LIVE: Paris artificial intelligence summit

Summary

Highlights

• The summit was co-chaired by President Macron and Prime Minister Modi and saw the participation of over 800 contributors from over 100 countries 🌏
• A significant focus was on AI being developed and governed responsibly, with an emphasis on science over science fiction 💡
• The event showcased AI's potential applications in climate change, job transformation, and AI Commons in sectors like healthcare and education 🌱
• There was a call for AI to foster dignity, agency, and community, aligning with sustainable and ethical practices ✨
• Highlights included discussions on voluntary commitments for AI safety and governance 🛡️

Key Takeaways

• AI Summit in Paris emphasizes global cooperation to harness AI for good 🌍
• Countries are encouraged to develop their own AI solutions, not just be consumers 🤝
• AI has the potential to help mitigate climate change and aid in job transformation 🌱
• The development of AI should focus on enhancing human dignity, agency, and community 😊
• AI is transitioning from just observing to actively participating in real-world applications 🤖

Overview

Chapters

• 00:00 - 03:00: Welcome and Introduction This chapter seems to start with a musical introduction, followed by the speaker expressing confusion or a mistake with the term 'oops'.
• 03:00 - 08:00: Opening Remarks by Unknown Speaker In this chapter, an unknown speaker opens with a warm welcome, addressing distinguished guests, heads of state, governments, honorable ministers, and colleagues. The speaker expresses delight in welcoming the attendees.
• 08:00 - 35:00: Speech by Dr. F.F. The chapter titled 'Speech by Dr. F.F.' revolves around an address delivered at an AI Summit held in Paris. The event is notably co-chaired by President Macron and Prime Minister Modi, emphasizing its significance on a global scale. Dr. F.F. opens the speech by quoting the esteemed scientist Marie Curie, highlighting a theme of courage and the importance of understanding in the face of fear. This sets a tone for the summit, encouraging a pursuit of knowledge and collaboration in artificial intelligence.
• 35:00 - 65:00: Panel Discussions and Presentations The chapter titled 'Panel Discussions and Presentations' reflects on the importance of understanding to alleviate fear. This is highlighted by a speaker who, after months of travel and preparation, expresses delight in welcoming the audience to an iconic venue known as 'the Grand Pet,' which epitomizes innovation.
• 65:00 - 68:00: Summit Breakout Sessions Introduction This chapter introduces the summit breakout sessions, emphasizing the importance of exchanging ideas and cultures as a cornerstone of progress. It sets the tone for the proceedings by acknowledging that modernity is rooted in these exchanges. The introduction leads into a talk by Dr. F Fe, highlighting the significance of their presence at the event.
• 68:00 - 94:00: Presentations on AI in Nature Conservation The chapter discusses a summit focused on AI in nature conservation, emphasizing global openness and the fear among some countries that they may only be consumers of AI technology rather than developers.
• 94:00 - 123:00: AI in Disaster Response and Agriculture The chapter discusses the need to include contributors from diverse backgrounds in global conversations about AI advances, especially in disaster response and agriculture. It highlights a turning point with over 800 contributors from more than 100 countries and various sectors, such as governments, international organizations, academia, industry, and civil society, collaborating to prepare for such advances.
• 123:00 - 136:00: Panel on Voluntary Commitments and Governance The chapter titled 'Panel on Voluntary Commitments and Governance' begins with expressions of gratitude for those who shaped the summit. It emphasizes the importance of transitioning from viewing artificial intelligence as a science fiction concept to recognizing its real-world applications and unprecedented potential. The chapter acknowledges the immense hopes and possibilities that AI brings on a global scale.

LIVE: Paris artificial intelligence summit Transcription

• 00:00 - 00:30 what oops [Music]
• 00:30 - 01:00 ladies and gentlemen ladies and gentlemen heads of state and governments honorable ministers dear friends and colleagues I am delighted to welcome you
• 01:00 - 01:30 to this unprecedented AI Summit in Paris co-chaired by President maon and Prime Minister Modi I would like to start with of an inspiring woman and scientist Mari curri she said in life nothing is to be feared everything is to be understood
• 01:30 - 02:00 and she continued now is the time to understand more so that we may fear less I believe these words are particularly meaningful here today after months and of travel and preparation I am delighted to welcome you in this iconic venue the grand pet the grand pet is the Hallmark of innovation
• 02:00 - 02:30 and modernity it is in its very DNA and I think it reminds us today that progress stems from the exchanges of ideas and the exchanges of cultures this is the very spirit that has guided Us in the preparation of the action Summit before I give the floor to Dr F Fe whose presence today is both a
• 02:30 - 03:00 pleasure and a privilege allow me to highlight three key points first this Summit is a moment of true Global openness to the world many countries still fear that their only choice is to be customers in the development of AI that they're not able to develop their own Solutions researchers and innovators on
• 03:00 - 03:30 the other hand who are the ones who are driving these advances are too often sidelined in global conversations this must change and today marks a turning point over 800 contributors from more than 100 countries from governments from International organizations from Academia from industry from Civil Society have worked with us to prepare
• 03:30 - 04:00 and shaped this Summit I want to thank them all I want to thank you all second it is high time that we move from science fiction to the real world of applications of AI with its unprecedented poent potential artificial intelligence fuels both immense hopes and
• 04:00 - 04:30 times exaggerated fear today you will be able here at the Grand P to see concrete applications and use cases of AI they have been selected from more than 100 for more than 700 candidates worldwide and third now is the time to act together for shared progress I know we share the same determination that we want to shape our
• 04:30 - 05:00 future shared progress means on the one hand enabling a sustainable development of AI given the urgency of the environmental crisis we know that AI can help mitigate climate change but we also know that its current trajectory is unsustainable we must reconcile the digital transformation and the ecological transition we also need in terms of the transform of jobs to help ensure that IT
• 05:00 - 05:30 augments jobs and doesn't replace them on the other hand shared progress means developing AI for the public interest together we must build AI Commons we must enable the development of decentralized hubs worldwide for applications in health care in education in cultural access this is our Collective responsibility and must do it
• 05:30 - 06:00 together the gr today is welcoming you all tomorrow heads of state again in the morning and the whole ecosystem will be meeting at station F the largest incubator on behalf of the AI action Summit team welcome to Paris welcome to the Grand palet welcome
• 06:00 - 06:30 to the AI action [Applause] Summit thank you all and now f f the floor is yours thank you welcome good morning
• 06:30 - 07:00 I'm waiting for my slides to load I hope uh you're having a great time in this beautiful uh morning in Paris uh could I please have my slides okay thank you so much uh good morning everyone I'm truly honored and excited to be here with all of you and this week and to be delivering this opening speech and I really look forward to some of the most Act packed discussions and possibly
