Zum Inhalt springen
L

Das Video kommt von YouTube: erst beim Abspielen verbindet sich die Seite mit YouTube (Google).

Too fast to process? The global challenges of artificial intelligence • FRANCE 24 English

FRANCE 24 English42:36 11.391 Aufrufe veröffentlicht Auf YouTube

Das Wichtigste aus dem Video

Tipp auf eine Zeit – das Video springt genau dorthin.

Transkriptautomatisch erstellt · 267 Zeilen
Herunterladen
  1. [Music] who knew for me it was during the tedium of a 1986 summer job in the Paris suburb
  2. of buan bianu as I punched citizens names into the city school board's first ever computer network who knew that a lowly data entry gig made me an
  3. unwitting foot soldier in the transformation of capitalism and perhaps Humanity from two fingers on a keyboard fast forward to the presence in the age
  4. of terabytes and artificial intelligence it's France and India a co-host an AI Summit in Paris we ask if it spun so far beyond our control that any
  5. well-intentioned bid by world leaders to forge common rules and guidelines seems well too little too late right now it's Silicon Valley that's got the money to
  6. draw the best best and the brightest and the cheap energy to power the bigger and bigger data centers needed for AI add to that a new US Administration that's
  7. ready to start a trade war with all those in its path then how does the rest of the world defend privacy shared natural resources including strategic
  8. minerals in places like China and Africa and what to do about a digital age that so far seems to concentrate so much power and wealth into the hands of so
  9. few more broadly is the techn technology evolving faster than Humanity can process its potentials and its dangers We'll be asking our panel today in the
  10. France 24 debate the global challenges of artificial intelligence and joining us from the AI action Summit the host of our Tech 24 program Peter obrien it's
  11. day one thanks for being with us you're welcome he's a mathematician by training who's advised president m
  12. Jer kulong is president of the uh po technique Institute of Paris you hosted the uh I guess you could call it the pre-summit last week which brought
  13. together some heavy thinkers already thanks for joining us thank you uh Bangladeshi American data scientist ruman chowri is the CEO and co-founder
  14. of Humane intelligence tell us what Humane intelligence is we are a tech nonprofit that's building the community of practice to understand the impact of
  15. AI on society all right thanks for joining us thanks as well to man has been covering AI for Fortune magazines since well before it became a thing
  16. Jeremy KH the author of mastering AI as Survival Guide to our superpowered future welcome to the show thanks for having me uh by the way you can listen
  17. like And subscribe to the France 24 debate uh wherever podcasts are streamed uh ruman what does AI mean for you well artificial intelligence has been
  18. compared to many things electricity oil so many things I think just for the average person artificial intelligence represents the next wave of
  19. technological Evolution things that will in Impact every aspect of our lives and in fact AI is really used in many things today we just don't see it so for the
  20. regular person AI may change how you live work and play how you live work and play so it changes us as humans I don't know if it changes us as humans I
  21. actually think the human condition has been actually fairly Universal across time this is why Shakespeare holds so much meaning for us even though here we
  22. are in a fancy studio with electricity and that's something unfathomable in this time I don't think it changes humans at all all right doesn't change
  23. humans uh but there are these these concerns and you know this worry that TI Kul that uh and a lot of the talk over and there's been plenty of media
  24. coverage in this country over the last 72 hours ahead of this Summit yep a lot of the talk has been explaining what artificial intelligence is
  25. and should we be reassured ing people or should they worry well first we've got to understand uh things before we worry or we are too
  26. too optimistic so the people at Institute po technique the Pari they they are part of the people who did artificial artificial intelligence The
  27. Institute poly technique the par is made of six engineering schools that started some of them in the 18th century some of them in the 20th century we are trying
  28. to turn them into an single institution adapted relevant for the 21st century but what they have always done is to keep Pace with the scientific
  29. revolutions and to anticipate them so we are strong in math and stats and computer science so that's what we are training uh students to to to do and I
  30. think we we can be adapted to to to these exciting times you're figuring out the how but do you think about the why yes indeed because there are social
  31. impacts indeed uh and we also have social scientists I I guess the it's too early to say but it's not too early to think about the
  32. potential consequences and the way we can regulate things uh Jeremy Khan we we heard the French President um who invited disruption in the speech he gave
