hello everybody thanks for having me thanks for indulging me if you will we are gonna venture like we haven't done today right a quick look at the future I've been introduced as a futurist I don't actually like that term really I don't know why Fabiana IO splendors host today chose that because I understand myself much more as a historian right I'm looking at what happened in the future what we're past imaginations of the future what did we imagine the future to be 50 years a hundred years hundred fifty years ago because that is actually quite instructive that gives you a way to realign what your own forecast for tomorrow are going to be whenever I think about you know exponential scales about what the internet what AI what all those new technologies can bring us I'm thrown back to this piece of art by Italian anarchist design studio slashed art collective called super studio and that was a proposal 14 1970 World Expo it's called the continuous monument and they imagine a globe-spanning hyper object
and network that everybody can access a network that will transmit information at once again this was 1970 our notion of computers looked like that just a couple years before ARPANET was still top-secret in a military intelligence research program was launched this is what a quote/unquote modem looked like this is what five megabytes were like in the time right as as five megabytes can you mad during your digital photography or videography on this and this was how the future of work was supposed to look like right this is how you work with computers of course if we look at one of the most capitalized companies today and how their work with computers look like it couldn't be further from our inner past notions of how that might look like this is Facebook's headquarter by the way so we tend to overestimate the impact of technologies in the short term and tend to severely underestimate the impact of technologies in the long term and I think Aviat that is that the degree with which we overestimate and
underestimate is quite different so what we overestimate has a different quality and then the underestimation in the long run but we're here to talk about the exponential society and we've heard a lot about that right so we might want to take a big bit of a look back have a look at Moore's law then the RSA mother of all Exponential's right framed in the late 90s late 60s by Gordon Moore co-founder of Intel basically one of the leading heads of why we're all here the integrated circuit it became a self-fulfilling prophecy we've talked about science-fiction ad turning of self-fulfilling prophecies earlier today this is what we're really talking about right processors doubling in speed and capacity at same cost every two years and if you just look at the marketing material it still holds up quite remarkably well to this day this is of course Apple size and launch of it so that made us to go from these computers as expensive you know building consuming entity is used to compute the flight paths of
intercontinental ballistic missiles retailing for I think the crater was about 35 million dollars inflation adjusted to this a computer on your SmartWatch on you on your watch on your wrist that you can actually talk to that's a nature of exponential change we're talking about artificial intelligence but before we go into that I want to take you on a little side story aside of you will who knows who this guy is where as an AI conference they should be like 50% of people should be raising their hand that's way too few okay you need to read up that entering he used to so we were talking about that a the fki a launch deep learning thing online so he co-founded Coursera for one of the first MOOCs if you will and lately he headed by dues artificial intelligence research lab so he's a guy who really knows what he's doing and he's got he had by those resources to pursue it right so he was one of the four at the best price we would it
can be for AI he had a presentation I think a month ago two months ago where he said AI will transform the world the way electricity did and we've heard that before in the video that in showed this whole day right and I like is hang on not only have I heard this one before this was about Internet of Things where I started my practice and now we do much more emerging tech but so yeah with the IOT given enough time we won't really think about them as connected to the network anymore much as we don't think about electricity as you know a special product requirement of course that's not really what he meant right what he meant was this would he met ubiquity and masses change and so it might be prudent to look back at how electricity actually developed when we take all this lighting for granted but it's really due to this guy Thomas Alva Edison who then got in there fisticuffs with those guys Westinghouse and Tesla over which platform which
protocol of you will should be used for distribution of electricity vested economic interests of stuff got really nasty they killed a couple of elephants as presentation how awful alternating current this Tesla for a long time for gotten kind of resurfaced thanks to those guys but what we forget is what a sea-change electrification actually was was huge political mandate behind that there was a whole ideology based around that fact right it doesn't really work out that well though and it led to huge concentration as energy tends to do and led to military applications this is the demon core if you've never heard of it I encourage you to look it up the demon cord the story is mind-boggling this is a supper subcritical plutonium core that they just you know experimented with two people died because of you know sudden criticality and a crazy radiation exposure I quite like that image because it reminds me of how nonchalant we talk about inner potential implications and experimentations with AI this is something we might want to consider
