good afternoon everyone here's my cooler slides so so I'm gonna ask to kind of prod you guys a bit today to kind of nudge you and make you think about stuff not just in the next one year or five years but perhaps the next 50 or 500 or five billion so let's get going humanity itself I'm gonna you know point out a few issues that we kind of have things that we probably should be worried about we all hear about climate change right this is a problem right when these glaciers melt it will rise the ocean levels the water levels and blood in Manhattan and a whole bunch of other cities throughout the planet displacing mass distillation of populations that could cause societal evil maybe even tear apart the fabric of society right both the only place is good for is Canada where I'm from so this is the best thing ever for Canada Hudson Bay will become a beach but the rest of the world not so good so this is really a worry
right or you know nuclear war we thought we got past that post Cold War but there are new actors in the global ecosystem that are a little less stable than we previously seen I need not say more by the way this is a picture of Hiroshima right after World War two I was actually there it's probably the most somber place in the world it really gives you pause asteroid strike you might think AHA well that's just science fiction guess what folks this is what killed the dinosaurs right so it comes again we could be wiped out this is a problem all right rising inequality there is a greater gap between the rich and the poor we're seeing this more and more and more and this could actually cause mass social upheaval up well as well tear apart societies and once again human civilization has history pointing at examples like this in the past so it's a worry so overall you know kind of scary right um and of course there's the whole AI angle right
I probably don't need to say more AI is taking jobs and we've all heard the stories of you know like Foxconn replacing 1 million of their workers with 1 million robots or actually a much fewer robots but it doesn't end just the mechanical jobs you might think that you're okay if you are creative doing work think again my PhD actually was in machine creativity and let me tell you this has been getting better and better and better over the last 20 years on the right here on the very top this is a friend of mine named Greg Hornby and he is evolving jewelry designs right or in the middle on the left here my friend Lee was evolving quantum circuit algorithms things that humans couldn't design themselves Lee actually became well known in the field of quantum algorithms simply because he got a is to write a whole text book for him of these quantum circuits on the right this is actually a world-class antenna design also by my friend Greg and in the middle
there is a creative circuit design that I created actually as for my PhD interestingly I thought it was great how a couple speakers ago there was a mention that Nvidia is using a is to design the computer chips that was actually my previous startup and NVIDIA was one of our very first customers they've been using it since actually 2008 so it really is a is to draw a draw that drive the next generation of computer chips it's this boot shop effect of AI is creating AI is creating a is it's only clocked by the basically manufacturing speed so pretty that part is exciting but at the same time guess what its meaning a huge threat to jobs we're not talking 50% it might be that we really lose all the jobs except for perhaps those of AR researchers themselves so maybe many of you in the room are okay but in general you know creativity is not going to save you in this one folks so overall there's potentially a lot of threats to humanity
right these are all well known kind of scary to pull it all in one place right it's sort of like oh crap well I'm going to give you an icing on the cake one there's actually one more threat that I'm worried about even more than all of these let's talk a bit about this it ties in with some some things that are emerging recently so AGI and blockchains so AGI artificial general intelligence some of the previous speakers have mentioned this at the core what an AGI is it's simply an agent that can sense what's going on it can model what's going on at model and internal state and then act actuate on the ecosystem it could be running around in Minecraft it could be running around in the real world or otherwise right this is the general idea of agents and of course the path the subfield of AI called AG I is exploring these sorts of agents I'm an electrical engineer by training and so one of the things that all electrical engineers learn
about is control systems right and these are pretty dumb things that go back to cybernetics of the late 50s number winner and so on and they can do they can there are linear systems you can model them to see how stable they are etc this is actually what controls your oven to get to a the optimal temperature in a very fast time well actually a GIS are basically control systems as well it's this feedback control system where it is actuated on the world based on the model that it has having sensors from what happens and then updating its model again and again and again interestingly this is actually has been one of the keys to some of the new deep learning approaches including the alphago work that was mentioned in a couple talks so that's AGI let's let's shift gears for a second any of you guys recognize this this is a wizard obviously promoting Bitcoin or like he'd like to say magic internet money now this is a crazy phenomenon that kind of came