  33. a a short while ago uh disruptions become something of a dirty word for some when they when they when they look at this yeah absolutely I mean I think
  34. people are people are worried about what this technology is going to do to society to to to their jobs uh you know people are very concerned about uh you
  35. the potential impacts of this it is and I think it it is right to say it's a disruptive technology uh I think it will change many things and and be fairly
  36. transformative I think there'll be some very good changes and I think there are some risks that we we should be mindful of and try to take action now to
  37. potentially avoid or mitigate all right the uh it already feels that uh the information age has begot the disinformation age a lot of the focus
  38. has been on this aspect of it uh among the general public that doesn't see what ruman was telling about which is uh the applications that go beyond asking chat
  39. GPT about a specific topic Eliza Herbert has that story Joe Biden and Donald Trump closer than ever or Emanuel macron embracing
  40. Marine Le Pen as deep fakes grow more and more prolific so do the abuses of artificial intelligence in just a few clicks an
  41. image can be falsified and virally spread across the internet chat Bots too are adding to disinformation according to experts
  42. programs that impersonate real people can be a dangerous digital tool especially when generating myths or rumors around election
  43. campaigns no important questions that we should tackle the future of the workforce of course the impact on our society on our democracies automation
  44. threatens to replace jobs like customer service retail and many other professions generative AI can clone voices used in impersonation scams and
  45. services such as chat GPT capable of producing text image and sounds directly endanger artists livelihoods copyright and the collective imagination
  46. how can we fail to see that this is the Trojan Horse for giving up the use of our most fundamental faculties another issue is on the battlefield militaries
  47. already use AI in Warfare for Mass surveillance and with autonomous drones and recently Google dropped its ban on using AI to develop weapons it could
  48. also be used in much more dangerous and and harmful ways uh example um kill anyone uh who fits the following description bias and
  49. discrimination privacy and data leaks and the environmental cost of the heavily energy dependent services are a few more of the risks that come with
  50. AI yeah I guess uh Peter Brian we could have entitled this show what could possibly go wrong uh the I I I I suppose these kinds of worst case scenarios
  51. spelled out there that's not what we heard from the French President earlier but it's high on everybody's
  52. mind yeah and I think one thing that went a little bit under the radar was that quick fire round that he did in his interview on Sunday nights where he was
  53. asked the question is AI a threat to our democracies and he hesitated and he finally did answer yes in fact I've read that that speech he made on S Sunday
  54. night as a classic macr on them top right he's supporting the AI act he actually even called for Global regulation of AI but at the same time
  55. clearly it was a massive um effort to drum up investment and that's absolutely absorbed this entire first day of the summit this headline that he's drummed
  56. up 109 billion EUR it is actually on a similar level to the kind of promise that we saw from Stargate in the US um but some
  57. definitely that I've spoken to at the sum feel like the other concerns around safety around risk have been sidelined while macron kind of rolls Back The
  58. Years and plays the best hits of the stup nation yeah let me let me bring in Roman on this can you it there are mixed messages and we heard him earlier from
  59. the French president on the one hand he's saying day two we're going to talk about regulation but it's day one come invest in France come invest in Europe
  60. right and and we're seeing that shift actually across the board even a lot of the people who called for safety and we're concerned about existential risk
  61. today are saying we should innovate I think there's two things being reflected here first is there's immense commercial opportunity and there's a lot of
  62. pressure from big and powerful companies for these governments uh you know to be aligned with them on the other end I do think one of the reasons we don't talk
  63. about AI abolition is that even for those of us who work on societal impact we do want to see a better Vision where artificial intelligence is used to
  64. improve human lives want to see how it's improved uh uh the lives of humans uh were at the same time um uh worried about uh who who has control and we
  65. heard this from the French President uh in that uh Sunday night speech uh that he uh that he gave uh to uh that he gave uh to French uh television and to uh
  66. journalist who had come from India uh Emanuel mcon uh talking uh about well uh the uh dangers that lie ahead and talking about the United States in
  67. particular uh raising concerns he says with its Tech dominance it affects everyone's life yet it's controlled by only a few and that's