the military applications are in a potential electricity generation we're quite obvious I mean this is the explosion and I think it's ten milliseconds after Trinity test site explosion led to nuclear power massive upfront cost massive massive capex investment if you will concentration effects across the electricity industry and then this happens right I mean we had turn our bill we had always our issue with nuclear but Fukushima really broke it and so ever since we're in the massively contrarian direction right solar is going through the roof wind is getting ever cheaper and and our forecasters can't keep up if you look at the official of Jem statistics on what UK and the British are not renowned for their generous subsidy scheme for total taxes placed just look at how poor the forecasting works out there it's crazy average prices for energy go down average prices for berries go down and suddenly there you have it that's almost exponential change if you look at it and suddenly those guys are rewriting the second time a whole industry
to the tune that the Bank of England not known to be the most alarmist or most progressive entity in the world is warning about the macroeconomic impact of the calm bubble because so much of our current valuations is tied up in future revenue flows and just looking at what happened to you as coal is mind-boggling top for use coal stocks of last four year lost ninety-nine percent of their value that is would you communicate if you equate artificial intelligence to electricity this is the history that you imply if you say AI is going to be the next electricity we've been talking about exponential change right and artificial intelligence artificial intelligence actually coined a day I'm quite surprised it didn't come up before today at a now-famous conference at Dartmouth us not UK in 1956 where they figured it can be too hard to completely model had brain works right should be done in a couple years with a generous research grant we still know we're close as we've just heard right but if we model
just the time and the growth that passed in between and we've got to be careful the Dartmouth conference was before even more as law existed right before we had a notion of what an integrated semiconductor is but if we look at this time span over 60 years compute power on an increment so it would have been increased by not 10x not 100x not even a million X but but just a bit more than 1 billion X that's the nature of exponential change that's make it so hard to forecast the future now of course we've heard all about the applications right self-driving is a thing voice-activated is a thing and note the words I've chosen right because that is one of the problems in AI research that you will encounter once it works nobody talks about it as artificial intelligence anything right self-driving cars as if driving cars it's not artificial intelligence toys activated systems are your voice' system that's not the artificial intelligence but you want to know what we're going to look at in
2030 don't you well and have let's have a look one thing to bear in mind when you look at the future is if you focus on just one dimension which is one factor or one variable you're bound to imagine these futures right that's always been lazy so let's have a look at four more or less plausible scenarios for what we could encounter in a short 13 years from now the first one has been discussed at length today preceding weeks months it's the mother of Exponential's it's the singularity and we're not talking about these kinds of singularities although you know the proponents kind of want to push that similarity you all know who this guy is right who doesn't know who this guy is no hands good I'll take that of course the technological singularity AI is getting ever better and never faster and once it hits the threshold it's getting better at better faster than we can control or we can get better and hence we have a runaway artificial intelligence which is going to
rule us all basically the nightmare scenario it's possible some extent it's even plausible but certainly not preferable let's have a look at a second scenario we're going to get into more interesting scenarios right we've just talked about self-driving right self-driving is going to rewrite the history of the industry there's no two questions about that but what self driving an electric really are is innovation across inputs and if you look at you know a newsroom circa 1980 and we all know News's in the gutter financially intellectually news has a problem but if you look at the inputs not that much has changed right it's gotten more efficient but the core product has made doesn't really change as arena change which really changes what happens on the output right digital distribution late all the difference and if you're working on any kind of future forward planting capacity I encourage you I really encourage you to have a closer look pay attention to what's happening in media and music because media and music were the first content types
first products if you will to go online because the data is native to the medium news and music is the carry in the coalmine for all industries where digitalization happens that's as an aside what we mean in this scenario is basically there's not so much going to be an exponential but a sigma little curve if you're in PR if you're in market research marketing what have you you've seen this chart you've heard of early adopters you're heard of the early majority that comes of agricultural research and kind of showcases the diffusion of new technology through industry what's really interesting though is if you apply a lot of Perez's theory of how innovation happens to the sigmoid curve you have the innovation on inputs and then outputs and usually a big chasm of financial crisis in the middle so that's probably what's going to happen in scenario two what's good to bear in line theory is the differentiation between innovation and inputs of a product and outputs of a product and Clay Shirky probably said this