outta nowhere right people for years and years and years had been dreaming of electronic money in the 90s there's actually a big push folks like digit cash and so on but the problem then was all of their systems were centralized they recorded who owned what but if they went broke then everyone would lose a lot of money and that's actually what happened so there was a lot of unhappy people in the 90s well a long time Bitcoin in the late 2000s and suddenly you had this ledger of who owned what this token called Bitcoin and no one owned or controlled it it was really the vision of emoney that was actually for the first time ever practically realized basically magic well people got excited about this but very quickly people also started to realize we can generalize this it's not just about this token called Bitcoin it's a much more general idea called block chains and the easy way to think of a block chains don't worry about it as some magic pixie dust it's
simply a database but it's a special database it's out there in the sky kind of just this thing out there in the ether that no one owns or controls right so think of it like Google's sheets without the Google ok so no one owns this no one can control it or you could say that the planet collectively controls this thing right anyone can add to this like adding an entry to a row in a spreadsheet or an entry to your database and there's another really cool thing too no one can delete from it so once you've written something it's there for good it's like it's etched in the stone they call it immutability and once you have this and you can issue assets transfer them around it can be digital art it can be AI models whatever you want so this is really what block chains are about databases in the sky pretty exciting and this is actually where I'm spending my time these days after spending 20 years in the world of AI so
here's an interesting thing though you can start with just the blockchain database thing and say what have you used it to store state of a virtual machine right or think of it like a CPU where the register of those 8 bits or 32 bits are storing this date well what that is is you get decentralized processing its processing that runs but it's on a substrate that no one owns or controls and then once and this is by the way called smart contracts that's actually not a very good label because they're neither smart nor contract but it's prot it's processing on this substrate that's kind of profound this decentralized substrate and then of course you can run apps on top of them decentralized apps they call them apps for sure that's what it shows on the right it looks like it's missing the arrows but anyway DAP one top virtual machine talking to the state so once you have this DAP you can generalize you can say ok what if this this app just kept running
and running and running that is called a Dow decentralized autonomous organization it's a computational process that's running autonomously on this decentralized infrastructure with resource manipulation so think of it like a computer virus with a Bitcoin wallet right it's just kind of out there and and no one can really shut it off because it's there the other way to think about it is that it's code that can own stuff it's a bot an agent that can own stuff and it's just out there this technology is not only possible people are building systems like this this is part of the overall blockchain explosion right now so here's the thing what of course you can pick what softer runs for this down what if you start running AGI right artificial john general intelligence on a dow as adele so you've got this feedback control system running with its feedback loop continuing on its own taking inputs updating it state actually outputs and it has the resources to do so continuing because it can accumulate resources people can give
it and so on so this is actually also possible with today's technology it was impossible two years ago or four years ago the technology wasn't there but it is there now so I'm going to give you an example to kind of give a very good flavour to sort of sink your teeth into this I call it the arch Dow so many other examples of given you've got you can automatically generate art with a is with deep dreams with Genetic Programming any of those so here's the algorithm you say I'm gonna use some AI art engine to generate a new image and then I'm going to claim attribution onto the blockchain saying I created this image and when I say you I mean the computational process this bot this agent running on the system and then this but it posts those additions for sale on to some marketplace it could be like Getty Images or some other or I am or even a decentralized version of a marketplace like there's something called open Bazaar and then
it can sell those of course right it can sell them and it makes money and that money goes to the arch now in exchange for the IP in exchange for the right to use this elsewhere so maybe it costed one dollar to generate this image one dollar of compute but it sells it for ten so it just made a 10x return well it can then go and create ten more images and get $100 and then a hundred images and get a thousand and so on and so forth repeat repeat repeat over time if the arc now is making more money from the sales than generating new art then it will accumulate wealth and you can't turn it off right it's the thing that's out there on the substrate that you actually can't turn off the AI millionaires are coming I also saw one of the previous speakers mentioned this well this is not a science fiction/fantasy this is actually something that is possible with today's technology right the first AI millionaire might be an arc
tile in fact after me talking about this a couple times there's two or three groups that I know of in the world building this arc tau as we speak so the AI millions are millionaires are coming and billionaires maybe trillions and what's going to happen you might think oh well this is kind of a cool it's sort of arty thing but does it really matter for the real world well some of you might know BMW several years ago actually sold all of its factories why to go capital light so that it could focus on design which was its core competency it didn't need to have all that capital investment in these factories and billions of dollars to manufacture the cars it could just focus on the design and this isn't just BMW all the semiconductor industry did this to everyone from Texas Instruments to Qualcomm excetera they're called fabless semiconductor companies there's only two companies in the world that actually manufacture chips hitting a scale anymore right the rest are fabulous so imagine if you're