• 07:00 - 07:30 even decisions so for many of us AI is often framed as a story of Technology at the modern world and certainly a story of the future that's why all of us are here but for me it's also a story that reaches so far into the very beginning of all of our existence that is started half a billion years
• 07:30 - 08:00 ago it was such a it was such a distant time that the very notion of sight didn't yet exist in fact the eyes themselves had yet to evolve and not a single living thing laid eyes on this world leing all life in the darkness you see on the screen now of course half a billion years years of history is a little long to unpack in
• 08:00 - 08:30 about 10 minutes so here is the tldr the moment Evolution endowed the simple creatures you see here on the screen the ability to sense the world around them and to react to these Sensations however dimly an arms race in evolution began what started as a passive experience the simple Act of
• 08:30 - 09:00 letting in light soon become much richer and far more active the nervous system began to evolve sight turned into Insight seeing become understanding understanding led to action and all these give rise to intelligence and this would reshape the very nature of life on Earth
• 09:00 - 09:30 forever today fast forward half a billion years later human intelligence has led our species to imagine and shape the way we work and live in every possible ways and we're no longer satisfied with just having Nature's gift of intelligence our curiosity now urges us to create machines that are just as intelligently as all of us if not
• 09:30 - 10:00 better hence the quest of the seminal technology that we are talking about this week began in the mid 20th century Alan Turing a great mathematician in UK was so pre that even before computers existed he was already grappling with the philosophy of endowing them with the cognitive ability rivaling our own to me his writing has
• 10:00 - 10:30 always been a dare daring Humanity to be bold enough to imagine thinking machines as he had imagined the very same Spirit of curiosity and ambition also gripped early computer scientists in America who not only convened the first research project to explore explor possibilities
• 10:30 - 11:00 of thinking machines they also coin the term artificial intelligence itself in the steamy summer of 1956 decades before it reached public Consciousness and this is the screenshot of the very research paper they wrote for this Workshop admittedly what was really amusing is they think that in two months they would solve a huge part of this
• 11:00 - 11:30 problem and uh unravel the mystery of intelligence maybe they were a little optimistic but you have to admire their audacity today we're 820 months into that two-month project but we made some solid progress another misconception of AI is that it's simply a matter of computers and engineering when in fact it has
• 11:30 - 12:00 always been a multidisiplinary pursuit of incredible Vibrance the modern AI era that we are living in now is a result of the convergence of three profoundly important yet distinctive technological and scientific advances first is the study of perceptual algorithms how living creatures like animals or even ourselves
• 12:00 - 12:30 make sense of the world around us and eventually the creation of mathematical models that allow MERS after Alan turing's staring conjecture at the Dartmouth AI summer workshop neurophysiologists hubo and viso have shed first first light on the hierarchical architecture of neuronal processing of the molan visual cortex
• 12:30 - 13:00 earning them a Nobel pricee and Transforming Our understanding of the visual processing around then psychologist Frank Rosen blot built one of the first prototypes of new network called perception this work would go on to inspire Decades of computer scientists especially early Pioneers like kiiko Fukushima Jeff Hinton y Lun Joshua Benjo
• 13:00 - 13:30 among others to device increasingly sophisticated models culminating in the Deep learning neuron Network algorithms we know today giving our machines uncanny ability uh of their own at the same time a second thread start to emerge cognitive scientists looking and into our own minds shedding light on
• 13:30 - 14:00 astonishing depth and sophistication of our own ability to perceive our surroundings a clear conclusion of their work has been was that there is no separating our brains from the environments in which they evolved from brains are simply machines locked away in the darkness of our skull
• 14:00 - 14:30 but eager students learning from every each glimpse of light Ting of touch Whisper of sound from the earliest moment of life for me in particular as a scientist coming of age at the dawn of the 21st century the when perceptual algorithms were almost the sole focus of my field the lesson of the cognitive science was
• 14:30 - 15:00 one of scale The evolutionary and developmental process have benefited from an enormous amount of data to drive our own learning my collaborators and students and I conjectured that the same would be true for machines but this time instead of collecting data using biological sensors they will come from Modern digital devic devices and the
• 15:00 - 15:30 internet this inspired my own Labs project imet the first internet scale training and evaluation data set for AI our hypothesis of data as the key to high capacity algorithms like new network brought them back to life in ways that previously unreachable and kicked off a global Trend in using big
• 15:30 - 16:00 data for AI now know as part of the scaling La of AI finally however none of these achievement would have been possible or would have made it to the first uh passing the starting block without the extraordin extraordinarily powerful to run them it began with the Milestones like development of the van architecture a paradigm of computer
• 16:00 - 16:30 architecture first proposed in the 1940s and is still followed today leading to the first microprocessors in the early 1970s but amusingly it was video games that helped push the raw speed of silicon into overdrive a cottage industry beginning in the early 1990s to make game Graphics more detailed
• 16:30 - 17:00 exploded into a global Powerhouse within just a couple of decades as giants like Nvidia built increasingly powerful Graphics processing units or gpus this turned out to be the last missing puzzle to en Network algorithms to learn from internet scale big data so if there are any gamers out there we
• 17:00 - 17:30 have to thank you for this the rest of course isn't just history it's the recipe for building the future in 2012 at the image net challenge my lab has pioneered each of these three ingredients algorithms data and compute came together for the first time at the scale big enough to transform my field almost
• 17:30 - 18:00 overnight it was the first time machines could understand and reliably describe of images millions and millions of them it's something we now take for granted today but it was an unsolved problem at the time and ability that felt like science fiction and it was the first Domino piece to fall in in a progression of Milestones that only seemed to come fast faster and faster as
• 18:00 - 18:30 the years pass by now over a decade later we're still exploring the implication of what this all means what began as a in academic curiosity is now enjoying year after year of breathless attention from Business Leaders entrepreneurs industry analysts and even politicians with no end in sight were now a decade or so into what historians
• 18:30 - 19:00 would surely refer as the first true era of AI by almost any metric you can name whether it's the Cs degrees AI programs uh investment funding new startups and many uh much else AI is a revolution that's only growing in scope and scale none of you need me to explain that over the last couple of years AI took another