  68. where the problem begins when tools that billions rely on to learn work treat others and shape their lives on the hands of a small group private interests
  69. without a global conversation about what's right where the limits should be and how to ensure access for all then we're not doing our job
  70. properly now Peter O'Brien uh the French President there doesn't say the words the United States but it's
  71. implicit everyone is looking forward some with trepidation I have to say a lot of people here in Europe for JD Vance's speech tomorrow if anyone tells
  72. you they know what he's going to say then don't Bank on it um we know he's from Silicon Valley we know he's um criticized the EU a lot in the past for
  73. overregulation but that tends to be about what he calls free speech concerns his his speech comes slap bang in the middle of Emanuel macron's tomorrow
  74. morning and uh you font allance um so that's if I can put it this way sort of unusual bedfellows um but the truth is that America is a huge concern for a lot
  75. of people working in AI regulation because he has W he has ripped up Joe Biden's um executive order putting safety guard rails around AI he's fired
  76. the head of the US's AI safety Institute France meanwhile has just started its own AI safety Institute that's one of the things to come out of the summit um
  77. but it's looking increasingly like it's the us against the world in AI as well as many other things in many other things and uh you spoke with the vice
  78. president of the European Commission on this yeah that's right she's here at this event at the Meson lashi which is just we trudged through a muddy
  79. construction site from the Guan palal to get here and interview her and first off I asked her is this event actually been a little bit too much about investment
  80. and a bit too much about the French of course we have one uh common single Market in the European Union so we want to make sure that we have also
  81. that kind of framework in the European Union that encourages Investments and Innovations when it comes to AI you know what JD Von's comments are like on the E
  82. EU is he going to come in and throw a grenade tomorrow in all of this of course all the countries they have their their own rules when it com
  83. for example the digital sector and other sectors and in the European Union we are very committed to our our rules we want to make sure that we have fair and safe
  84. and also Democratic environment when it comes to digitalization was Italy right to take action against deep
  85. seek all all the companies who are doing business and operating in the European Union of course have to respect our digital rules and it also comes to data
  86. protection so that is now something that is uh nation National authorities are doing and checking it for example with deeps is it not problematic that we're
  87. getting very different approaches from different data protection authorities from within the EU when we look our digital rules uh it would be better if
  88. we have always one uh uh one approach in the European level and we know that with the gdpr we have different ways of implementation in different member
  89. states and of course it's never good if we have that kind of fragmentation in our markets
  90. that's the eu's tech boss saying there that actually the EU needs to take a common approach for for AI worried about things splintering apart further yeah
  91. gdpr she mentions there just to tell our viewers that's part of the the the rules uh for uh uh Tech transparency put forth by by Brussels that uh US tech companies
  92. have to abide by and as as Peter O'Brien was saying might be in the crosshairs of that speech we hear from from JD Vance your thoughts on on the US vice
  93. president at this Summit and what you're expecting from him you know I'm a I'm a mere Citizen and I as a citizen I just think that Europe might find the proper
  94. balance between uh Innovation and regulation we have a model where the citizen is protected and we want to innovate yeah because we don't want to
  95. be the referee in a football match between China and the United States that's right that's right and the only way to do this is to be an
  96. and we are for once we did not miss the train I mean we France and Europe we are in the train I mean there is a lot of uncertainty uh who knows what will
  97. happen but we are players uh and I would like to advocate for two things we uh higher education institutions can do first education I mean we have to train
  98. people to AI we have to to train them to be critical and aware and that will help uh and the second thing is let us speak a little bit of the the potential uh
  99. upsides of AI and and the potential positive consequences uh every scientific presentation I hear these days whether in biology chemistry
  100. physics uh uses AI with a power that that is incredible so as far as you have to to test molecules and to test proteins and so so on so forth I mean
  101. you are going to to the the the the very scientific methods are changing for the better and well that's that's worth it in isn't it when we speak of diagnosis
  102. of cancer right and Jeremy Khan you you heard there t Kuno alluding to the fact that uh France has its own AI Champion
