AI: What is the future of artificial intelligence? – BBC News

should we automate away all the jobsincluding the fulfilling ones should weallow AI machines to flood the internetwith propaganda and fake news should wedevelop non-human Minds smarter than ourown machines that might one dayoutnumber us or outsmart us do we risklosing control now you might think thatsounds like some futuristic script froma Terminator movie but last month someof the most well-known figures who areinvolved in the development and trainingof artificial intelligence call for amoratorium until we better understandwhere we're going an open letter wassigned by thousands of entrepreneursacademics and scientists including ElonMusk who wants the training ofintelligence haltered for at least sixmonths we're going to dig deep into thisover the next 20 minutes or so in thecompany of two people who know a thingor two about it joining me is the techinvestor even burfield the Evan burfieldthe author of regulatory hacking Aplaybook for startups he's in Texas I'mProfessor Gary Marcus is in Vancouverhe's the professor emeritus at New YorkUniversity and author of rebooting AIProfessor let me start with youum clearly with such advances that we'reseeing we have to set some guard railswho do you think should be in charge ofthatuh were we both professors I think thatwe need Global governance for AI I thinkthat we have a lot of patchworks rightnow almost balkanizedum the worst case from the company'sperspective and the world's perspectiveis if there's 193 jurisdictions eachdeciding their own rules requiring theirown training of these modelsum each run by governments that don'thave much specific expertise in AI sowhat I called for an economist editorialearlier this weekend in a TED talkearlier this week was to have a globalsystem modeled on something like theinternational atomic energy Authoritywhere the world comes together and sayswe have a new threat here but it'sreally a new set of threats and we needto work together on this so I think thenumber one thing is it should be Globaland the number two thing is it can't bejustum policy but it also has to be aresearch side because we need to inventnew tools like we had to invent forfighting spam and cyber warfare and soforth there's so many different threatsas you mentioned around misinformationcyber crime and so forth so we need tohave a kind of standing organizationThis Global and well-financed to try tobuild tools to mitigate those threats soEvan burfield there are many many peoplewho who just want to press the pausebutton until we work out some of thesethings but I can already see and I'veheard uh the reasons why that probablyisn't possible and and that is becausenot everybody will stop and people areworried about losing competitiveAdvantage so how do we best go aboutthisyeah I think that's exactly rightum you know the the challenge with amoratorium is that it's incredibly hardto enforceum the responsible actors would be morelikely to follow it the irresponsibleactors wouldn't but that's actually notso much my concern with the moratoriumthere's absolutely questions we need tobe asking about the governance of AIabout what industry can do whatgovernment can do uh I think the letterdid spark a conversation Schumer isworking on a new AI bill here in the U.Srumors McCarthy's working on arepublican versionbut what I think is actually much moreimportant is to start to have theconversations about how we prepare oursociety our economy uh our politicalsystem democracy itself for all of theimplications of AI that are coming oneway or another and I suspect we'll see ayear from now we'll go this was lessimpactful than we thought five yearsfrom now it will be an absolute tsunamiof upheaval and we have this windowright now where we can have thisconversation and we can get creative andI think we've got to use it and amoratorium gives us this false sense ofsecurity that we have control and canstop it versus figuring out how we ridethis tsunami and try to direct it in amuch better Direction Professor Marcusdid wedid we learn anything from the the lasttechnological Advance the advance of theinternet of social media are therelessons from that which we let's face itwe didn't do very well that areapplicable hereI think the number one lesson is youdon't want to close the Barn Door afterthe horse has left I think you knowwe're very late in in figuring out whatto do about social media I think weprobably handled privacy alone in thewrong wayum we wound up with so much polarizationand hostility we wound up withmisinformationum I think we waited too long to act Ithink the number one lesson is we shouldget on it right now and I agree with theother panelists that the moratorium Youcould argue about the merits whether itwas the right thing or the wrong thingit was absolutely the right thing toraise this and get it on everybody'sagenda this is not something we want sixmonths from now something we need now soEvan burfield when you talk about atsunami in five years time what doeswhat does that look likeuh look I'm down here in Austin Texas uhat Capitol Factory startup accelerateI've spent the whole day uh you knowmeeting with startups and there's not astartup right now out there that is notapplying these AI generative modelsthese large language models to everyinteresting problem of the Sun andthere's all of the the scary dystopianpossibilities that you uh led into thissegment with but there's also uhincredible advances in how to make workmore fulfilling and more impactful howto apply uh tremendous personalizationto Medicine based on our genetics ourenvironment the particular issues we'rehaving how do you make government moreresponsive and feel more like aconcierge