• 19:00 - 19:30 even more staggering turn with the Advent of large language models which uh which scale each of these three ingredients of modern AI to even bigger extent a new architecture called Transformer trained on effectively the entire internet of data powered by a staggering number of the latest and best chips the result as we have all seen was an
• 19:30 - 20:00 even bigger explosion in machine ability than those of the last de decade we now take it for granted that AI can talk to us fluently in natural language answering questions on almost any topic we ask and even generate complex forms of almost uh any imagery voice music and videos the astonishing success
• 20:00 - 20:30 of chbt which set record for user adoption says it all about the impact of this Innovation that's has made on everyday lives and the capabilities aren't just superficial if you look at this chart the performance of AI models across a whole range of Benchmark from handwriting recognition to PhD level
• 20:30 - 21:00 scientific questions has skyrocketed in recent years with some of the performance trajectory on some of the most challenging tasks to humans are nearly vertical and starting from the fluency of languages large language models are now taking on a more active role in problem solving breaking down tasks and planning steps towards achieving goals
• 21:00 - 21:30 in real world a gentic AI as meing a calling now in 2025 is the latest chapter of this consequential technology for many users and businesses but there is still more yet to come there's more than language intelligence think about humans and our holistic being as an intelligent being a new chapter is Dawning one in which the
• 21:30 - 22:00 capacity for Idol perception is expanding into something far more active in my own world of cameras and robots AI is starting to create understand reason and interact in int intangible 3D spaces whether digital or physical it can be used to interact with people and
• 22:00 - 22:30 things and we call this spatial and embodi intelligence take a image like this one as an example our visual intelligence make it easy for us to recognize everything that's in it the cat the pot of plant the table and of course that glass of milk but is that really all perception our perception is giving us I'm willing to bet if some of you look
• 22:30 - 23:00 at this picture there's more than a few of you um are just understanding this image but also you're dying to reach out and grab the glass of milk before it shatters over the floor this is a trivial example but it highlights how profound the difference is from merely observing to acting it illustrates the inflection point point I believe were
• 23:00 - 23:30 about to be reaching at which AI goes from an observer to an actor alongside Us in fact with respect to my hero Alan Turing I now believe his vision of the future was too narrow and too inward facing Evolution itself has taught us unequivocally that the real power of intelligence isn't merely to think but
• 23:30 - 24:00 to drive action with those thoughts just think about how much humans uh spatial intelligence has empowered us to build our civilization from ancient pyramids to Industrial Revolution from scientific discovery to artistic expression how will our relationship with the world around us change even further when AI expands our spatial and
• 24:00 - 24:30 embodied intelligence What will what will it help and help us to create and discover what kind of future are we already in the process of building well here's a glimpse of the rapidly evolving technology of spatial intelligence I'm showing you here four examples by my students at Stanford and
• 24:30 - 25:00 colleagues at World Labs ranging from the top uh the upper left uh panel is the semantic labeling of complex everyday videos to the upper right panel artistic style transfer to the lower left panel generative AI algorithms creating videos from text prompts to the lower right panel uh making 3D worlds from any image and that's right you're
• 25:00 - 25:30 indeed seeing the Beloved painting by vano of a French cafe brought into an imaginary 3D World by our generative AI models and equally exciting robotics a form of embod AI is also rapid advancing these two work from my la are the latest example of combining robotic learning
• 25:30 - 26:00 with large language and vision models allowing robots to perform everyday human tasks in much more open realistic settings compared to the previous generation of robots that were highly programmed and choreographed these are exciting possibilities but if AI were truly become not only thinking machines but
• 26:00 - 26:30 also doing machines our Collective responsibility in shepherding this technology becomes even more urgent and important I don't think it's an exaggeration to say all this has brought us to a moment of civilizational consequences so how are we going to navigate this I think this is also the purpose of our gathering this week this question has guided my work for years
• 26:30 - 27:00 while I'm certain there are no easy answer to be found one theme continuously to run across all my work this is what I call human Center AI which consists of three simple values dignity agency and Community First is dignity in the face of more and more power ful Technologies we as humans are often confronting the
• 27:00 - 27:30 question of what defines us stripping away from all the tasks we're able to do the fundamental pride of being who we are making decisions on our own and conducting actions by ourselves is still Central to our very being nothing would excit me more if this technology can help protect and even return that sense
• 27:30 - 28:00 of dignity to all of us especially to the most vulnerable and this is an example how Robotics and AI technology can one day help indow the sense of autonomy to even the most paralyzed patients in this Stanford project by my students a robotic arm is controlled purely by the thoughts of the person
• 28:00 - 28:30 collected non invasively through EEG and the recording of brain waves our AI algorithms are able to decode that person's thoughts and instruction and guide a robotic arm to make a full Japanese Sukiyaki meal second is agency a core tenant of my my lab's work has been exploring
• 28:30 - 29:00 applications of AI that augment what humans are capable of instead of replacing them jobs will inevitably be impacted by the progress of AI just as every major technological shifts in history has reshaped the labor market ining AI help to augment and Empower us from
• 29:00 - 29:30 creativity to Health Care from scientific discovery to manufacturing much of AI skill set are complementary to that of humans there's so much opportunity for us to take advantage of this digital or physical collaborator to superpower ourselves my lab's work in the past decade in AI healthc care delivery has
• 29:30 - 30:00 shown me many opportunities for AI to help to improve the quality of care and to alleviate the burdens of our health care workers here are three examples of using smart cameras powered by AI algorithm to help hospitals to improve hand hygiene practice by clinicians to help document patient Mobility exercises
• 30:00 - 30:30 to and to assist surgical instrument tracking last but not the least is community so much of the last decade has been a story of Technology tearing us apart filter bubbles rage bait and all the likes AI is at another fork in this road with one leading to AI that over Shadow real social experiences and
• 30:30 - 31:00 personally tailor content that confirms our biases but on the other path leads towards a world in which AI help us to build better bigger and more robust communities such as educational assistance that bring the ability to learn to more and more people kids and grown-ups alike here's just two quick examples on the left us Ai and and VR to support