  103. it's a company's one of its co-founders in Alum of two two of them excuse me of of of poly technique uh and uh it's called Mistral uh and and uh your
  104. thoughts uh I mean Mr Char's are great it's a it's a good company um you know they're good Founders and they've put out some good models um I think the
  105. issue has been a question of scale um the model sizes that they have put out are not as big as what you've seen from sort of open AI or anthropic and I think
  106. there's this issue of of what is it going to take to sort of get to the next level of capabilities uh for a while it looked like to get to the next level we
  107. was going to take these massive data centers uh which is why open AI has done this partnership to to build Stargate um it's not clear it wasn't clear that MRA
  108. would have that kind of capital to build something that big now there's some debate in the field actually about whether that's really necessary uh there
  109. are these models uh that are coming out now that are smaller uh that do very well deep seek is a smaller model uh this is China's version China yeah this
  110. is China's uh this model from a a Chinese startup it's a very Innovative model in a lot of ways one of the things that's Innovative about it is they
  111. showed that you could take a smaller model and get it to do these kind of reasoning steps that are very similar to these very big models and so that's
  112. interesting and it's possible that mrr could take you know borrow some of those Innovations from from Deep seek every the whole world sort of studying this
  113. model it's an open source model meaning anyone can download it anyone can examine it they publish quite a lot of information about how they engineered it
  114. um and the whole world can kind of learn from that and I'm sure the guys at mrr are looking at that and thinking are there things here that we can borrow so
  115. if uh something like deep seek is open source does it need to be banned emanu mol says he's against that for now you heard the the hesitation there in
  116. that clip that Peter O'Brien played us of the Texar at the European commission yeah well the model itself you can take the entire model and run it on your own
  117. equipment and then there's very little risk to privacy or or security the issue has been if they also deep seek also offers the chance to interact with the
  118. model through their own servers so then you are talking to a system in China and there there has been concern about what what is happening with that data and
  119. does the Chinese government potentially have have potential to to access that data and there's some relationship in that service with China Telecom which uh
  120. deep seek has not you know was was very quiet about that seems that there is perhaps some relationship there so there are concerns if you're using the app
  121. that you would download on your phone then you're talking to a server in China there there's some concern but if a company wants to take the model and run
  122. it on their own equipment there's much less concern Roman chowri what's your religion when it comes to the uh the nationality of these uh of these large
  123. language models right well first in talking about open source look open source is not new it was not invented with AI open source has existed for
  124. decades and actually has been net good for competition we have to ask ourselves in a market like AI that is almost naturally a monopoly or an oligopoly
  125. because it is so Capital intensive to build models don't we want to also spur Innovation by allowing small you know smaller actors to thrive as as Jeremy's
  126. pointed out deep seek and even you know uh meta llama models allow you to download and utilize these models for yourselves or even to make a little
  127. startup so I I think you know we have to look at the actors who are saying open source is bad and how they might benefit from it what we want is a competitive
  128. market cuz right now the concern more broadly just going Beyond artificial intelligence is uh the uh the most powerful control the secret sauce
  129. whether that's algorithm RMS or uh whether it's the the model for for making uh artificial intelligence that's right so my nonprofit human intelligence
  130. we work on test and evaluation uh We've pioneered things like public red teaming it's where you bring in people to evaluate models now unfortunately that
  131. can only go so far because today the model owners write their own homework they grade their own tests and then they publish their own findings telling the
  132. world about how great their models perform even when you think about all of the news and the hype around deep seek versus open AI what you have to look at
  133. are the metrics by which they analyze performance now it's things like medical exams legal exams physics questions and coding questions very little of those
  134. questions were about impact on society May might it perpetuate discrimination uh it does it hold up against cyber security standards so as long as we let
  135. companies write their own tests and grade their own homework we're really not understanding what's happening so whose job is it to enforce
  136. accountability uh I think it's an ecosystem I think government plays a role but as we've talked about the United States that can