to Citizens all of that isalso being worked onum and I think figuring out how we putthe guard rails in place around some ofthe scarier things which isn't justabout regulating Aya it's about changingour social policy changing our Marketpolicies themselvesso that we can mitigate some of that andand direct this into the the much morehopeful and optimistic Direction why whyon that point though Evan is it is itimaginable in the current scenario youare around it all the time that aresearch lab would cross a critical linehere without even noticingI I'm I'm personally skeptic you know uhGary's written some wonderful pointsabout the fact that we are we are veryvery far I believe from artificialgeneral intelligence and the Terminatorscenarios I I think we've got to be veryaware of right now is simply that thistechnology is already right now today ata state if it did not Advance anyfurther where its application is goingto profoundly change how we live ourlives how we work how we engage witheach other in communities how ourdemocracies function uh the impacts onour democracies are going to be feltright in the 2024 political cycle herein the U.S that's what I think we needto be talking about and preparing forthe scenarios of AI is like nuclearweapons we have to ban it immediately Ithink are much less applicable to thethe much more realistic uh changes thatare already happening around us rightnow that are going to accelerate milesum you've just come back from Washingtonand I know that you've been talking topolicymakers about the specific issuesin fact the reason we're talking aboutit tonight is because you tweeted no onehas a clue I I mean is that is it asblunt as that that nobody reallyunderstands it there is practically nowork being done on ithi Christian you're you're spot on thethree biggest challenges right now withpolicy makers are one this wascompletely foreseeable there were someof us in Washington talking about this10 or 15 years ago policymakers weren'tpaying attention and most of the thinktanks in Washington really failed tostart a conversation about the Practicalthings that needed to be done to preparefor the age of AI so we're behind theball from a policy making standpoint thesecond thing I would emphasize what Evanburfield just said there is a wavecoming and you can do two things when awave is coming you can get crushed by itor you can ride the wave and to useanother analogy right now the discussionin Washington is about whether to putthe genie back in the bottle or not thatshouldn't be the discussion it should bewhat three wishes should we ask thegenie and that's the discussion thatshould be had about how to handle Ai anduse it for good purposes and finally theother problem is policy makers are notthinking two steps forward on thechessboard it's AI right now but in inthis decade AI is going to besupercharged by other Technologies likeQuantum Computing that are going to givemachines genuine human-like emotion whatare we doing to prepare for that weshould be having that conversation nowthere needs to be institutions inWashington that focus on that so Jackwe've had a discussion about that inthis country and the UK government hasdecided that it doesn't need a dedicatedUK regulator for AI so who's overseeingitthat's a very good question I mean uh II was on stage last night with thechancellor Jeremy Hunt and I asked himabout this you know he's the guy incharge of the UK economyum and he was really quite dismissivehe's he in in the sense that he said youknow this is something that is going tohappen and we have always embraced newtechnologies in this country and weshould do so again uh it's full steamahead was the phrase that he usedum you know he was very very positive hedid he did not want to talk about thepossibility that people would lose theirjobs because of this technology he onlysaw it as a purely positive thing and hewas not keen to talk about the way itshould be regulated now you know I'm noexpert on this still if I'm a politicsguy but what I do know is that is howWestminster works and how um politicalsystems work and I can tell you now andyou'll know this Christian there is noway our political system is set up todeal with this challenge absolutely nochance the speed at which decisions aremade in Westminster and I suspect inother major political centers is far tooslow to cope with the pace at which thistechnology is coming the policy makersdo not understand it at all this is justsomething that is going to wash over usand we're going to have to cross ourfingers you know the UK government putout a white paper which is what theycall their draft strategy on AI theother day I mean just the very name ofit white paper tells you how old schoolthis is yeah you know it's out of datealready and that thing took you knowyears for them to put togetherum we just don't have the sort of nimblesmall system smart thinking people setup to deal with this and I'd be verysurprised if that's different in the USor indeed in many of the other big Powercenters well they clearly don'tunderstand it Professor Marcus do theycall you in to try and get you toexplain it to themthat I was talking to people in the U.Sand Canadian government yesterday I'vebeen called a lot latelyum I think there is an awareness thatpeople don't quite know what to do andthey are increasingly turning to meum and also turning to all of my youknow academic colleagues and so forth soI think that there's at least arecognition and people know what theydon't know I do think that the UK whitepaper saying that you won't have acentral office of AI is certainly forall the reasons that were kind ofimplicitly just said which isum the government is going to beill-equipped to deal with the speed