• 31:00 - 31:30 dyslexic college students as personalized learning tools on the right is an AI expert created by University of Buffalo in the US focusing on augmenting the shortage of speech language pathologist to to come to early intervention for children aged three 3 to 10 with speech and language
• 31:30 - 32:00 processing challenges all these bring me to the last idea I would like to share with you today at this civilizational moment how can we work together to govern AI in a way that preserves the astonishing potential in a rational practical and responsible way for starters it's essential that we govern on the basis of
• 32:00 - 32:30 science not science fiction from the main street to the Wall Street so much of today's AI conversations are colored by sensationalism and hybo resulting in misleading policies of AI governance instead we need to apply a much more scientific method in assessing and measuring AI capability and limitations
• 32:30 - 33:00 which can consequentially led to more precise actionable policies grounded in reality in turn this brings me to the second point of embracing a pragmatic attitude instead of a ideological one when it comes to AI governance AI is poised to be a powerful technology that that can help us to live and work better
• 33:00 - 33:30 if used well so instead of hindering the Upstream exploration and research of this still very young technology we should put more focus on its practical applications so that we can ensure its benevolent usage and to guard against harmful outcome finally we need to invest in far more healthier a vibrant AI ecosystem
• 33:30 - 34:00 where Academia entrepreneurs um open-source communities and public sector can all participate and play their critical role in alongside big companies in driving this technology forward if AI is going to change the world we need everyone from all walks of life to have a role in shaping this change earlier in this talk
• 34:00 - 34:30 I spoke about three critical ingredients that brought uh brought us modern data and compute if the bulk of these resources are concentrated in only a handful of companies the AI ecosystem will suffer from the lack of curiosity driven research top educational talents opsource efforts and multidisiplinary
• 34:30 - 35:00 Explorations so 75 years ago Alan Turing saw a glimpse of the future and was inspired enough to DARE Humanity to build a thinking machine today we have taken tan stair to a level that he probably could not have imagined the technological advances of the age of
• 35:00 - 35:30 AI is breathtaking I believe time has come for a new kind of dare rather than simply asking if we can create AI let's ask if we can create AI as a force of good in short today I would like to DARE all of us to build a human centered AI thank
• 35:30 - 36:00 you thank you so much dear f f thank you let them take picture of that us yeah I'll be so proud thank you thank you so much for doing this thank you I much appreciate thank you so much de f for addressing us I really liked the fact that you highlighted science not science
• 36:00 - 36:30 fiction and I was very impressed by the embodied AI capabilities that you showed us now let me start the continuation of this program I would like to bring to the stage or to call to the stage both Ena Fred and Nick Thompson who will be hosting this whole event today please Nick Ena please join me welcome wonderful welcome thank
• 36:30 - 37:00 you Nicholas Thompson American I'll be one of the moderators on stage today we're going to have incredible series of pitches they're awesome we're going to see red teamers we're going to see scientific innovation we're going to see AI Ed for good lots of great stuff coming Ena justu Ena no I don't speak yes um but yeah I mean it's such an important time you know in 25 years of covering Tech and silicon Valley I've never seen a technology move as fast there's so much
• 37:00 - 37:30 potential and there's a lot still to be worked out and not a lot of time to do it so look forward to hearing some of the Innovations all right wonderful and before we start we will uh see the art of sopia Crespo stay tuned and um see you soon all right we'll be back shortly
• 37:30 - 38:00 spe
• 38:00 - 38:30 and I just also would like to um clarify
• 38:30 - 39:00 for everyone that the breakout sessions the main stage the auditorium they're all starting now so you can either stay here and look at uh the art and then the use cases but you can also make your way
• 39:00 - 39:30 to the main stage the auditorium the breakout sessions thank you it's in the
• 39:30 - 40:00 e
• 40:00 - 40:30 e
• 40:30 - 41:00 e
• 41:00 - 41:30 e
• 41:30 - 42:00 e
• 42:00 - 42:30 e
• 42:30 - 43:00 e
• 43:00 - 43:30 e
• 43:30 - 44:00 e
• 44:00 - 44:30 e
• 44:30 - 45:00 e
• 45:00 - 45:30 e
• 45:30 - 46:00 e e
• 46:00 - 46:30 [Music]
• 46:30 - 47:00 welcome back everyone I'm Ena freed Chief technology correspondent at axios I'm very excited to share with you some of the presentations that we have on how AI is being used today um and I would be remiss we talked about the fact that this is a field that has been going on
• 47:00 - 47:30 for a while and we talked about Alan Turing earlier in faf's talk uh my good friend the former Chief technology officer of the United States Megan Smith Reminds me that aah love lace also played a role she was in the 1800s suggesting that we need a calculus of the brain if you will uh so with that I'm honored that one of the first speakers is one of the women helping lead this um and that is Claudia Gomez who leads a project called tech for nature from Mexico
• 47:30 - 48:00 [Applause] Claudia Conan over here and first what a beautiful stage to talk about such important topics and to be honest AI should be a tool that helps us all not only drive mere efficiency but Collective purpose we know that AI is transforming every industry but there's particularly one critical area that needs our full attention as an
• 48:00 - 48:30 AI ecosystem which is nature conservation and regeneration and this becomes not only an environmental issue it becomes an issue of technological access it becomes an issue of economy it becomes an issue of fairness it becomes an issue of economic stability and let me tell you a bit more about the region I'm from Latin America and the Caribbean this region is home to more than 40% of the world's
• 48:30 - 49:00 biodiversity and this is the same biodiversity that is one that is regulating the climate around the world one that is being the pillars of innovation and the pillars of all Industries however biodiversity is being lost in alarming rates at se mines the organization I founded and I lead we believe this is not a problem of solution it is a problem of imagination it is a
• 49:00 - 49:30 problem of interoperability of joining the finance the conservation and the technological worlds to bring purpose in this field and to drive it both ethically and context wise for us it has meant to explore the frontiers of innovation and how to use these tools to redefine paradigms paradigms that are centered in nature paradigm times that are centered in an economy that has a new frontier
• 49:30 - 50:00 based on balance on equity on Justice specifically we're working with AI systems to understand how to protect better the Jaguars in southern Mexico the peninsula of Yucatan a Mayan indigenous people Rich culture area and what we're doing is we're developing together with a local community which is mayam based and the local student students from local universities two deep learning models that are redefining
• 50:00 - 50:30 what is possible for conservation the models do not only allow us to recognize patterns through camera traps with thousands of images coming daily through these iot systems but also it helps us to understand the presence of jaguars and the individual species based on the patterns of The Furs or resets as we know and this this is not only a mere topic of understanding the presence of a Jaguar
