of social media is actually much more attenuated than than we thought so those I think would be the two things more AI a bit less social media polarization Panic the interactions of
people in society is harder to predict than the technological Curves in some ways I guess yeah no that's absolutely right yeah and the name of your platform the exponential view maybe just for listeners why you called it that because I think it does quite neatly describe I
guess your outlook on the world and the impact technology has on society and and policy yeah thank you for that you know the the exponential uh is you fundamentally this idea of uh a smooth accelerating curve it's it's the most powerful force in in the world it's compound interest something improving by
10 or 20 or 30% every year uh and what that does is it gives you these very very surprising nonlinear uh effects U and we saw that during the covid uh pandemic when the virus would go through these exponential phases and You' talk on publicly and people would say Well
only five people got it yesterday but it was it it was two the day before that and one the day before that so we know that exponentials exist in the natural world they're not very common but in the technology World they are really at the heart of how uh the technology improved
and how the markets uh expand Cam's going to ask you I know about what the realities of the next two years are going to look like in AI but before we do that let's talk about some numbers and when things go exponential sure and I was listening to your podcast the
other day and you were talking I think about an outlook for 2024 for this year and you reflecting on some of the penetration of things like chat GPT or generative models in large corporates being at around 5% of usage in some corporates which sounds to some people like incredibly low but for you 5% is
roughly the magic number when things then tend to go very quickly is that right yes that's right I I mean 5% is uh is huge uh 5% is 300 million people which is by the way where chat GPT is on a on a monthly basis when you look at the adoption curves of products in a
market uh there's the there's the usual sort of s-shaped curve and you have this exponential phase that that that tapers off it's roughly at the 5 to 10% Mark that you normally see the the inflection um when more and more customers want to buy a product and the question is why
does that happen and um the Jeffrey Moore who is a you know tech tech strategist and historian and sort of business uh thinker uh talks about this quite a lot he calls it sort of crossing the chasm uh essentially you go through different segments of buyer Persona U
from the innovators who are probably like the three of us will buy technology just it's new and doesn't matter whether it works or not through to early adopters who willing to take a lot of risk and the technology they buy they just kind of need to know that it's
worked for someone trust you then get to the early majority and the early majority needs to see people like them buying and using the technology and that point from early adopter to early majority can happen between the 5 and 10% penetration uh Mark and that's typically what I start to look for with
with markets electric vehicle uh sales or you AI Enterprise penetration five or 6% is a moment where you would expect to see an acceleration and look that's a horis that's not a law of physics so once something gets to about 5% penetration how long does it then take
generally to go from there to mass adoption I think historically it's been sort of 15 years is is that right do you think that will hold for for AI and some of the other technology we're going to talk about today it it's historically been 15 years and that's not a long
amount of time and it was 15 years for one of my favorite examples is the transatlantic Europe to New York Route for passengers which went from sailing ships to steam ships in just under 15 years and the Titanic disaster happened in between which was terrible PR and what we've seen is that that has
compressed so smartphones were seven years and one when we look at Enterprise AI we're coming into a market where every company already has computers and smartphones many companies are on the cloud it takes one switch of a feature flag in Microsoft to provision AI to 100
million or 500 million Enterprise customers so my expectation would be that the point at which we go from 5 to 75% of this technology will be much much faster than 15 years maybe 5 to seven years if it's 5 to S years we are today two years on two years in from when it
started so right yeah we're we're quite far along I find it can be helpful you know in our line of work we spend a lot of our days thinking about AI but for those who who perhaps don't um I think stories can be very helpful kind of real real world examples the internet Chang
the world in so many ways we couldn't imagine you know we order through through our phones with a click of a button we meet our partners online um you know we now have all the world's information at the click of a fingertips that was very hard to imagine in 99 when the internet was really just starting to
take off so maybe what are some sort of tangible ways that um this exponential technology AI will kind of change everyday life I mean some that come to mind are autonomous cars driv around um you know robots potentially in our houses you know doing doing certain sort
