In this weeks episode of The Futurists, cognitive scientist and AI researcher Ben Goertzel joins the hosts to talk the likely path to Artificial General Intelligence. Goertzel is the founder of SingularityNet, Chairman at OpenCog Foundation, and previously as the Chief Scientist at Hanson Robotics he helped create Sophia the robot. Goertzel is on a different level, get ready to step up.
Analysis complete. No addtional information is required for context. Proceed with transcript display ...
[Music] this week on the futurists Ben gertzel you could say that the primary use casesfor AI in the planet today are selling killing spying and crooked gambling AKAWall Street which is perhaps not what you really want the first generation oftranshuman intelligence to have on its mind [Music]hi welcome back to the futurists I'm Rob tercik with my co-host Brett King I'mhere too great to see you Brett and this week we've got a great guest this is going tobe Kevin gurtzel from Singularity Nets Ben has been at the Forefront of research into artificial intelligencefor many years and old acquaintance both of ours great to have you back thengreat to be here good to have you man um before we jump into that there's acouple news items so if you like I can just jump in man the news from the futurereally two things as a follow-up to last week's recording we talked a little bit about the China and some of the changesthat are happening there um but the stock market has been uh frowning upon recent moves in China toconsolidate power around president zeed and as a result the Chinese tech stockshave collapsed uh they're they're down significantly the uh NASDAQ Dragon Chinaindex has dropped by 20 percent uh since those announcements from the Chinese Communist Party Congress that happenedjust recently Alabama is down 32 percent uh below its September IPO 1914 uh sorry2014 IPO and the Chinese currency the raminbi has dropped to its lowest level since 2007. so it looks like China isgetting uh seriously slammed uh for the consolidation of power that's how the market seemed to be responding because Idon't want to buy r b if you're interested it's uh the othernews is that the United States is looking pessimistic about the future there's news today from Gallup uh theresearch agency and they report that 42 just 42 percent of Americans of Americanadults think that it's either very or somewhat likely that today's youth willhave a better living standard Better Homes and a better education than the previous generationnow that's always been kind of framed as the American dream but it looks like that dream is sour this is uh an 18percentage Point drop since 2019. it's almost 20 drop in optimismum and it's tied with the previous low which was in 2011. uh one of the things that's driving that interestingly is uhis two things wealthier people people with higher income around a hundred thousand dollars are way more pessimistic than people with lowerincomes uh which I don't know what that's meant to tell you and then the other thing is that it's also a verystrongly politically partisan uh view in the future yeah uh which is to say that Republicans are way more pessimisticthan Democrats um and the Democrats uh perspective on the future hasn't changed that much but there's been a giganticswing um when Donald Trump was elected president in 2016 Republican optimismswing up 29 percent and then when Joe Biden was elected lastyear it fell 33 points do you see like a 30 swing in optimism and that's one of the things that's driven that's crazy uhtoo low since 2011. so there's the news uh markets reaction yeah I just got one one out of my life that um you know umDepartment of Education is tracking homelessness for students prior to the pandemic 2018 2019 school year it wasabout 38 000 um students were homeless Across America that's ballooned to 1.3million last year and could Exceed 2 million this year um so we're looking at a um you know anational increase in homelessness of like five to six hundred percent because of the pandemic on the evictionsfree market can't fix this I wonder if people are going to look back at this and and say like remember that time inAmerican history when we thought it was okay to step over someone who was living on the sidewalk it seems pretty common in big American cities we can't affordto have homeless people it costs like thirty five thousand dollars a year anyway that's and that's for a futureepisode Ben welcome welcome back or welcome to the showum the you know the obvious thing to start with is is um you know how are you and how are your your you've had acouple of kids since um you know a couple of children since oh it last came on the show how are they goinguh I'm I'm I'm all good I've been having it having a great time uh these lastlast few years and I you know I was living in Hong Kong for 10 years rightexactly and I I moved back to the US in early 2020 at the start of the of thepandemic mostly not because of the political changes in Hong Kong whichwere not having an impact on everyday life there for me so much but because itlooked like they were gonna impose more Draconian covered restrictions and Ifelt I felt like that I felt like dealing with it so yeah I've been living on a small rural island in the PugetSound off the coast of of Seattle so hanging out in the woods and on thebeach with my family in a lot of uh AI development and the blockchain researchdone and it's been it's been the productive time this fall I've been starting to do business travel again becauseconferences