• 50:30 - 51:00 but I don't know if you know but the Jaguars are considered an umbrella species which means that if the Jaguar is healthy the whole ecosystem around the Jaguar where is present is healthy it becomes a proxy for stability and for ecological wellness and also the Jaguar has a rich and profound cultural importance it is a symbol for not only the Mayan indigenous peoples but through a lot of indigenous peoples around Latin America and the Caribbean it is a symbol
• 51:00 - 51:30 of power it is a symbol of balance it is a symbol of just Transitions and thus you can understand why we're focusing not only in the biological importance of the Jaguar but in the cultural one as well having Ai and multi- sector collaborations and the Partnerships with local community as an interface to create a reality that is more nature-based that relies on
• 51:30 - 52:00 regeneration and redefines the possibilities of justness and equity and fairness and this project that we're helping Drive in the south of Mexico is having a lot of specific Corey salts that are multi-level one we have developed a tool and a platform that is open to anyone that is interested in Jaguar conservation throughout any region to be able to use our models and use our tools openly without cost to be able to advance their costs on this conservation
• 52:00 - 52:30 second we're joining these efforts with another interoperable system where we're using AI sound pattern recognitions to identify the presence of other species for us it has meant to identify more than 117 species in that area which are endemic and most of them in endangered species actually some scientists didn't even know that some specific species were present in that area and the data that we're having through these AI
• 52:30 - 53:00 driven systems are helping integrate a platform where we can integrally understand the health of the ecosystem third we're driving policy what are we doing exactly this project was a base to create the Statewide policy of the 30x30 conservation Statewide policy on protecting and conserving 30% of land and sea by 2030 and this is not only important for the policy level but it also serves as a
• 53:00 - 53:30 metaphor for Global change through this project and through the data decisionmaking this region and this natural reserve was also registered in the global natural reserves of iucn and lastly it became the footprint the metaphor and the inspiration for the Jaguar Regional biocord Alliance which we have presented at Cup last year together with the interamerican velopment bank and partners to be able to scale a philosophy of
• 53:30 - 54:00 interoperability and when I see interoperability we're going Beyond data Beyond technology but interoperability of knowledge systems of possibilities and what we're seeing here are two main things the first is that we cannot separate technology and AI driven systems from the social the cultural the economic and the systems that embed
• 54:00 - 54:30 Injustice they have to go hand in hand together with ethical considerations and context wise philosophies and second it has shown and has become a metaphor the Jaguar of the type of economies that we should create economists that yes are led by AI but at its core have a principle of regenerative Justice like the Jaguar of
• 54:30 - 55:00 balance and a transformation based in equity thank you very much so you know one of the things about nature and AI is the intense amount of energy use that's being used for these AI systems how do you reconcile the work that you're doing with AI with the fact that the industry you're a part of uses so much energy it is a key question to this work and any type of work that uses
• 55:00 - 55:30 AI so we believe in green AI what does that mean we have to recognize that we have to walk towards an AI that we mitigate and we reduce enery consumption and at the same time we have to mitigate the ecological footprint for for us we're working with tech providers that are conscious of the energy use and while we transition to that energy efficient system we're mitigating our footprint with regenerative ecosystem
• 55:30 - 56:00 approaches awesome well I I had the opportunity uh your Project's incredible I also had the opportunity at our AI Summit that axio did in San Francisco I don't know if you're aware the Earth species project they're helping us talk to animals I don't know if Jaguars will be a part of it yes indeed and we're also working with whales so that will we will present that coming up thank you so much castan gzm so our next speaker uh comes to us from Nepal is working on a project
• 56:00 - 56:30 called satellites for Humanity please welcome nian badio hello hello check you're on thank you uh bonjo mid Mish my name is n Bak uh today I'm representing Anar praan Nepal and it's umbrella project satellites for Humanity now first going back to the project let's give some
• 56:30 - 57:00 brief about the organization itself Anar is a nonprofit uh Aerospace company specializing in space Tech especially Cubs ground stations ground station terminals open source and proprietary software systems with AI capabilities and we are also doing uh different research and development projects on uh AI models for onboard system systems and ground stations as well so now let's uh imagine that you're
• 57:00 - 57:30 sipping a cup of tea in a very beautiful mountain range and all of a sudden a wall of water comes down storming sweeping everything around you that's uh Glacier Lake outward flood and because of climate change a lot of those incidents are being frequent in a country like Nepal and Himalayan countries uh incidents like flash floods uh landslides forest fire drought and so on so satellite for humanities aimed to
• 57:30 - 58:00 fill the gap between the incident and the response now that response could be post disaster or pre- disaster using some early warning uh systems in place so in 2021 uh in mumi River there was a very big flood so we customed our uh ground station terminals uh with a communication to a cubat in space uh that would provide us data of weather uh
• 58:00 - 58:30 from cubat to ground station of Nas Nepal Academy of Science and Technology we were able to use those data to model some uh early warning system predictive models unfortunately the uh ground station terminal was also Swept Away uh but uh we got the data and we got the learning so that's fine uh after that we we went to uh collaboration with KATU Metropolitan
• 58:30 - 59:00 city with a objective of utilizing all the open source satellite datas uh and uh commercial datas and Commercial uh AI models as well to enhance our systems and provide uh satellite data analytics so under the project satellite data analytics project we delivered them with some flood hazards uh risk hazards and and all kind of satellite uh data analytics with AI insights uh from which
• 59:00 - 59:30 we were able to derive the problems pinpoint the problems and provide alternative viable Solutions uh so with that solution in place we were able to create some positive impact among uh 2 million residents of that metro metro City and around 6 million commuters of that place uh to make more our system more more robust we are more inclined to AI models like age AI image classification
• 59:30 - 60:00 models uh image segmentation models and for ground stations for making our early warning system more robust uh we are using predictive models from the data that we are getting from uh different glacial lakes and rivers also the population estimation model uh that we are working on we have been working on for few years now now uh actually uses uh satellite data end to end to provide the estimation of
• 60:00 - 60:30 population of a building so the state of the art model can provide a GD level population estimation uh while we challenged ourself and decreased the special Dimension uh and uh also increase the accuracy by some point we are able to predict the population of a building of katman Metropol City with accuracy of around 87% till now and we are working on that to make it more general for similar Urban uh