  137. also sometimes be questionable that's over for four years at least well we'll yeah to be determined but I I think your prediction is probably correct this is
  138. also why we need a robust Civil Society technical organizations and also I'm a big proponent of independent algorithmic Auditors one of the things that I and
  139. others are pushing for in the United States is something called Safe Harbor it exists for ethical hackers as it relates to software and I think it needs
  140. to exist for people who want to understand the impact of algorithms and not be sued by companies can they do that can they take on Silicon Valley we
  141. can try your thoughts on this jerman yeah I look I think that it's it's going to be difficult with the current Administration to see the US government
  142. doing something on the federal level anytime soon but there are states in the US that are trying on their own to pass rules including some states with quite a
  143. lot of power there was a bill in California that uh got to the governor's desk it was vetoed but there's going to be a sort of second attempt or third
  144. attempt in California to pass legislation there's some legislation in Colorado and those things can actually have an impact and the companies don't
  145. love it because they'd rather see you know One National rule that it makes it much easier for them to comply to uh but they you know these State efforts can
  146. actually have impact we've seen that with privacy the state of Illinois has a very uh strong data privacy and biometric Privacy Law it's a state law
  147. but it's had National impact and so you might see those kind of efforts actually hold you know creating some accountability for these companies this
  148. kind of Civil Society approach that's an apem to what France is all about we're a very top- down country that's right nevertheless yeah we try to cure
  149. ourselves and and I I I like the the idea that maybe we are not going to a super intelligence uh ruled by somebody but to a a network of intelligences that
  150. uh that would interact uh between themselves and that could be controlled or or we could be critical about about them I also love the idea of open uh
  151. open source uh yesterday I was at a meeting of uh so called AI Alliance and when you see that IBM and meta are in favor of open source source because they
  152. have understood that it helps competition well that's that's good news we're actually a member of AI Alliance as well yes so there is this other issue
  153. which is we don't yet know how much energy all of this Computing is going to require uh Emanuel mcon uh talking up that big investment that Peter O'Brien
  154. was telling us about um and uh more and more let's take the example of the United States it's comp contemplating the opening of its first nuclear power
  155. plants in decades studies predict they'll consume up to 9% uh uh artificial intelligence will consume up to 99% of the country's
  156. energy by 2030 here in Europe data already consumes data centers already consume a whopping 3% of electricity and that could climb to 8% by 2030 Peter
  157. O'Brien that's the equivalent of uh the whole electricity output of uh a country like Spain of course a lot could happen between then but still uh this requires
  158. a lot of of energy and uh a lot of the protesters are about that yes there's a few things here one is that the kind of efficiency saving we
  159. saw with deep seek um might show that it doesn't have to be an exponential increase in energy usage another thing is you're totally correct depending on
  160. the country there will be a different environmental impact um I think one of the main reasons that macron has drummed up this support for France is because of
  161. its abundant nuclear power because a lot of companies don't want to be seen to be pushing out loads more CO2 in the atmosphere so if these kind of data
  162. centers are sent up in other countries that are more reliant on hydrocarbons there might be a lot more emissions in fact I was talking to Stuart Russell who
  163. um is a scientist at Berkeley he's basically he's literally written the book on AI uh the AI textbook which is in more than 1,500 uh un ities around
  164. the world he's got an interesting take on this he says look at the statistics we currently have at its current rate we can't and this you know is far from the
  165. consensus that we've seen emerging in some of the media and a lot of the people talking about AI he says that at the moment it is a tiny fraction of the
  166. world's carbon emissions AI data centers but he's someone who believes in AI potentially having an exponential increase in terms of its resource use so
  167. so he said in that case if it does continue on an exponential path then we do need to start to worry but as I said earlier deep Seekers potentially a bit
  168. of a counter example as to some somewhere where efficiencies can be made yeah because the planet's resources Roman chadri are
  169. finite and I'll also add that deep seek is built on uh reusing essentially a fundamentally uh you know bigger trained model so deep seek is able to be