of is is a sort of futuristic one but I think perhaps not that far away so maybe give us some color on on on how you think we could imagine this future world 10 years from now 10 years is a really mean time frame Cameron but I'll do my I'll do my best what I would imagine uh
will happen with the AI systems is that they will get progressively more robust and we will push more and more of our tasks to those systems whether the task is uh we're reading a business plan to pull out the most Salient points whether it's going through uh a thousand
customer feedback emails to identify clusters of complaints and opportunities we can already start to do those things in a um in an app based environment where I have to sort of load this these this material into chat GPT and get the results but I can see that becoming
highly highly automated and each individual having a network of autonomous agents that would be take carrying out the little tasks that they have to get get done and how many agents we might have probably into the hundreds or or thousands I mean I already have a workflow today where when
I come up with an idea uh I will pass it through a network of different AIS that have got different personalities one's a bit like a business school Professor one's a bit like a physicist uh another is is a great writer with a New Yorker profile and the three of them argue
between themselves until they get to a con consensus that my idea is strong enough you have that set up today I have that set up today so that's me using three agents plus one because there's a judge four agents and we're you kind of Year Dot of all of this um so I would
imagine that we'll have a big big constellation of agents that will support us in business and per personally so many that we'll probably have to have an Uber agent like an AI Chief of Staff the way that you know the billionaires do who helps essentially direct that work appropriately yeah uh
and and and how what does that what does that feel like I think we do have some um analoges for that because we can read the profiles of billionaires on you know on on Twitter or on the blogs or in you know the Sunday papers that's incredible I mean that's fantastic they've already
got your your agent you're going to have to give us a tip on how to set that up after this uh well i' I you know I wish it would they were completely reliable they are they work well about uh 20% of the time okay but it cost me 50 cents to do this so you know why why
not and if it doesn't work it doesn't work but but I just want to come back to this point about you know living with a load of um uh AI systems supporting us because it's not about everybody becoming a a billionaire it's about helping us navigate our human scale
World um in a way that is better attuned to us being as people and I remember the days where when I was a kid my parents would interact with the insurance company or the gas company once every year or once every couple of years that is not what life feels like now the world has become
very very complicated and quite hard for us to to navigate and I think that there's a what I've discovered in using AI in the last year or two is that it's created an abstraction layer over a lot of gnarly things that I now don't need to deal with one interesting mental
model we had a giant is talking about AI as sort of a parallel labor force and we' really seeing the power of large language models kind of filter through to healthcare education where there is a lot of this kind of administrative type type of work obviously labor is is a key
input of all economic equations how do you think that's going to impact the world positively or negatively I know that's a very big hard question uh well you know I think as any good accountant starts at the top of the p&l which is the sales so let's start with the
positives um that we we'll deal with you know you should see productivity improvements and productivity improvements should translate into higher economic growth and greater degrees of prosperity and I know it's important for for giant and for your portfolio companies and listeners is
what does this do for the sustainability uh equation and I think it's incredibly positive for the sustainability equation because often the reason do things in Dumb ways is because we don't know how to do them in smart ways and so most of us probably use LED lights in our homes
well why don't we used them 40 years ago right well because the blue LED had not been invented so we didn't know how to we didn't have the knowledge to do it and a lot of the problems that we deal with are problems of Discovery and problems of of knowledge so when we look
at the Topline uh AI question we've got these tools that help us make discover faster once they they're made and found we can commercialize them more quickly and since most of them will end up being efficient because who doesn't want to do things in a more sustainable less
resource intent intensive way it really really supports the sustainability uh Direction and Ambitions that we you know we might have and it does so in a way that is sort of economically self-sufficient as let's talk about Ai and sustainability and energy in particular so we had Mustafa suon the
founder of Deep Mind on the show a while back and he said this is going to be solved we are entering an era of abundant almost free energy and AI will get us there so it's all going to be fine do you share that confidence that AI is going to just completely transform