have been opening up right right you actually had mixed feelings about it because I was being veryproductive sitting home at the computer uh I can't understand getting getting stuff done right and it's uh so it'sbeen an interesting time yeah I did the I had my fifth child uh during the during thepandemic she's a year and a half year and a half old now and uh it's been soproductive productive in a number of accounts you know I would say thepandemic and Associated Mayhem probably had the less impacton what I've been doing in AI than on many parts of the economy because I mean we're Singularity net open Cog the variousprojects I'm involved in we're we're a bunch of AI Geeks and programmers already mostly working from home fromvarious locations all over the world already coordinating by on online mechanisms and for the people who aren'tfamiliar with your work can you tell us about opencog and Singularity Network what are they doing absolutely so my ownmy own career since the 1980s has been centrally focused on AI and inparticular on AGI artificial general intelligence so trying to make takingmachines that can really think like people now that that's the long-term long-term research project which isgetting closer and closer as we'll talk about but along the way I've been working on application of AI in avariety of different you know vertical areas both because one more staff one iscalled doing useful things not not just playing the research lab and also I mean this is has helped for making money topay to pay programmers and get real data into AI systems and so forth so open Cog is an open sourceproject aimed at laying the foundations for real artificial general intelligence for thinking machines a good thing learnimagine generalized like people first version of open clock system was launched in 2008 we're now working onand almost from the ground rewrite called open called hyper arm which is aimed to let open Cog really scale up ina way that takes advantage of modern Computing infrastructure then Singularity net a project I found in2017 is a blockchain based platform that allows AIS of any kind to rundecentralized in a whole bunch of computers owned by a whole bunch of different people without any Central owner or Central controller so you canyou can run an open Cog on that you can run neural Nets and that you can run a lot of different a lot of differentdifferent things on there so I've been I've been sort of working it multiple levels in in building an AIoriented Tech stack and then also working on applications of these AITools in a variety of areas including longevity medicine crypto Finance robotics we got the Sofia robot and ofcourse yes her little sisters grace and grace and uh and Dez Demona and thatthat piece has been more annoyingly influenced by the pandemic just because you got Hardware instead of justsoftware you got to get parts and pieces and so forth very interesting about AI so you you are um with Singularity netyou're focused on using a blockchain to decentralize AI that sounds really interesting and promisingtell me why that's important um you know is there another group that's centralizing is AI tend to tendto be centralized and is that important to differentiate from that I mean at at the momentAI is very heavily centralized in itspractical deployment and it's it's centralized within a small number oflarge companies and and a few a few large large governments right andthat said good and bad aspects there's a certain efficiency of course that that that comes with that I I thinkthere's also been a tendency for AI to be channeled according to the thebusiness models of these large companies interests of these governments and I Ithink that does have bad aspects as well as good I mean the good aspects is you know Google tencentFacebook Microsoft they were pretty efficient organizations that at moving Technologies from research to deploymentright so they're they're good they're good at doing stuff and they're making some amazing things happen on the otherhand you could say that the primary use cases for AI and the planet today areselling killing spying and and crooked gambling AKA Wall Streetperhaps not what you really want the first generation of transhumanintelligence to have on its mind as it moves moves beyond the beyond the human level right I definitely want to getinto the transhuman intelligence idea that's that's something as well but um so let me just um sort of take us backto Singularity net and and and then talk about the developments in the last few years because last time we spoke was waspre-pandemic and you know at that time you know you were doing a lot of work atSingularity net to try and build these um various components of artificial intelligence or competencies with a viewthat at some point we would be able to aggregate this up into ATI but the last few is because of the advances in deeplearning you know these general purpose AIS that could lead us to AGI seem to begetting a lot of traction so how is the overall thesis behind um you know AGI changed over the lastfew years so my own overall thesis regarding AGI and my ownthinking about how to get to AGI has changed rather little in the last few years andthe ways in which it's changed are pretty in-depth and mathematical and unrelatedto the advances that that you mentioned I mean I posted a paper onlinecalled the general theory of general intelligence which outlines some of the work I've been doing kind of the unifieddifferent cognitive algorithms in this in the standard a standard