I mean he applies it to communications tools but really it applies to anything you can't really innovate across user behaviors if the technological foundations are still unsettled that's why Facebook couldn't emerge in the 90s but early in the 2000 because everybody had internet access right we didn't have to worry about 56k modems anymore and that's true across a wider media and internet ecosystem as long as you know the technological solution a situation of internet access was unsolved there was no room for innovation on the outside layer now let's have a look at another scenario we're getting into slightly multivariate world now I think this is what has people really worried it's the centralization scenario right it's the future of autumn is Asian a lot of people like to talk about data as the new oil and we've just talked about what we imply when we talk about AI as the new electricity I want you to think about what meaning people carry over when they talk about data as a new oil because most likely
they're talking about centralization effects right we're arguing right now that you know the web is leading to a whole new class of not oculus we're traditional antitrust approaches aren't really effective we're worrying about a IAR supplanting us making of it you know our future a jobless future there is one guy in the picture has anybody spotted in it because if you think and if you buy into the notion as data as a new oil you're assuming implicitly that there is going to be huge upfront investment and centralization effects which is going to turn our cherished world after you know strong middle classes of where in Germany a strong middle that stand as the backbone of the economy into world rule but power loss on a lot of internet platforms we already observe that right that's a worry right now in policy it's a very plausible scenario averages over and if you haven't read Piketty or at least heard of Piketty and you worry about the effects of autumn automation I encourage you to read that
book or at least the introduction because it's it's a big book and now the fourth scenario probably my favorite probably the longest to expand on is world of asymmetries we've talked about you know how renewable power prices go down and go down like crazy talked about Barry packs decreasing in price like crazy we talked about Moore's law which you know drives down the cost of computation crushed severely and taken together these are three deflation Airy price pressures that make access to technology so oh damn sheep and what happens if access to technology gets cheaper and cheaper is that we like to pretend that it enables all this massive creativity but it also leads to world of volatility uncertainty complexity and ambiguity I don't have to tell you that we're living in a quite volatile world as a political scientist the last couple of months have kind of been mind-blowing just in the speed of unravelling and we know this right we know that technology enables this kind of massive uncertainty change disruptions the urea was
first one to put the finger on it ever since Christmas we'll put out his book the lifetime of a typical S&P 500 firm has decreased by about a third so with that in mind the deflationary pressures of technology we might want to have a look at what AI could be like is it really going to be like electricity on the changes or is it going to be more like that good Berg's printing press of course 1492 that changed the course of history right Germany we're celebrating this year 500 years of this guy putting 95 grievances against the door of some church which then led to 30 years of atrocious war I don't always want to fast and end it in a treaty Investar which is very basis on which we now think as nation-state right the whole international order is basically a function of Gutenberg inventing the printing press that was a time of massive volatility and uncertainty complexity ambiguity no doubt so I'm a political scientist I trained in international relations with a focus
on what we then called new Wars what is now commonly known as a symmetric warfare this is an aerial shot of Baghdad by the way we know how this works out already consumer drones a luxury year five years ago are now used not in fact that but in Mosul as inserting weapons right there dropping grenades the US military is on record that one of their close allies actually shut down one of these drones they cost $200 with a patriot unit with a price tag of about 1.3 million US dollars that's really messed up economics if we're looking at a future of AI under deflationary price pressures under assumptions of decentralization we're going to end up in a messed up world there's malware right now not me right Mir I was just interesting because of the externalities involved that is actively seeking out whether the computer that infects has been infected by other malware and removes it so in all the scenarios we're talking about why do we think a eyes will cooperate and what happens
if we just forecast the cost reduction the access of cheap energy for a couple of years when everybody can train their own AI these are four scenarios for more or less plausible scenarios just you know to lighten up the mood a little pick your poison there's two things I really want to leave you with I mean I know these futures can be depressing is these two first pretend to severely overestimate the effect of technology in the short-term and severely underestimate the effect of the long term secondly we can only really innovate over user behaviors over the outputs of any given system once we've set up the technical challenges on the infant's level and with that I thank you all for being here fabien we'll be right up here with the battle thank you