energy abundance and to a large degree solve the climate crisis I can agree with both halves of the sentence independently but maybe not together okay so I do think that we are on a a path towards really really radically changing the energy system and I wrote a
a piece that was in the financial times a few months ago saying energy is no longer a commodity it's a technology for the last couple of hundred years since you had a fossil Energy System energy costs haven't really come down I mean oil and gas basically cost the same it's all dependent on the
commodity markets and the the regional autocrat energy is now a technology which means there are learning rates which means things get cheaper every year the cost of sell of panels has come down I mean choose whatever number we like it's in the high 9s uh and and so
once that happens and you get to this this point where the technology is so much cheaper than the Alternatives the markets expand dramatically if you look at China today every 5 days China is onboarding the equivalent solar capacity of the entirety of hinley SE for American
listeners hinley SE is a 3.2 gwatt uh nuclear reactor the the UK has been trying to build for 47 Millennia or something seemingly going to cost 92 quadrillion dollars or anyway someone will put in the correct numbers afterwards I'm sure but but that's every 5 days and their rate of is is
accelerating and so without AI we already have a path to um let's go to the global South right the single solar panel that every household needs in order to run a fridge and lighting and some some Cooling and and charge mobile phones we already have a path to um the
sub $50 per kilowatt hour batteries I mean that that's the price for lfp Batteries coming out of China right now they've come down 50% in a year and and so energy abundance or vastly more energy availability at much much lower cost is all is already coming and we don't need AI for it okay yeah so
there's the other half of the question let's do the other half let's do the other half yeah so I I mean I think the other half of the question which is you know how does AI help or solve the the climate crisis I mean I think it does help more more than harm there are a lot
of headlines about the increasing demand demands AI are putting on building out data centers and the electrical system um and I think we have to put that to some context the energy system is huge electricity is 20% of it data centers is a small sliver of that and AI is
incrementally a small amount uh of that one stat that really stuck out to me I think is that the computational need and demand for AI is doubling every 100 days so talking about exponential that's pretty I mean maybe maybe we're extrapolating too far that that will
continue but it seems like there may be no end to the demand for these systems um and yeah I mean perhaps it's a red herring but just the numbers being invested into Data Centers do suggest it's quite quite quite quite something I think Amazon invested 50 billion in the
first half of this year into new data center build outs and they're committing to 150 over the next I think couple years quite it's it's quite a lot numb but let's let's unpick a little bit of that so there is a uh you know there is a a Mexican standoff going on between the big tech companies
because the they all believe that AI is going to be this huge thing and so that it's too the cost of not participating weapon right game theory is is too high and the first company to say they're not going to play is quite likely to get tanked by investors who sudden believe well they
don't know what they're doing so there's that Dynamic I think the second is that they really are physical and materiality constraints uh at play uh the you know the Silicon Valley uh folk are talking about wanting a 5 gwatt data center to process large language models or
whatever comes after them there are 12,000 utility scale power plants in the US and only one is bigger than 5 gaws so uh so we're talking about a really remarkable ask and it's a little bit it I I I think it's quite hard to see how those engineering Pieces come together
in a threeyear time frame and then I'd add the third thing which is that it's so expensive to run big models that as soon as a big model gets built people try to make them more and more efficient and we just saw on the day that we recorded this uh this episode uh
that quen is a Chinese uh reasoning model it's about as good as open AI 01 model uh which is the one that they launched in preview in in September but it's actually quite a small model it's you 32 billion parameters which sounds like it's a lot but it's you know in the
world of hundreds of billions see some equivalent of Moors law in sort of model size sort of getting smaller and smaller but maintaining or getting better in efficiency well because we're we're really dumb when we build them the first time right and then we just we get
better at building them and it will generally you will benefit from making your model smaller and more efficient because you can just use your existing resources more better once you do that right so there's an economic incentive to do it as well I think that's very
much where Greg Jackson for example from octopus energy is coming from when he just says we're going to get so much more efficient on the full stack everything from the compute the chips the way the data censers are run is going to get way way more efficient there is going to be clearly some level