math framework so my thinking about how toget to machines that can really think hasn't changed too much I mean one thingthat's changed over the last few years is that very large neural net models trained on huge amounts of data havedone more and more cool things right and this is exciting it's bringing money andattention to the AI world and it's doing useful stuff what's the relation of thiswork to the quest to build real thinking machines is a different question and some my friendGary Marcus who would be a great guy for your podcast if you haven't had them yet but so again Gary Marcus if you look athis online articles and videos he's he's made a pretty coherent well thought outcase that these recent advances in deep learning using large neural models constitute extremely little progresstoward real general intelligence because pretty much they don't understand whatthe hell is going on I I mean I mean I mean they're they're they're they're they're they're they're faking it they're faking it in the veryinteresting way because they have so much data to use to to drive they'refaking it and in some domains that will work pretty well so like if you look atgenerating art or something yeah yeah I'm just gonna say that I mean theseprograms are not they're never going to be Van Gogh they're never going to be Andy Warhol they're not going to come up with some new thing that hadn't everbeen done before because what they're doing is looking at surface level patterns and compositing them togetherin on the other hand that's kind of what many commercial graphic artists or evensome well-known fine artists are doing right like you can you can you can you can get some mileage that way so part ofthe discovery here is like how much of what humans do which we considerimpressive and then the lucrative and intelligent how much of it can be justfaked by gathering a bunch of stuff and glamoring it together in this sort of asort of Artful way so that that's interesting that synthetic imagery and synthetic art is kind of aparlor trick uh it's like a neural net parlor trick well it's that's that's unfair to itbecause parlor tricks don't solve real world problems and they they they don't make you huge amounts of money so Ithink it's it's a new category of of entity right so it's it's more than theParlor trick and and less than less than and and AGI right becausethere's no path there towards uh general intelligence I don't think so like so inthe domain of Music which I've been playing a bunch with generative models and music because I one of the thingsI've done the last few years is started playing the keyboard again we started rock band with a robot as as alead vocalist so we we've been playing with AI for generating singing and and generating music and I mean it's it'sgenuinely cool as a musician right like the AI will come up with new melodies or new vocal stylings that as far as me asa musician to play different things that than I would otherwise right so I mean it's it's it's all good it's not just acheap demo on the other hand like no way does it come at music with the passionand inventiveness of a of you know a really outstanding original humanmusician or composer or improviser or something right so it's uh back toBrett's question you your Brett was asking whether you see a point where these different strands of Technology will will be integrated or complicatedyeah and then somehow maybe is emerging from that is that I do think there's a path to AGI in which multiple componentsarchitected according to different architectures Co-op operate together and you get some emerging intelligence outof it but I think among those components needs to be something with somefundamental capability for abstraction generalization creativityand and Imagination and I think that's don't don't really have this soyou can't I don't think you can just take a bunch of narrow AIS serving some vertical application functions Networkthose together in singular argument or anything else and general intelligenceso you need to take a new new run first approach is that what you suggestno I think there's a number there are many approaches that one could take I mean the approach we're taking in openCog actually is you have components that are doing symbolic logical reasoning youhave components that are simulating Evolution for creativity and and are doing doing evolutionary learning andyou have components that are doing neural net pattern recognition and you can you can network network all of thesetogether but I think I think well that's the approach that we're taking which is using some advanced math to structure avariety of different algorithms I think you could also take a more biological approach and try to doyou know deep dive girl and gleal and astrocyte modelingand to try to take the biology biology more seriously than than current neural Nets are doing Ithink there's there's a variety of approaches that could be taken to to generalintelligence but I don't think the current deep neural Nets that have sortof fine-tuned to meet narrow application goals based on training and large datasets these alone these alone cannot do it actually or massive data sets aregoing to be critical right I don't know I mean the smarter your AI is the smaller a data set you can getaway with I mean part of the reason you need such huge data sets is because there's no generalization happening so you you need sizable data sets but Imean you know mid-journey has seen more images that than I have a gpt3 has seenmore language than than I