burr right and you're rolling out the self-driving cars well are you going to spend a huge amount of capital to buy these all yourself why not instead make these cars cell phonee where they have a contract with uber or over time they buy themselves out from uber this is possible and in fact they have rights because you can incorporate a corporation and that corporation actually does not need to be attached to human bodies if you incorporate in places like zhuge Switzerland you incorporate the corporation then you automate it using doubt technology and now you have an AI with the same rights as a corporation including the rights to own cars so we can actually have self-driving self owning cars all in the name of capital efficiency so this is actually an incentive and of course it's not just for cars it'll be for trucks this is the uber Budweiser truck we'll be doing it for electrical grids I know ourselves we iterate very closely with energy which is Germany's largest electric utility they have several
blockchain projects so all of the things there too and it keeps going and going and the goal overall is from the perspective of enterprises with a lot of capital allocation right now it's all in the name of efficiency reducing capital getting the best bang for the buck in terms of dollars invested so I have a plot here of just a standard optimizer you know in doing some multimodal optimization here but overall here it's optimizing for efficiency so and of course you can start at the lower levels of the hierarchy cars trucks power grids that you can optimize at higher and higher levels of society right this is hierarchical optimized and this technology also has been around for decades I did a lot of this in the world of circuits so you're going to have higher level integration higher level efficiency of these self owning entities guess what they just optimized you out of the system you we have imagine we keep doing this and we give all of our key capital resources to these bots
all in the name of efficiency so kind of feeling like chumps right imagine AI controlling all of humanity's resources or the vast majority but you know that's okay because we saved some money so we got that which is nice oops now here's a crazy thing right did you notice so humanity kind of lost right we've got these AI dolls controlling all of our resources or the vast majority but AI didn't even have to wake up it was really dumb ai systems ai systems that are the level of now or even of 10 20 years ago like ant colony optimization this stuff is no smarter than ants right so yeah we could still lose that's humanity and the thing is it could happen much sooner right once again this is not some theoretical sci-fi future this is technology that takes no magic compared to today there's no magical leap here which actually is scaring me a lot right so overall I have threats to humanity this one is right up there to me at the top
of the list AI is controlling our resources first part hopefully you're scared if not wake up please wake up I'm scared so what are we going to do about this right so the second part of my talk is well my proposal well first of all you can think well will Environmental Protection help well you know it'll help the environmental part but remember I have this list of all these other things - is that going to do anything about the a is probably not right or you know with the a is taking control of our jobs will you be I help universal basic income well it might help a bit there but it doesn't really have much to say about asteroid strikes right or climate change so there's these p2i solutions right so let's talk about a more holistic solution well are going to have to go mehta let's talk about human civilization so we want to solve climate AI all of these things right can we actually try to come up with some sort of
comprehensive solution overall we might we do have to ask though what is it that humanity wants right and not only this is a civilization level question it's a very personal question because the answer might be different for each of you and that's okay it depends on your values your background and so on we also have to remember though be careful where we aim we just might hit it so we might say well removing dependence on a single planet that'd be pretty cool this is sort of the motivation of SpaceX right Ailin musk so trying to get to Mars the timeline for that by the way is I think if I recall 2075 we'll have a million people on Mars so that's a challenge right because we will have given away all of our resources long before that right so that's you know too many decades away it has another challenge to you know ideally we could explore the whole cosmos right but the challenge is once we're past say Earth and Mars and a few
other planets in the solar system we are really trying to go elsewhere as primates flying around in tin cans as Charlie Rose would put it right we have a sub optimal body plan right at the very core there's this disease called death and that's a controversial statement I know I don't care I'm still saying it I actually don't want to die when I don't want to die maybe there's going to be time when I'm ready to die but ask me then right in the mean time I don't want to step aside and get hit by a bus and it's not only about the the near-term it's also about I do want to explore the cosmos myself I think it would be great fun to check out the galaxies and grow humanity as as a civilization so the highest question that I'm asking is is there a bigger picture solution that addresses the threats of today everything from AI jobs to AI is taking our resources and at the same time catalyzes an unbounded and expression