of increase in energy demand because of data centers one of the approaches big Tech has taken is to go all in on nuclear so it feels like a nuclear Renaissance is coming uh we're looking very closely at that at Giant both fusion and fion what's your take on
nuclear and the world that that plays well I love how um uh Fusion companies have rebranded fishing as nuclear and called themselves Fusion first of all uh that's a little you know nice little uh bit of marketing uh patter you know we have two types of of nuclear power we
have nuclear fishion which is the splitting of a a heavy atom of uranium uh and in that uh or or plutonium perhaps and in that that split some heat is released and there's lots of radioactivity and that is the one that we control around the world the Chinese and the Koreans are very good at
building them the British have forgotten how to do it the French have you largely power their grid um because of a big buildout in the 70s through nuclear then you have Fusion which is the bringing together of much lighter particles normally uh isotopes of of hydrogen uh
we've only ever used done Fusion successfully in bombs uh when we've tested them in in the 50s so controlling Fusion has been a big challenge but in the last few years a number of private companies and a couple of research Labs um the national ignition facility for example in the US
have started to make some significant strides and it might be in a few years we'll only be 5 years away from Fusion being being ready and I think Fusion would be a great uh addition to our energy portfolio but where the uh where the large tech companies are today is
saying we need the constant predictable output uh of nuclear efficient reactors these can be 800 megawatt 1 and a half gaw in size and they're available 92% of the time so the data centers always get their juice interesting you know I I am cautiously optimistic about about Fusion
it just it feels like we have a lot of different um approaches from the the sort of technical architecture uh in different parts of the world in the UK in Germany of course in the U but across the US and there's enough of a heterogeneity it's not like we're all Chasing After You the sharing of the
photo on the iPhone right where it can only be one winner and and you you'd feel that that would also create um what you often need with a new technology is um you know the the supply chain that's that's required I mean Fusion is is hard and it in the process it fires off lots
of neutrons that you have to capture in a you know some kind of material blanket a lithium waterfall or whatever it whatever it is and so there's a lot of requirements in in software as well and if there are enough players going after the you know the actual reactors I think
it creates something of a market further Downstream now you can make the same case for Quantum Computing which is also P persistently five years away I'm kind of cautiously optimistic about Fusion we should come back to Quantum but I want to ask you you built this amazing
content platform kind of go to you as as the guru what's going happen in the future but you also actually put your money where your mouth is you invest in startups so how is your view on the exponential age affected how you invest your money in startups what do you look
for what do you what have you decided not to invest in I still look for really really compelling Founders and having to feel that a founder or founding team is has just got got it they've got something about them and that actually I want to speak to them for the next
decade uh which I which I'll need to but the thing that I now bring to to bear is a rough question of what you'd call technoeconomics which is as the is there a a relationship where as you get better at the technology your fundamental economics improve this is outside of e
economies of scale these are economies of learning so I'll give you I'll give you an example um of one of these is a Rice University spinout which is building a material that is much stronger than steel it's half the weight of um aluminium it's as uh conductive as copper uh and it is carbon negative in
its production because it's carbon nanom material so once you're producing at its scale it sucks CO2 out and when I invested they were a million times the price of steel which is going to make it find tough to find a buyer but in 18 months we've got that down to a thousand
times and we have a path to being four or five times the cost of Steel at which point many many markets open up because of the other characteristics but there's no reason there's no reason law of physics or chemical reason why that cost curve can't continue to decline to be
below the price of steel and much much much below the price of steel in you know the coming years so that's a type of thing that I would I would look for and I wouldn't have looked for things like that you know five years ago because I didn't have the ideas as sort
of solid in my head so not just economies of scale but sort of a compounding technological competitive advantage of some sort yeah that is that it that can come from some sense of of learning learning effects now that the issue with learning effects is that they're really hard to predict a
priorize so all the work that's been done on learning rates is sort of expost right you you you get the data and you see what's happened with the cost of nuclear reactors and it turns out they have a negative learning rate so the more you build the more expensive they