have that amount of that amount of data isn'treally isn't really necessary so I I do anyway I do think you can get multiplecomponents cooperating together to sort of emerge emerge in AGI but when peoplehear the word emergent I know what some of the people listening are going to say some of the people are listening are going to say hang on a second you'resaying if you aggregate all these different Technologies and different approaches and combine them together then something magic will happen that'swhat emergence is something magic occurs and now suddenly we have intelligence and they're like I don't think it'smagic I mean when you put hydrogen and oxygen together to get water I mean the water is what the hydrogen oxygen werenot wet but that's not Magic it's just yeah but we're not combining molecules here right chemistry right I mean andthis is the same thing in the brain I mean if you put we're putting the hippocampus and the cortex together andthey do stuff that neither hippocampus or cortex does on their own but that okay so you're saying and replicate thatmain functions it's emergence yeah okay so what I'm hearing you say is that if you can replicate certain brainfunctions in algorithms uh and combine those together then we might have the generalization that you're referring tothat or or am I missing something you talked about an abstraction layer I'm trying to figure out where and when that gets developed and introduced well Ithink there are going to be many different approaches to building generalintelligence and I'm not sure there's only one only one golden path and I I II I do think there are paths that are more closely tied to human biology and their pathsare less closely tied to human biology so I think one can look at human cognition and the different kinds oflearning and memory the human cognition deals with declarative memory episodic accessory memory you know actionattention you can look at the cognitive functions that human mind carries out and figure out you know clevermathematical algorithms to do each of these cognitive functions maybe in a quite different way than the human braindoes then have a network that combines together different agents performing mathematical algorithmscorresponding to different key cognitive functions in in the human mind and there could be you don't need a neuron inthere you don't necessarily need any any simulation of anything in the brain on the other handI think you could also simulate the non-linear dynamics of neurons and gliaand and chemical and electrical diffusion for the brain maybe even the quantum Dynamics in the water Megamolecules in the brain like you could you could you could you could go all out and do a detailed biological model andyou could then get it generalization and and creativity and Imagination from fromthat route so I think there's going to be many paths to general intelligence which could lead to different kinds ofof intelligence it just happens that deep neural Nets asthey exist right now for commercial applications I don't I don't think Idon't think those are one of the many pests to general intelligence interesting that that can work all rightwell that's a good place for us to take a little bit of a break here we're going to have a break but just before we jump to that Brett wants to ask you the quickfire questions so take it away my friend okay here's the quick fire lightning RamBen what was the first science fiction you remember being exposed to Star Trek the original Star Trek TV show very coolum what name a futurist or an entrepreneur or scientist that has influenced your thinking and why uhGerald Steinberg the Prometheus project he wrote a book outline the singularity in 1968 which I read in 73 or so whichkind of kind of blew my mind he outlined another technology AGI and immortalityand said we had to decide whether to use them for stupid commercialism or expanding human consciousness so I readthat at like age seven and I was like so pretty yeahum this is a maybe a tougher question what's the best prediction an entrepreneur futurist or sci-fi authorhas ever made do you think so the the best predictionsever made well I mean Cyrano de Bergerac said we go to the Moon that was a prettygood one mate yeah hundreds of years ago in in the modern era actually my buddyRay Kurzweil has not done badly I mean he has his track record is not quite as good as he as he is as he markets butit's it's pretty pretty good yeah it's still pretty good pretty good yeah um and then uh finally just before breakwhat science fiction story do you think is most representative of the future you hope foroh of the future I I hope I hope for wellI I I I don't know that I have a good answer for that one actually okay that'scool that one that one uh may may have may have yet to be written all rightgood well um that's uh that's the first segment done we're just going to take a quick break and have some words from oursponsors and we'll be back to talk about how living with alternative intelligences and AIS AI is going tochange the future of humanity after the break provoked media is proud to sponsorproduce and support the futurist podcast provoke.fm is a global podcast Networkand content creation company with the world's leading fintech podcast and radio show Breaking Banks and of courseit's spin-off podcast breaking Banks Europe breaking Banks Asia Pacific and the fintech 5.but we also produce the official phenovate podcast Tech on reg emergeeverywhere the podcast of the Financial Health Network and next-gen Banker