  170. efficient because existing Foundation models exist so we couldn't have one without the other uh you know I'm not a climate and AI expert however I Channel
  171. my friend Dr Sasha luchon climate lead at hugging face and she's been doing amazing work on things like energy star ratings for AI models and it's also
  172. important to differentiate what we mean there's a difference between the cost of energy in training a model which is quite High versus you know and that's a
  173. sunk cost versus performing an individual task asking chat GPT 10 questions right if chat GPT is already even built and then we're distilling
  174. models on top of it that's what the process for deep seek is called distillation uh is that cheaper easier and better but I don't think it it gets
  175. rid of the question of what's the climate impact of AI models just because it's less than what we use today we know the planet is burning we know that the
  176. environment is declining right we know the environment is declining and uh like you're saying the the these these large language models riding piggyback and now
  177. we have this ambition right the Indians have been talking up the fact that uh they want their large language models to be not just in English but also in all
  178. the languages of the uh of their country that too is going to cost money uh absolutely and I believe India's launched a competition uh to for for
  179. homegrown Foundation models and my understanding is that the government is hosting this competition to uh spark their startup ecosystem in what's called
  180. Sovereign AI it's a it's a phrase I heard quite a bit uh in my international travels as us science Envoy in the Biden Administration science Envoy for AI uh
  181. Sovereign AI is a big topic of conversation people want their own homegrown models uh for things like data protection privacy security but also
  182. cultural context linguistic completeness uh what people don't want is sort of a a cultural encroachment and values that are different from theirs yeah CU
  183. already capitalism has killed a lot of languages uh people uh in France for instance don't speak with regional accents as much anymore speak local
  184. dialects because they all watch the same 8:00 news when I was growing up now with artificial intelligence is that accelerate that homogenization of
  185. culture that's definitely a risk um and if we don't do things like building Sovereign models or more local models it's definitely a risk that you could
  186. get more homogeneization um that is something I think people need to be aware of on the other hand whether it really makes sense for everyone to
  187. train their own large model in every case I'm not sure it does I'm not sure it's the most sort of you know we're talking about environmentally
  188. sustainable path it's not clear that would be necessarily the most environmentally sustainable path either um I but I think there's there's ways to
  189. sort of use these models where you could counteract some of some of those those Trends um you know you could the models actually have a fair bit of uh knowledge
  190. they they sort of suck up the world's internet uh worth of data and within that you can you can sometimes take a model and fine-tune it to perform fairly
  191. well a local local task or a local language it's much harder for languages that are not digitized in anyway so-called low resour languages there
  192. it's a much bigger problem because there just isn't Digital Data on those languages and there you might have to do things but there's some ways AI can even
  193. help there so you could do um for instance uh AI is very good at doing taking voice to text so one ideas if you can create some kind of um chatbot or or
  194. voice interface for people in low resource countries you actually have a way of digitizing their data you can actually get them to tell sort of their
  195. traditional stories uh to something that then can take that and create digital uh documents digital artifacts that can then be used to train a model but right
  196. now we don't have the the data for some of these low resource languages to train effective models is that something that um yes indeed we did not mention so far
  197. very much the the the issue of data to to to have a good artificial intelligence what do we need we need computing power we need the talents qual
  198. qualified personal we need a business model but we also need clean data and interesting data and that's why imagine all the library are are connected all
  199. the the the incredible works of art or literature of every country is is is treated by AI that would be fantastic and indeed you have to collect data in
  200. some parts of the world of the world where no digitalization has taken place fantastic or is it theft because a lot of these large language models rely on
  201. taking everybody's personal back yes what that that's what's extremely interest interesting in the current times what should be open what should be
  202. Pro proprietary what should where is privacy I mean that's a in in his speeech uh uh Peter O'Brien in his speech mcon talked about copyright but
  203. what was he what was he meaning by that copyright is one of the issues that's among the most thorny because one