• 60:30 - 61:00 geography uh for the model to be open sourced and used on different places as well along with that we have uh three different satellite projects uh we completed two and uh we are on the verse of uh completion of the third the slippers to sat project now with each iteration of satellite project we have uh enhanced the systems systems Hardware uh communication and specifically the AI
• 61:00 - 61:30 models that we have been using so we started with AI model for image AI model for image classification and uh we opt for image classification as well as segmentation and uh this third one is is more better model than that and we are also testing the feasibility of smaller language model on a on Satellite computer itself it's like imagine having a uh capacity of analysis and a rational
• 61:30 - 62:00 thinking in a mini computer mini satellite it's like having a mini astronaut is exciting and what's more exciting is uh collaboration with like-minded people and organization like you uh on projects like these and I'm very sure that we will be able to achieve something significant by this year so thank you for the opportunity and uh thank you for uh giving your attention Mery have a good
• 62:00 - 62:30 day thank you good job um so I mean talk about sort of how how did you end up in this work what brought you to the project personally yeah so in Nepal space technology was a new thing a big thing and a new thing uh so I started uh I fortunately I got a very good Mentor on first place and I was also very excited about space and Technologies from Charwood so it
• 62:30 - 63:00 clicked and also the narrative that that the organization draws uh about not only being fancy about this space Tech but also actually solving some real world problems step by step uh initially it looks small but it has a huge impact and uh Believing on those objectives it just motivates to do more to be involved in more projects to comp complete more of these uh so yeah it's it's like it's exciting the coming projects are
• 63:00 - 63:30 exciting the projects that we have done are exciting so yeah and how can people check out your work uh we have uh websites uh we are in booth number one so you can come and have a chat with us uh and we also have a LinkedIn pre QR somewhere over there all right so if you want to hear more about Nan project in in Booth one thank you so much thank you so much thank you so we're going to shift a little bit from uh from what we were were talking
• 63:30 - 64:00 about the first couple projects looking at uh things to we're going to also switch languages so uh next up please help me welcome Raman faru from GA and this presentation will be in French thank you very much mer
• 64:00 - 64:30 start man
• 64:30 - 65:00 comp
• 65:00 - 65:30 open
• 65:30 - 66:00 Maxim
• 66:00 - 66:30 Serv
• 66:30 - 67:00 far
• 67:00 - 67:30 open
• 67:30 - 68:00 fa
• 68:00 - 68:30 [Music] [Music] uh
• 68:30 - 69:00 [Music]
• 69:00 - 69:30 open source
• 69:30 - 70:00 comp
• 70:00 - 70:30 Dev
• 70:30 - 71:00 Val Val
• 71:00 - 71:30 [Music]
• 71:30 - 72:00 Roman Meri um so I was trying to use Google Translate real quick but I it wasn't keeping Pace what's one thing you hope that the English speakers will take away the fact that in France we have a u a kind of lab an experience and a feedback
• 72:00 - 72:30 from uh 15 years of act Tech and the maturity to see how uh valorize aggregate all the uh environmental data that we have around in the farming uh sector and around the farming sector to create more knowledge uh more uh new strategies uh more to be more helpful with the data for farmers and all over the world we have uh in heart to share that experience and to launch it uh
• 72:30 - 73:00 anywhere uh we can find the same parameters and problematics and eating better being more safe with what we produce is a I think a challenge that we share uh all over the world and there's so much we need to improve our food uh technology especially when you think the the amount of aable land the amount of land we can grow food on is shrinking our population is growing climate's
• 73:00 - 73:30 getting trickier but if we can use technology to spread the knowledge we'll be better off exactly what we think and what we believe in excellent thank you so much R thanks both thanks to all of you uh so thank you so much this is our first session we'll be back around 11: uh so 15 minutes come back and we'll have more presentations and uh I'll be joined on stage my colleague Nick Thompson from the Atlantic will be uh taking over
• 73:30 - 74:00 [Music] thanks
• 74:00 - 74:30 e
• 74:30 - 75:00 e
• 75:00 - 75:30 e
• 75:30 - 76:00 e
• 76:00 - 76:30 e
• 76:30 - 77:00 e
• 77:00 - 77:30 e
• 77:30 - 78:00 e
• 78:00 - 78:30 e
• 78:30 - 79:00 e
• 79:00 - 79:30 e
• 79:30 - 80:00 e
• 80:00 - 80:30 e
• 80:30 - 81:00 e
• 81:00 - 81:30 e
• 81:30 - 82:00 e
• 82:00 - 82:30 e
• 82:30 - 83:00 e
• 83:00 - 83:30 e
• 83:30 - 84:00 e
• 84:00 - 84:30 e
• 84:30 - 85:00 e
• 85:00 - 85:30 e
• 85:30 - 86:00 e
• 86:00 - 86:30 e
• 86:30 - 87:00 e
• 87:00 - 87:30 e
• 87:30 - 88:00 e
• 88:00 - 88:30 e
• 88:30 - 89:00 e
• 89:00 - 89:30 e
• 89:30 - 90:00 e
• 90:00 - 90:30 e
• 90:30 - 91:00 e
• 91:00 - 91:30 e
• 91:30 - 92:00 e
• 92:00 - 92:30 e
• 92:30 - 93:00 e
• 93:00 - 93:30 e
• 93:30 - 94:00 e
• 94:00 - 94:30 e
• 94:30 - 95:00 e
• 95:00 - 95:30 e
• 95:30 - 96:00 e
• 96:00 - 96:30 e
• 96:30 - 97:00 e
• 97:00 - 97:30 e
• 97:30 - 98:00 e
• 98:00 - 98:30 e
• 98:30 - 99:00 e
• 99:00 - 99:30 e
• 99:30 - 100:00 e
• 100:00 - 100:30 e
• 100:30 - 101:00 e
• 101:00 - 101:30 e
• 101:30 - 102:00 e
• 102:00 - 102:30 e
• 102:30 - 103:00 e
• 103:00 - 103:30 e
• 103:30 - 104:00 e
• 104:00 - 104:30 e
• 104:30 - 105:00 e
• 105:00 - 105:30 e
• 105:30 - 106:00 e
• 106:00 - 106:30 e
• 106:30 - 107:00 e
• 107:00 - 107:30 e
• 107:30 - 108:00 e
• 108:00 - 108:30 e
• 108:30 - 109:00 e
• 109:00 - 109:30 e
• 109:30 - 110:00 e
• 110:00 - 110:30 e e
• 110:30 - 111:00 science technology and Innovation at the
• 111:00 - 111:30 oecd the organization for economic cooperation and development it's my great pleasure to be here today to moderate this panel where we're going to discuss voluntary commitments and in
• 111:30 - 112:00 particular the Hiroshima AI code of conduct uh that I had the pleasure of launching on Friday last week at the oecd as the pace of innovation increases in AI the uh Pro proliferation of voluntary uh Frameworks as well as regulatory Frameworks is on the top of everybody's mind including ours at the oecd our mission is to bring country together in dialogue but also to harmonize approaches to gather data and to advise policy makers on how to create
• 112:00 - 112:30 safe secure and trustworthy AI last week was a significant milestone for us in the launch of the Hiroshima AI reporting framework for the code of conduct it's the first uh truly Global approach to implementation of a voluntary code of conduct uh for advanced AI systems uh there are others of course and I just want to mention a couple in May 2024 the international AI Community took a significant step in the sole Frontier AI safety commitments these
• 112:30 - 113:00 commitments at the time were signed by 16 leading AI developers and they focused specifically on Frontier AI systems and established concrete pledges for testing information sharing and responsible development practices the hirosima AI code of conduct is safety focused but it's also broader than that it's more comprehensive and takes on on issues like intellectual property privacy and data issues and so uh it's it's my great
• 113:00 - 113:30 pleasure to have these five distinguished panelists with me today we're going to discuss the role of voluntary commitments there are lots of them out there what role do they play how can they work together with regulation it's our deep belief at the oecd that it's takes an ecosystem to manage to govern and to nurture this technology toward uh diffusion across our economies and Societies in a way that will uh Advance economic growth and human uh uh Prosperity so with that I'm