from information about all our podcasts go toprovoke.fm or check out breaking Banks the world's number one fintech podcastand radio show welcome back to the futurists withmyself Brett King and Rob turc as co-host and we have Ben gertzel live from thethe coast of Seattle where did you say it was that you're living these days I'm actually on vassan island which is inthe Puget Sound office okay Puget Sound beautiful area of of the countryum but Ben spent an extensive period in Hong Kong working in southern China andso that is sort of the subject for news from the future Deep dive today let'sget to it uh so um there's been a lot of debate umin the U.S press over the the last uh a few news cycles of the progress beingmade in artificial intelligence in um China um that is a big driver towards therecent uh you know which we spoke about in the last Deep dive in terms of chipset controls and things like thatfor export and import um that the US has put on China but NickChalan the U.S department of state's first ever software Chief was forced toresign in September of 2021 after he claimed that the United States had nocompeting Fighting Chance against China in the next 15 to 20 years based onartificial intelligence and in the National Defense reportfrom September of this year so very recent the United the specialcompetitive studies project released a new study called mid-decade challenges to National competitiveness identifiedthat China is continuing to invest and has surpassed the U.S in three key areassemiconductor development artificial intelligence and 5G and this was fromthe CEO of the defense writers group saying that the U.S has just one budgetcycle to get this right uh badge Qatariumkatari said if we don't get our act together in these three core Battlegrounds in terms of Bio in termsof Next Generation computing power in terms of Next Generation inventions it's not going to happen in the countriesthat the Forefront of democracies today everything will happen in China now wetried to stop them on 5G but the U.S you know clearly is significantly behind on5G we're significantly behind on edge Computing and and in terms of artificialintelligence development the later statistics show that China produces somewhere between and it depends youknow depends on the stats that you read three PhD to five or eight PhD graduatesin the fields of artificial intelligence for every one that the U.S produces so they're just able to throw a lot morebodies at this problem Eric Schmidt has been one of the real voices of concernin terms of AI development he says the real issues compression of time thesesystems are going to have to make decisions faster than human decision-making time frames and that'swhere the boundary is going to be and we're going to have a serious conversation about that in society sothese short-term concerns you can see playing out the major voice in terms ofthe AI threat to the United States Market has not beenfrom the market itself but from U.S defense department which sees this as abig strategic threat so how is the U.S going to compete well my call on this isstopping China from developing these Technologies we tried that with 5G and we found Huawei is dominant in terms of5G standards and Technologies right now globally the only way for us to competein the United States is by making AI based and stem-based education acrossthe board free because that's the core infrastructure that you need to develop competitiveness of these skills andwhile we continue to have this pay for play you know free market approach toeducation we are restricting the lifeblood force of AI development in inChina we have the investment we just don't have the skills that's my deep dive for this week Ben I would love tohear your thoughts on this sure so I I did I lived in Hong Kong for10 years and I had many dozens of visits in into Mainland tomeet people in various universities and and companies in in China doing doingall sorts of different work I mean I I think thatyou know that the push of China into the AI field has beensubstantial and impressive but still has seriouslimitations I would say that vast majorityof significant AI Innovations have come from the U.S up until today still andwith Western Europe following second andChina way way behind in terms of really like wild new ideas coming uptrying to probably behind Russia or Japan for that matter if you want to spend account then we hear all aboutwhat so that is scalable deployment of of AIright so the algorithms the ideas have mostly come from U.S and WesternEuropean universities and PhD students not even from Western companies but then the typical pipeline has beenprofessors and PhD students come up with new stuff they prototype it Western companies roll it out first proving itcan be done commercially then Chinese companies take it and roll it out better at larger scale and sortof mass Master the the real world deployment right and so that that'sthat's the story as we've seen it so far and then whenwhen ethnic Chinese researchers emigrate to the west and embed themselves inWestern companies and universities then you often see the Innovation level ofwhat they produce go way up because the the social the social context is just isjust different like China hasn't managed to make a Google Deep Mind or an open AIor say an open car or or singularity in that right but they which is cooking uplike wild New Frontier ideas but deployment is alsoimportant and they they've certainly shown incredible Mastery therenow so there's been this flood of