get except in Korea um and uh you you have high learning rates in solar photovoltaics or in wind turbine and of course in Silicon chips but you it's quite hard at the outset to predict what that learning rate will be so what you need to do what I do is I you know I'm a
small investor so I don't do the depth of diligence that a giant does but I'll talk to the founder and see whether they understand this process and see whether they've started to deconstruct the problem and have realized where there may be strategies that will improve
processes and costs and where they realize they're taking a shortcut because we just need to get it out the door yeah yeah particularly relevant investing I think in climate Tech but across the board but absolutely climate Tech as well yeah yeah well I've looked
at your portfolio and uh I mean I do love some of the the companies uh in there I'm actually meeting uh uh Edward from synonym in uh in in a few weeks as well yeah I mean and that I think synony is a great example of a company that sits directly in that theme so the three
of us get to as a part of our jobs go and invest in in companies in our case our full-time jobs for listeners who are not investors either Angel Investors or Venture investors but who want to do well out of AI in their financial lives to to build wealth either when they're
thinking about investing their pension or just in terms of their career and winning in in the era of AI what should people be thinking about I guess first question just with their their pension portfolio is it just invest in the NASDAQ because that's where the ultimate
AI winners will go um or or indeed just all all in on Nidia um and then secondly from a career point of view how do people who might be either in their or indeed you know 40s 50s think about career reinvention so that they're not left behind you know the way we used to
think about that in terms of the internet and it was deemed to be crucial that you learn how to Google or you you'd be left behind yeah yeah I I still find that my wife is a much better googler than I am as well and I will never get as as good as she is on the on the personal portfolio standpoint I
actually really wrestle with this question um on the one hand you know there's we well shown Orthodoxy you should just buy your tracker fund and not worry about it too much and just recognize that markets go up but then if you're in the UK our markets have always
underperformed relative to the US so should you actually skew geographically to the US I think the US has got all the Hallmarks of a country that will do really really well over the next 20 or 30 years and then Tommy to your question um you know should you just go into the
into the NASDAQ and again portfolio Theory says you shouldn't do that right you should have um you know few bets here or there but of course once you've seen what's happened and to to the S&P 500 and you've seen what's happened to cap flows into the American uh Market I
think I do I do wonder about that and then I start to think should should the advice be have 80% of it in a low cost basket of ETFs and have 20% directionally in in certain places but then how do you make that decision about that 20% and who can you trust so I think it's a really really really hard
question I mean there are some things that I think you know I feel uh you know quite um that I've got quite strong conviction around one is that the US is likely to do pretty well all things considered that the The Runaway of with with big Tech and they're now very very dominant
as a share of US market capitalization in fact stock market capitalization sorry concentration is higher now in the US and it has been in more than 100 years um is I I do think that there will be a drag through to reindustrialization in the US right everything that Trump says and frankly
what Biden was saying previously says there will be Tailwinds to invest in those industries that have been forgotten about over the last 30 years and that could provide an opportunity and and and the other thing I think is to think about will be what happens with
uh stock market volatility in a process where lots of companies will find they not competitive so think about what's happened to the German car companies collapsing profits no way to get around China's cost leadership around EVs and so is there a way of of playing with
with you know with volatility it's hard I think it is very sobering though the point you make about Europe what's happened to the German big cap stocks and if you look at the footsie how many companies on the foots 100 are going to actually benefit in a major way from the
age of AI very very few and it's it's a I think a huge problem we're not here to give Financial advice but I I would if I were a listener looking at investing in the UK even trackers I would really be thinking about that so there was something fascinating in asml which is a
a Dutch laser company that makes the machines that make chips they had a this summary of the Semiconductor in ecosystem and they said about 50% of the ebit in the semiconductor ecosystem that's earnings before interest and tax is reinvested in R&D that's about 850 billion their math
is terrible $850 billion total revenue total ebit it was about 510 billion this as much more than 50% and R&D growth was running at 12% historically uh ebit growth was at 10% because the semiconductor industry is investing in the future and and I really struggle when I look at R&D s across