• 113:30 - 114:00 going to turn to my first panelist uh Vice Minister from uh for international affairs from the ministry of internal affairs and Communications of Japan Mr Tako imagawa please good morning thank you a it's great honor I'm sorry it's great H it's a great honor to be here and thank you for having me here in the in this uh important session first of all let me talk about recent progresses by the G7 and OCD on the international governance
• 114:00 - 114:30 in last two years Japan contribut contributed to the launch of the hosim May process under the Japan's G7 presidency in 2023 and developed the international guiding principles and international code of conduct for advanc developers with the G7 member states and the oecd in order to ensure the effect effess of implementing code of conduct by AI developers we have developed the reporting framework as just explained
• 114:30 - 115:00 under the Italian G7 presidency and launched this this mechanism last Friday before the uh just before the AI action Summit because the focus of this Summit is action it is very important for the hirosima process to Foster AI developers to move forward and to put this code of conduct into action we hope that this reporting framework will promote Innovation while improving transparency and accountability in the development
• 115:00 - 115:30 development of generative Ai and we also hope that efforts based on the code of conduct will spread globally we also hope to encourage many AI developers to participate in the reporting framework in addition as the code of conduct is a living document we would like we' like to appreciate receiving feedbacks from the participating organizations to promote safe secure and trusty AI not only companies but also governments should play should play an
• 115:30 - 116:00 important role Japan is in the process of establishing a new leg legislative framework for AI we are trying to strike the right balance between Innovation and risk management and also uh to ensure consistency with International Frameworks including the hirosima process our goal one of the most Innovation friendly countries in the world and our new legislation will serve as one of the successful models for the
• 116:00 - 116:30 global Society finally uh Japan will contribute to uh Japan will continue to contribute to International efforts and cooperate with cooperate with the international Society to further Advance the hosay process and we also would like to invite more uh companies to participate in this in this reporting framework thank you thank you Vice Minister it's been a great pleasure to cooperate with Japan under your leadership in the G7 followed by Italy as you say to bring this
• 116:30 - 117:00 publicly Let's uh turn now to um some some people closer to the action uh and I'm going to ask uh uh Miss Lisa Soder who is a senior policy re researcher at interface to talk a little bit about your experience um both on the technical side and on the research side and and what you think the role of voluntary commitments can play in the policy ecosystem wonderful thank you so much and a pleasure to be here today um so
• 117:00 - 117:30 yeah I think it's really great to see so much consensus both in industry and government um that we need to take action on AI governance um and we with these voluntary commitments uh we really do have some wonderful tools at our hand um to get some first feedback loops um into to governance so to get some first experience what kind of tools work what um what doesn't um and so this is really great to see um also that the um
• 117:30 - 118:00 reporting framework has been launched um last week but I think there's still one area uh where we do need to move faster and take a stronger stance and this is really centering all around this notion of accountability we're talking here about an industry where there are a lot of companies that have this Mantra of let's move fast and break things um at the very core of their company DNA often um and so I don't think this puts in a really good um starting position to
• 118:00 - 118:30 also trust them um to uh keep their promises and not break them um so what we really need for these uh voluntary commitments to be a real success um and to actually use them as this beautiful tool to get some first feedback loops and move very fast in in our governance efforts um is to establish accountability mechanisms to give you like a very quick example maybe from um a lot of the reports that have been updated um this past week um from the
• 118:30 - 119:00 companies in terms of how they're complying um with with the frontier safety commitment um that was launched in um there's a lot of this language in there around um we may decide or we aim to decide to share information with governments where we think it might be appropriate if there's risk to Public Safety and um and this is obviously not enough um from a citizen perspective from a government perspective to trust these companies um to do this um and and
• 119:00 - 119:30 take action on this alone so going forward I really hope to see uh more work on transparency framework is a really important First Step but we do have to go deeper and not just ask some high level questions for create accountability mechanism compliance checks we need to build our governance capacity and ultimately I do think it's also an important step to think about how can we put these um commitments also on a legislative footing so I'm seeing
• 119:30 - 120:00 here someone from the AI office thank you so much for your work on the code of practice please keep up the ambition levels uh also people from the UK AI Bill uh really please use this tool um as as a great starting point where you can build up your governance effort from there um and um keep going forward thank you Lis so uh you've highlighted well that uh we're sort of at the beginning of what is going to be a very long journey and
• 120:00 - 120:30 uh having spent time many years in the private sector myself I can understand that it might be daunting even for large organizations to look across the landscape and decide what to volunteer for how to report what to tell who and why and so I'm curious since we have um two companies here uh on the pan if you might talk a little bit Peter about um your experience there or your thoughts around the role of these voluntary Frameworks and how harmonization may or
• 120:30 - 121:00 may not help in the context of uh of uh a of accountability to Lisa's point and so I with that I'll introduce Peter sarin who's a co-founder in CVP at um AMD Silo AI thanks it's it's an honor to be to be here um um I mean I I I obviously represent a very practical perspective from from many different
• 121:00 - 121:30 dimensions uh and and feel sort of less accomplished than many on the panel related to the process through uh say uh Soul Bletchley Park and and and Hiroshima and and so forth um but maybe I can add sort of you know uh some to to what we are actually seeing um and have seen for the past many years so um yeah I mean I Silo um Europe's largest private AI lab that was eventually acquired by AMD in the summer
• 121:30 - 122:00 um but we've been in the space for for uh quite a few years we were founded end of 2017 um and been been in various different context contributing to models and and and building models uh both open source large language models but also a wide range of other models um I mean I I think uh I I think eventually um you know we obviously believe in are are committed to
• 122:00 - 122:30 legislation to uh the reg regulatory Frameworks and um and also sort of voluntary commitments um um we do we I mean we do at the same time see that there's um quite significant complexity uh by being a practitioner in in in the space um and I think that complexity comes from the fact that there's no one such thing as as AI um I mean there's just so