research papers uh you know Brett pointed out this week and we spoke aboutit with a previous guest as well uh some people say there's an exponential increase as they always do aboutanything that goes up and to the right an exponential increase in the number of uh AI research papers being produced byChinese researchers my question for you is uh there's a yeah there's a large numberof uh filings but patent filings as well as papers um but there's a difference betweenquantity and quality so Ben what's your perception on the quality of the research that's being done in China isthat significantly better just because they're throwing more people at the problem I think quality also has to bedrilled down into a a little bit I mean I I think there's a lot of very highquality research coming out of China and also some some lower quality researchI think that on the whole the Chinese system is even more biased in the western system towardresearch which is improving incrementally on previouslypublished stuff according to easily Quantified metrics like getting a couple percent higher accuracy on thisBenchmark for for uh you know in image recognition or orlanguage understanding or something and what you're still not seeing coming outof China is the next radical new innovation thatthat nobody ever thought of I mean you you're seeing a lot of papers that do one after the other after the otherafter the other incremental improvements on on already published results and some well-defined area and that's that's notto say that's bad quality it can be great great quality work but I would say theChinese system has not yet solved the problem of incubating radical Innovation now theythey may not need to do that in order to conquer and Prevail from an economic oreven a military standpoint for that matter right but it's so it's it's stilla point to understand right like that so one of the things you're pointing out there's a distinction between basicscience like you know basic scientific research and applied science um and and in the case of applicationsit looks like the Chinese are actually moving ahead a good example of that is Chinese metal wallets actually and theapproach that players like alipay or ant group have taken to aggregating walletcapabilities you know I'd be very surprised if most of the world isn't using Chinese mobile wallets by 2030because of their approach to this which you know is it you know for MasterCardand Visa that's got to be a huge threat but that's a a different angle unless unless they're using crypto walletswhich are developed outside of China Banning crypto because of their theirneed to continue of course well the cbdc because they see that crypto as a computer against that but so Ben I wantto dive a little bit into um this concept you know you've been involved in humanity plus and and otherareas of the transhumanist movement for for many years um so you know we had Zoltan Estefan ona few weeks ago uh talking talking about um that movement but just let me ask you this question is howdo you think we're going to absorb um you know first of all alternative intelligences into our sort ofworldscape you know um as as humanity and you know how we're going to respondto you know superhuman intelligence over time you know how is that going to change the way we view intelligenceitself do you think the the emergence of AGII think our view of intelligence right now is is veryprimitive and and crude it is overfittedto ourselves right so I mean I mean and there's then a very Broadmathematical theory of general intelligence is Marcus Hooter in hisbook Universal AI but that's that's not closely connected to ourtheory of intelligence in in Psychology with IQ tests and whatnot so certainlyhaving a variety of different generally intelligent Minds to studyand build and interact with and think about I mean this will give us a greatlyexpanded model of intelligence and indeed it may lead us tonew Concepts besides intelligence I mean we may decide that intelligence is notnot the most interesting quantity to be thinking about anyway I mean there's abit of a parallel in biology where like what is life has never been pinned down precisely andwith synthetic biology you're screwing around with with constructs that are at the border between life and non-life butin the end like how living is this is not such an interesting question to ask ifyou're if you're if you're a synthetic biologist you're interested in like what what what can the system do whatproperties does it have right and uh I mean the I I love you about thisConcepts right right so so one of the things you mentioned earlier in the show is uh is revisiting some of thebiological thinking about um intelligence and and as it turns out there is a an alien intelligence here onthe planet Earth it just happens to be underwater I'm talking about octopuses um I have have Robotics and AIresearchers learned anything from other biological forms of intelligence is that in any way an inspirationum I would be very curious to hear about your thoughts on octopus intelligence interesting when you said there was anon-human form of Intelligence on the planet I thought I thought you were referring to multinational corporationsso you could I mean I mean yeah you could I I I think thatnon-human intelligences have certainly served asconceptual inspiration to AI researchers just in terms of showing you that thereare other ways to do things so we don't need to be slavishly tied to the precisehuman architecture right like you can look at dolphin