ofthis and if you just leave it to 20different Regulatory Agencies Each ofwhich don't have expertise you're askingfor trouble you're asking for a lack ofcoordination and it's just not realisticthat all of those agencies are going tobe up on things so there needs to be Ithink at least some Central oversight Ithink the United States should considera cabinet level AI officer and youshould consider something comparableum you need some people maybe like a G7then we have a G7 meeting of foreignministers we need a G7 meeting of ofIAAI ministers is that is thateffectively what you're saying well Imean I'm calling for something similarwhich is a global organization uh kindof like the IMF or an internationalatomic energy agencyum where you have a lot of experts youhave a lot of people in government youhave a lot of people in uh the companiesand yeah you have regular meetingsyou're like well this week the new thingis this thing called this is a realexample called Auto gbt where you haveai's training other AIS what do we doabout that how big a threat is it is ita small threat big threat like if youhave a research arm then you can saylet's do some experiments here and tryto figure out what the limits are rightnow instead you have like 193 countriesmaybe some of them have read the newsabout this major news Discovery some ofthem haven't even aren't even aware ofit and there's like no coordination herethat just can't be the right way you'renodding Evan because this is the keyissue it's miles as miles discussed it'snot human competitive intelligence it'sit's what happens after AI gets smarterthan humor intelligence right amazingbut you know I I can't go to aconference I I actually live inWashington DC most the time I can't goto a dinner or a conference a meetingwithout the word I being discussed andthey're all talking about chat GPT andGary's right is that even Auto gpptthere was a an experiment run last weekcalled chaos GPT where they took aneutered version of Auto GPT and told itto go out and figure out the mostefficient way to destroy Humanity it wasa it was sort of a test and it's set towork doing it there's there's a lot ofthis stuff is moving incredibly fast andfiguring out how you can educate policymakers about how to mitigate regulatebring transparency to some of thosethreats while not preventing what can bebreathtaking advances inum how we live our lives in much morefulfilling and purposeful ways andSociety I think that's that's a lot ofthe trick here to Echo Jack's pointthough about white papers and the wayyou know government moves I I tend toagree you know miles may be moreoptimistic than I am but I tend to agreeI think a lot of the big changes thatare going to need to happen probablywon't happen until uh there's some sortof provoking event some sort of Crisis Idon't think that though prevents us fromstarting to have the conversations atleast the Way Washington tends to workat least you want to have the the policycontainer the framework the ideas readysome sort of consensus being built sothat when the opportunity presentsitself kind of like a a VC who sees agreat startup right when the opportunitypresents itself you're ready to jump onit you're ready to move forward and Ithink that that has to be happeningright nowgo aheadthe opportunity that I see right now isto build some Global governance I thinkyou have the governments are afraid ofthe technology companies the technologycompanies are afraid the governments aregoing to shut them down as they did inItaly and this means everybody has someincentive to go to the table that's rareand I think we should be seizing thatopportunity right now to try to dosomething coherent that is dynamicenough to cope with the speed of thechange to take advantage of the goodthings and and to avoid the bad thingsbut we need that coordination now and wecan't just leave this to the usualmechanisms it's just too slow miles oneof the more worrying things that yousaid was that speaking McCarthy waslooking at an AI for for republicans andyou know one of the experiences we haveof recent years is that the Russianswere able to interfere in a democracyand who knows arguably it's been debatedwhether they were able to change some ofthe results through what they wereputting onto the internet I mean we'reinto a whole new ball game for democracyif AI can put out misinformation andpropagandait's not if it's how much and when it'sgoing to happen probably in West 2024election wow yeah there's no question Imean this this coming election cycle inthe United States it's a big concern forelection security authorities it shouldbe but I yeah I got to go back to whatthe other panelists said in order torespond to it effectively we've got tostart with education and right now Imean I've tried to brief policy makerson this it's like explaining particlephysics to a chocolate chip cookie Imean there's just not recognition aboutwhat's happening if I was President JoeBiden right now I will put I would putthe entire cabinet on Air Force One Iwould fly them to Silicon Valley and wewould spend the week educate educatingthem about what's happening becausethere aren't just these securityimplications for the elections as Evannotes there's also really positiveimplications I mean there's the abilityto address major health care problemshunger homelessness and to do it in realtime and we are missing someopportunities by policy makers not beingeducated on the subject but of coursesecurity has to come first and in orderto protect elections or anything elseit's got to start with you know policymakers becoming technologists beingeducatedfascinating conversation we're gonnahave to leave it there Evan burfieldGary Marcus thank you very much indeedfor joining us