Europe both from a government level and at a um at a company level they are not where they need to be and in the few cases where they are in the German car industry they've played out really badly totally looking looking forward to perhaps a brighter a brighter note sorry
than the UK the UK's imminent demise um what you know we've talked a lot about AI today um but what's another technological leap that you're really excited about that may may change the world for the better well you know think Technologies rarely they rarely leap
they tend to move in smooth smooth curves occasionally though you do get these sort of par paradigmatic switches and I think funnily enough it wasn't a technology of productization chat GPT was a really good example but I am super super super super excited about solar
panels and about the way they're going to change the energy system and let me explain why I'm so excited about them energy is ultimately it's it's health and energy is clean water energy is fresh food energy is prosperity energy turns out to be wealth it turns out to be opportunity
when energy is expensive people can't access those things when institutions are weak and you can't build the old school top- down Energy System people can't access energy great example is Pakistan so my family is from Pakistan and Pakistan's Energy System is is
kind of rubbish right it's totally under strain the grid can't distribute enough there's load shedding all the time and Pakistan in the last year has become the fifth or sixth largest importer globally of solar panels and has done something that was you know deliciously Advanced
which is that they managed to get GDP growth while reducing energy usage now in reality they hadn't reduced energy usage what had happened was that businesses because they couldn't rely on the state to create the right institutions to allow for top down energy investment went often bought
solar panels so they could power their factories and their offices ongoing and and the government statisticians can't count that energy consumption but it's happening and so what they've been able to do is address a fundamental State failure solely because solar panels are
now so cheap they're cheaper than fence panels uh in many countries uh and and go and solve this problem themselves and we have to remember that you know four four billion people don't have access to the mobile internet today there are no countries that are energy
poor and Rich every country that's energy poor is also poor in the ways that we think of as poor and so having this declining cost um of of solar panels is extremely helpful and what really makes a difference is that they are modular so you don't have to spend a
billion dollars to get into the energy business which is what you had to do 20 years ago to build a coal plant you can do it with a $1,000 or $150 and and that is a piece I'm really really excited about because I think we can you know you see those photos of the
the Earth at night and uh you know subsaharan Africa is just dark and there are little pots in in in uh South Africa where there's lots of street lights and I think we can we can really achieve something collectively uh by the economies of learning and solar PV modularity uh of the panels and then
just the individual in initiative to be able to find a 100 bucks or 200 bucks to start getting reliable uh sources of energy so we we've talked about solar we talked about Fusion we've talked about AI let's just talk about you briefly we ask all our guests on giant ideas what
is it about you that has enabled you to have this extraordinarily successful career so for you people really come to you to see the future and and it's not future gazing it's you rooted in very deep analysis that you do yeah how did you get the confidence to I guess become
that Authority that that people go to for advice on the future well thank you for your really kind words uh some of it is just accident I mean I was born in the 70s and it was computers and and space operas on TV and so of course we believed we believed in all of that and
at the same time the threats were very industrial it was acid rain the ozone hole and you know nuclear weapons uh and but with from my own character I love ideas and I love getting ideas from people and I get love getting ideas from books and I love connecting all the ideas and saying well what if we did and
could you think about this and and that because I get so much out of that I can do that all day every day the whole time and and then I just think a very very um simple uh piece of of it's not even wisdom because it's so obvious you just keep doing what you're doing you just
keep doing what you're doing and maybe it's survivorship bias or maybe you actually genuinely just get good at what you're doing but I think repeating that that process uh and for me it is you know ideas from wherever from whomever trying to bring them together trying to
make sense of them and being cast because of the 70s and the 80s in a world that felt uh like technology was doing really exciting things amazing aim thank you absolute treat for our listeners we loved it thank you so much thank so much appreciate it if you
enjoyed this podcast please do share it with a friend and if you'd like to hear from more Visionaries sharing their giant ideas this podcast is available on Spotify apple or YouTube and of course we did discuss investing on this episode but please do not take this as Financial advice [Music]