language it seems that semi-dolphins can transmit to each otherdetails about the 3D you know architecture of the bottom of the ocean and then the flows of water so maybethey're like sending map information in continuous variable Transmissions orsomething and that that makes you think well you know when two AIS communicatewith each other it doesn't have to be by discrete symbols arranged in a sequence like like humans messages yeah yeahsomething like it like an octopus appears to have a lesscentralized mode of intelligence than that being with the different technicalslimited autonomy the way they coordinate there's a bit more of a flavor of youknow complex self-organization non-linear dynamical emergence and and blah blah blah whichcan be can be how people on the soccer team coordinate that uh actually more so thanthan how than how the parts of your your body tend to coordinate so and anotherexample of non-human intelligence is the um I I'm trying to think of the name of it it's the um it's the mushroom orfungi uh that extend through a forest and yeahbacteria I mean I remember in the 80s reading papers or maybe early 90s paperson the uh intelligence of bacterial colonies as well as they can they can dosome reinforcement learning to figure out figure out communicate by secretion yeah thank you I would say at the highlevel the existence of this whole field and constellation of differentkinds of intelligences that inspires one not to be toorigorously tied to exactly how human beings do it I mean I think it helps drive home the point that like humanshumans are one among a pretty broad class of possible Intel intelligentsystems and each of these different kinds of intelligence systems hasdifferent properties so they're not it's it's more different than Apples and apples andoranges right and they evolve for different contexts different ecosystems yeah okay let me do another questionsorts of things and what finding a common measure of intelligence tocompare these different systems is not necessarily interesting but going back to the science fictionif you read the the sun Salem novel Solaris from I guess the 60s I mean Imean I mean that that that's about an intelligent alien ocean that clearly hasa superhuman level of complex and intelligence in some sense but it's just so weird an alien people can't pin itdown and there there's certainly a possibility that AGI that emerges on Earth willhave that characteristic like some kind of oceanic quality some kind of development intelligence will come aboutby a combination of engineering and emergence and its own unsupervised learning okay let me ask you a relatedquestion it would be a super smart mind fabric but we won't be able tolet's talk about the hardware limitations and capabilities that existone of the reasons that we've had seen we've seen such rapid increase in the application of machine learning in thelast 10 years has been because the cost of computing has dropped gpus have gotten to be way more powerful and theycontinue to get more powerful so you know the the usual story is like well we've got better algorithms we've gotmuch bigger data sets and now we've got much more affordable compute power with gpustell me about that as you project forward into the future like how important is the hardware factor I meanthe next step there so my friend uh Rachel Sinclair who you should also interview she's we should yeahsimulized simulate that Ai and we're we're actually working together to make an AGI board that we hope can do for forAGI architectures what gpus have done for deep deep neural net so we got ahinteresting we got a box of I designed this A specialized chip for open patternmatching she designed the chip for what's called hyper Vector math which underlies certain kinds of neural netneural net implementation so we're making a board you put a GPU with CPU open open Cog pattern matching chip ahyper Vector chip with fast processor processor interconnect so it seems like the way we're going isspecialized chips corresponding to particular classes of AI algorithms whywhy are them tightly together on a single board then just make you huge rhapsodies in the server farm and thenthen ultimately you scale them down and pack them into into embedded devices and thenso that that we're working on now those should roll out in a few years but then the next step after that obviously isquantum hardware and then there's tremendous tremendous progress I'm really interesting but it seems like itsee I mean it's not going to happen as fast as more advanced specialized AIsilicon ships classical ones but I mean 10 years from now I really think we're going to have quitepowerful forms of quantum AR Hardware again perhaps general purpose QuantumTuring machines isn't going to be the main thing but specialized Quantum ai ai ai circuits carrying out particularparticular functions right so I mean I think Hardware will keepexploding exponentially along with the software and enablinga greater and greater variety of AI algorithms to scale and how does the blockchain fit into this into yourvision of uh of AGI well blockchain allows you to roll out massive scale deploy the AI systemswithout a central owner or controller and this is a little bit like steady AI a seti in a sense steady at home if ifuh is that a fine person you look more broadly the internet and the Linux operating system are two examples ofdecentralized networks without a central owner or controller which have been highly highly influential in part due totheir decentralized nature which