  204. of the outcomes we have from the Summit is a copyright Charter that's been signed by a lot of Publishers and a lot of um sort of artists Representatives um
  205. writers Representatives musicians Representatives around the world so a lot of them have signed this Charter calling for protections of of copyright
  206. and there's a couple of things in here it's it's it's not just about um how we make sure people are fairly compensated for their work their work essentially
  207. being data mined but also what happens to stuff that's generated by AI is that then under copyright it does it fall under the copyright of the person who
  208. produced the work so in terms of the outcomes of this Summit it's notable that none of the AI companies are signaries of this Charter so that's not
  209. going to be enough to get them to change if for instance they have stolen stolen data um we saw actually last week the US copyright office released its kind of
  210. opinion if you like on the latter Point whether stuff produced by AI is falls under copyright and they essentially said um you know in the vast majority of
  211. cases no it doesn't because it's been produced by an AI therefore it shouldn't come under under copyright and they said actually for now copyright law is strong
  212. enough on its own to be able to deal with lot of the questions around um around Ai and copyright but that still leaves a massive question over the
  213. enforcement because there's a lot of work out there that has it appears to have been been taken without permission and um I know it's a huge question of
  214. you know can this ever be be compensated because at the moment we've now got publishing houses for instance that are signing deals we saw recently AFP here
  215. in Paris sign a deal with MRA giving them access to all of their journalists work uh in respon you know in return for for some some sort of gain from MRA but
  216. um the question is whether whether this is enough and and whether it'll be uh and I know there's definitely a lot of smaller artists that NE or writers that
  217. aren't necessarily represented by big houses like AFP um that that have a very different perspective yeah AFP the French news agency uh Roman Chad we saw
  218. those images of leaders beginning to arrive for that uh a dinner at the hosted at the French presidential Palace for uh those attending the AI Summit
  219. this question of who owns the internet and who owns what's in it I think the internet has undergone a frankly sad change the internet started off as a
  220. place of openness it was free a free for anybody and yes we needed to protect people on the internet but increasingly it's become a privatized and monetized
  221. space and what suffers is the quality of what's on there you had mentioned earlier about the quality of data that goes in data sets now if you think about
  222. AFP or you know the Atlantic or all these places just handing over their data sets to big companies the people who suffer the most are actually the
  223. people who have written the Articles the journalists who are not being adequately compensated because one of the questions is what does adequate compensation even
  224. look like right what do it mean to have a revenue model but but it is also self-defeating because no AI can ever write as well as a journalist like
  225. Jeremy wrote a book uh recently and like no AI could write that book and it's built on his years of EXP experience and an AI could mimic Jeremy if you wanted
  226. it to but it's not going to get that native knowledge and one of the things people talk about is something called Model collapse where models are
  227. increasingly trained on data made by Ai and and you know it's almost like an inbreeding that leads to and if you think about like you know animals uh if
  228. you inbreed too much then it has these fundamental genetic flaws and the same thing happens to AI models you cannot simply keep training a model on its own
  229. data because after a while it removes itself from reality so you'll forever need this you know genuine source of information well if we're making an
  230. exploitative system where we are simply uh you know the the the producers but not the receivers of the benefits that is you know not a system that's
  231. sustainable Jeremy are are you reassured there that no matter come hell or high water you'll always need a good copy editor yeah well that's a human being
  232. well copy editor is an interesting one CU that's where maybe the systems are good enough now to do some of that but no I'm assured that that ran thinks that
  233. AI can't write as well as I can that's also I mean people always ask me on the book did did I use AI to write it and I did not because and and they said well
  234. did you try and I did actually spend a little bit of time could I get this thing to write as well as I could I could not I could not prompt it to and I
  235. tried to give it some examples in my writing I could not get it to write as well as I could um and but I I'm not sure that's going to be true forever
  236. actually to be fair but what I do think is going to be true uh is that you're still going to need journalists and you need journalists to find facts and
  237. particularly to go out and investigate uh and and to hold public figures to account and to go out and and find new facts that are not already in the public