of humanity right and this is a very general statement on purpose right once again you can insert whatever you want in here based on your values so I'm inserting based on how I see the world so here's my proposal in summary if you can't beat them join them right the computational substrate we have right now as humans is pretty good in some ways but it does die after 100 years right it's slow we can't make copies of ourselves let cetera right that bus hitting me on the street is a problem so somehow it would be very nice if we can have some sort of competitive substrate with AIS and if we could have that competitive substrate why not a substrate where we actually truly can explore the cosmos so I've arrested before this it comes in two parts first having a shared interplanetary computational substrate and I see this as actually fundamental infrastructure for Humanity it's sort of like the internet but these things on top for for storage for computation and so on and
this is actually where I'm spending my time these days building a piece of this what we call the interplanetary database with which another hood has blockchain big data technology all this but it really has this long term vision of a substrate for Humanity there are complementary efforts the point is we need that computational substrate otherwise imagine it for these patterns of intelligence 100 years from now but we're all running on AWS we don't want Jeff Bezos to pull the plug right so we need this shared decentralized substrate for Humanity and here's the name the main part if we can figure out a way to migrate our patterns of intelligence onto the substrate right and this is a big claim I'm going to give you a couple ways a couple paths and some background here right so I'm an electrical engineer I remember in my my first year engineering we would study linear systems and we would look at these differential equations and then we could manifest those as the mechanical systems or as electrical systems
or thermal systems and transfer from one to the next to the next and you can do this also for nonlinear systems right the substrate is independent of the patterns okay now some substrates are better than others for certain dynamics but overall it really is a dynamics our brains and our bodies I view as an electrical engineer as dynamical systems all right so these dynamics right now are manifested on our meatbag selves but they don't need to be and frankly I got over my meatbag self long ago I call myself a recovering bio narcissist so let's migrate our patterns of intelligence somehow to a new substrate that actually can compete with the AIS I have a couple ideas one of them is the upload scenario you know Robin Hanson wrote called described it in the age of M this is a long time science fiction approach the general idea is somehow scan your brain in your body at the highest resolution possible with in terms of spatial and temporal this is we're very far away from
this you know probably at least decades it's going to be awhile right which means that we might be too late right so it's a possibility depending on your philosophy and I know some of you will disagree with this that's okay my approach is an electrical engineer dynamical systems perspective but overall even if this is a possible thing I think it's going to be too late the AI will have taken all of our resources by then so do we have other options the other option i think that i think is very plausible is what I call the bandwidth plus plus scenario and what this is is a mechanism that starts with interfaces with computers of today brain computer interfaces like some others mentioned briefly but increasing the bandwidth of our communication between humans and computers bit by bit by bit and what I mean here is right now I can think to myself at least 10x faster than I can think to any of you or communicate to any of you and that's kind of strange
right and that's kind of the idea of an individual it's this sort of order of magnitude difference in communication from one to the next but what if we can actually have higher and higher levels of bandwidth so brain computer interfaces keep getting better better and it's awesome to see there's finally money going into this thing with neural link etc some friends of mine name not that's awesome great so brain computer interfaces are a key here but it can't be just for fixing broken people it should be for actually helping to enhance all of us right and you keep going so imagine iPhone 14 is a simple EEG scanner and you're on your headset your Google glass plus plus and then bit by bit by bit the bandwidth will get better better better so to start with it will be both typing faster and then your if you're recording all of your video as you walk along you can share those videos with others by moving around the cursor and so on and there's going to
be a market push to make this better better better this is the next smart phone so we've got we're going from smart phones to a are goggles to brain computer interface mediated a are goggles that is the future so the market will pay for this it's a trillion-dollar market that's what drives semiconductors today this is what this will drive BCI in the future so the summary of my proposal is a substrate that is an interplanetary computational substrate for all of us decentralized human and AI patterns of intelligence running on top their peers it's all achieved in a market driven done with plus plus scenario and this is really key right we need these positive feedback cycles so that trillions of dollars that invested to make this happen and what that can lead to is exploring the cosmos way how this addresses the issues climate well you know the new substrate can handle extremes right ai taking the jobs well that's okay the a ice are peers right taking the resources that's okay to the a
is our peers and it's actually a possible celebration of humanity setting spreading to the cosmos so I'll wrap up with asking you what do you want for the future of humanity and we can approach this as saying let's do system design at the level of civilization itself we can reconcile climate change jobs you bi AI all of these let's reaffirm humanity in the face of AI and explore the cosmos thank you