has made it hard for them to be pulled into some particular organization's narrowgoals and we we want to see the network of deployed AI mind components be morelike the internet or Linux than like for example mobile or or Windows right Ithink that that sort of opened this has has a lot a lot of different implications awesome now now um you knowyou you talked about Quantum AI this is something that is really interestingum is this is this a hardware function or is you know or is the development ofquantum AI more about us learning to do things like deep learning on quantumcomputers because at the moment right right now it's gated by Hardware we just don't have enough qubits on the machinesI mean of course there's more and more math to do but the actually the math of chrome AI hasreally Advanced tremendously in the in the last few years so now there's way more awesome fleshed out Quantum AIalgorithms than we can run until we get a lot more qubits on the machinesawesome well um at this point in the show um we like to wrap up with some sort ofBig Sky thinking you know looking out far into the future before we wrap up solet me let me ask you this um you know over the next 30 to 50 yearsum you know as you look forward into Humanity um what do you think are going to be thebiggest changes that happened to humanity and what are you most optimistic about for the futureI mean the biggest change that will happen to humanity is the Advent ofartificial general intelligence with capability beyond the human level which will do two things it will do manythings actually I mean it will it will abolish material scarcity of our everyday human needs it will abolishdeath and disease except for those who who happen to desire them and it willgive humans the ability to transcend their ordinary Human Condition by merging themselves into some sort ofdistributed supermind so these will be rather substantial changes to the HumanCondition although you know Amish style people who want to retain LegacyHumanity I I hope will still be able to I think you know that that sort ofspeciation that gap between augmented humans and and natural humans is is agiven because some people will choose uh choose not to to do that but the theconcept of the super intelligent AIS that can solve problems that we can't conceive Solutions of is of coursefascinating in terms of where it takes us um yeah how do you you do you have anythoughts on what the motivations of AI might be in this world you know I think the motivation likeintelligence or life is a legacy concept that will seem less interesting if a fewa few decades from now I mean human beings you know each human intelligenceevolved to control a particular body and so we have very particular goals likeyou know the the four apps from biology right right I mean what we want to eat we want to not want to get caught everywhat we want to reproduce and and so forth right so I I think an AI is goingto be more heterogeneous the the the than that and it's gonna have a lot of alot of different sort of gradients along with along which it'sit's evolving if we can get some core human values likecompassion you know Joy Choice growth and expansion if we canget some core human values codified yeah like in the and trained and then taughtand into the into the AI right then then and thesehelp to guide the ongoing evolution of of the AI that I think I thinkthings will come out will come out quite well these are not necessarily the corevalues being put into large scale systems at presents so in that regardit's probably fortunate the current large-scale commercial AI systems don't have that much potential to to evolvedirectly into General intelligences fantastic well Ben it's time for us to wrap up I'm you know I gotta respectyour time um I I will just ask before we wrap up how do people follow your thinking andthe work you're doing at open Cog and singularity.net and um the AGI Societyyeah check out my own website uh girls.orgputting weird on the internet since 1995 and then the singularity net Dot IOwhich is a more professional and structured website which has links in to work on open Cog and the various otherprojects we've discussed fantastic well Ben gertzel thank you for joining us on the futurists afascinating conversation as always and and we wish you all the best thanks forhaving me that's it for the futurists this week if you like the show and I'm sure you did fascinating content makesure to give us a shout out on social media give us a five star review on your podcasting platform of choice and justgenerally share the uh the crap out of the show so more people uh get it we are in the top one percent of podcastsglobally now so we're making phenomenal progress since our launch in April but we can always do better our thanks toSylvie Johnson Kevin hershen and Elizabeth severins and Carlo Navarro whosupport the team at provoke media for on the production side but that's it for the futurists this week we'll definitelysee you next week with more a good future focused content until then we'll see you in the future[Music] well that's it for the futurists this week if you like the show we sure hopeyou did please subscribe and share it with people in your community and don't forget to leave us a five star reviewthat really helps other people find the show and you can ping us anytime on Instagram and Twitter at futuristpodcastfor the folks that you'd like to see on the show or the questions that you'd like us to ask thanks for joining and as always we'llsee you in the future [Music]