  238. domain that will remain and maybe even become an even more important sort of skill going forward um so whether the actual composition of the of the product
  239. uh you know is done by the the human sitting at a keyboard writing or it whether it's sort of synthesized by and and created by AI I'm not sure but the
  240. actual fact finding very much will still have to be done by journalists and then when I think it comes to sort of the writing of of works of of fiction of
  241. novels and things those are a form of Art and and I think art is ultimately about the communication from one person to another it's about actually the
  242. intention of the artist and these AI systems have no intention they're not they're not people they have no uh lived experience which they're trying to
  243. convey um so I think ultimately there's going to be something um inauthentic always about a novel written by Ai and and I hope people recognize that
  244. inauthenticity and prefer actually the auth communication you get between artist and and human audience we talk about the beauty of mathematics you're a
  245. mathematician K so but at the at the end of the day uh when it comes to things that are more of a technical ilk um the machines can take
  246. over well as far as math are concerned I don't know ter who is probably the best live being mathematician said few months ago that he used AI as U um an
  247. undergrad moderately gifted undergrad student I don't know whether it's true still true or not you remember when we used to say that about uh about Chess uh
  248. the yes the Mach that's why we should we should yeah nevertheless as Yan luk says AI does not understand the way the world the world works I mean how reality works
  249. the so uh I do think that at least for some time uh the efficient combination will be the the combination of human intelligence and artificial intelligence
  250. I mean again as far as diagnosis is concerned that's that's obvious that you you need Specialists to properly interpret the thousands of images that
  251. have been uh sorted out by artificial intelligence so well let's work with that let's try to be assessed by artificial intelligence to go to go
  252. further probably not to be replaced and what about winners and losers cuz uh the the French saying and you heard the French President saying come invest with
  253. us we got nuclear energy uh we have lots of land and we have smart people who go to PO technique but and data because we are a centralized state so right yes not
  254. all EU members are are so well endowed so are there going to be winners and losers now is it well uh I guess at least in in Europe there there I mean as
  255. far as higher education institutions are concerned the field I I know best I we we try to collaborate and I mean we are in an alliance with the tomb with techon
  256. and dfl so that's Europe in a broad sense with the the Danish and and the the Netherlands and we we tackle the same big challenges AI
  257. climate defense and security so yeah there is a European game where there should be winners uh and losers very very briefly
  258. there Peter O'Brien uh can we say this day one is the easy part I don't know if uh macron would say it's been easy because he's obviously
  259. working on a lot of these contracts for many years uh and tomorrow I mean yes I I think so because it is really the politics of it that are going
  260. to be difficult macon's clearly said he's in favor of a a Global regulation of sorts and and what I think is interesting is China we haven't talked
  261. about China but in the last six months they really have moved towards more of a a concern on AI risk and safety in in a way perhaps as the French get have got
  262. less interested in that but um China will be present the vice Premier will be there tomorrow and it seems at least on some um some topics they'll want to
  263. collaborate but I do do think even you know I've seen some people say the wording that France has gone for seems like an almost like a bit of a a bit of
  264. a middle finger to the US in in some ways because talking about things like inclusivity inclusivity and sustainability are not what the US uh
  265. wants wants to hear all right and we'll be of course covering it for you right here on France 24 Peter O'Brien many thanks for for joining us live there I
  266. want to thank TI K uh Roman a chowri Jeremy Khan thank you for being with us here in the France 24 today
  267. [Music]

Zum Nachlesen

Künstliche IntelligenzKünstliche Intelligenz (kurz KI, englisch artificial intelligence, kurz AI) ist ein Forschungs- und Anwendungsgebiet der Informatik. Anwendungen künstlicher IntelligenzDie Anwendung der künstlichen Intelligenz erstreckt sich über verschiedene Gebiete der Wissenschaft, der Wirtschaft und des Militärs. Existenzielles Risiko durch künstliche IntelligenzExistenzielles Risiko durch künstliche Intelligenz bezeichnet die Hypothese, dass wesentliche Fortschritte bei der Entwicklung von Artificial General … Verordnung über künstliche IntelligenzDie Verordnung über künstliche Intelligenz (KI) (informell KI-Verordnung, KI-VO, englisch AI Act oder EU AI Act) ist ein Rechtsakt der Europäischen Union …