• 122:30 - 123:00 many different perspectives to what AI is in the real world be it sort of general purpose uh AI models Frontier AI or however you define it or then sort of AI models that are embedded into everything from I don't know say vacuum cleaners to uh lawn mowers to toothbrushes and you know that is the reality we are seeing in in in the real world um and and eventually that that sort of practitioners are are seeing in
• 123:00 - 123:30 the in the real world and and I think a lot of the um say large language models in the real world they aren't necessarily like these GP aai type of models or or uh general purpose AI assistants but they're actually relying on open source and they're hundreds and thousands of them out there for various purposes optimized for that specific purpose um and and open source obviously add sort of quite quite a quite a
• 123:30 - 124:00 different dimension to the space as well which I think is sort of already in itself on its own tackling some of the sort of transparency and accountability Dimensions uh that that we are looking at with these sort of voluntary uh commitments um as well so I I do see sort of voluntary commitments as a way to maybe bridge the gap between sort of um practical implementation and and legislation um and and help us move
• 124:00 - 124:30 forward but at the same time I think it is it is adding uh sort of yet another complexity that practitioners eventually sort of need to spend time with yeah thank you for that um I'm going to turn now to Lisa Soder who's sorry sorry Sarah it's the Sarah hooker for the who's the VP of research at cohere and curious on the research side do these voluntary commitments come across your radar what do you think of
• 124:30 - 125:00 them um coher has been committing to to to at least the so um the so safety commitments and was involved a bit in the hoshima AI process as well so please share with with us your perspective yeah well it's lovely to be here I mean I think that so coh is a frontier AI company and so it works on a global first perspective so actually a lot of our Focus it's very refreshing to be at this Summit and to see the focus on global Ai and um multilingual AI that
• 125:00 - 125:30 serves a broader range of interests um I lead our research team so my focus is on the Next Generation models and I think on a personal level being a researcher in AI um frankly uh seeing something which has traditionally been reserved for conferences immediately translate to impact around the world is um at least personally makes safety and thinking about these topics very pressing um I'll share candidly a few observations so we
• 125:30 - 126:00 submitted a safety framework this was decided at so um many other Frontier AI companies submitted uh there is a wide variance so it sounds like uh Lisa you mentioned you've read a few of them uh I would say ours is very centered on like um privacy and security and how do we think about data and multi lingual impact there's others that are much more focused on maybe more sci-fi long-term risks everything has to have a portfolio
• 126:00 - 126:30 but the first thing I'll say is I wish uh luck for the you know policy makers who have to reconcile this so there's an interesting aspect to this which is part of having voluntary commitments and I think what is interesting about the oecd hirosima commitments is standardizing some of the questions that so this is very helpful I think this is a first step but the longevity of this will rest on whether it is possible to reconcile
• 126:30 - 127:00 and that it's actionable at the end of the day and so if this is a nice moment we're all here there's very few rooms in the world that have all of us at the same time but part of that discussion has to be an acknowledgement that there's a lot of work ahead and whether these are you know whether this is a you know three voluntary commitments a year type of thing or whether has longevity depends on firstly how do you reconcile these very different perspectives of risk but secondly how is it actionable in the long run so this is a starting
• 127:00 - 127:30 point for this conversation thank you Sarah I I think we we fly agree and it's why um I think the Japan and then Italy had the vision of taking uh principles on a paper and trying to standardize the format because we have experience in other domains of Technology where we have transparency reports on many things out of companies that are not comparable across uh across organizations and certainly not at an international level so agree it's definitely a first step and the code I
• 127:30 - 128:00 think itself will have to evolve over time as technology advances um but the the objective behind it is also for it to be public transparent and so that it can inform our research at the oecd but certainly everybody out there so with that I'm going to turn to Professor to talk to us maybe a little bit more um from an from an academic perspective Ive uh and his thoughts on um voluntary commitments uh but also more broadly um The Experience uh in Academia looking at
• 128:00 - 128:30 uh AI safety and and AI uh development so professor John mcdermid he was a professor of software engineering at the University of York you have the floor yeah thank you very much and again I'm very um pleased to be here and I'd actually like to pick up on some things Sarah was saying about making things actionable and my colleague and I and nework have worked on safety of software intensive systems for nearly 40 years the last seven on autonomy and systems using AI um supported by the L's
• 128:30 - 129:00 register foundation so I think there are cases where autonomy is valuable but actually with Frontier AI I think they are best used based on our experience to actually augment human skills ra than to try to supplant them um governments have a role to play in managing risks and the hiroshim um process I think is very welcome there I think the critical steps actually lie with the the companies they're the ones who can actually shape the models and
• 129:00 - 129:30 and make them safe so from that point of view the frontier ey safety equipments I think are very welcome but actually implementing them making them manageable I think is a real challenge I'd like to talk about some things from the traditional safety engering world but has adapt to to software intensive systems they're going to help doing that and the first thing is we we have to understand the context of use um and to us given how broad these Foundation models are that means we need to link
• 129:30 - 130:00 what we call Upstream safety which is around the frontier models themselves with Downstream safety what happens for an application in particular context of use um to me one of the really critical needs is better measures of risk assessment these ideas of criticality levels I think are a starting point we need to understand the scale of deployment things that can be deployed on millions or even billions of um uh devices in several hours or days is is a
• 130:00 - 130:30 is a huge risk factor that we need to take into account um we need to talk about societal impact we also need to understand the ability to remediate harm if something occurs and we can't remediate it once it's happened we must avoid it if it can be rectified we can take a different attitude to that risk we also need to worry about about single points of failure and what happened recently with crowd streak I think is a good example of of that um we need to be able to design for safety and for
• 130:30 - 131:00 Assurance and that's something where I think the AI safety world can very much learn from the safety critical systems industry another really critical aspect is being able to diagnose problems if problems arise and we can't work out why they occurred and prevent them occurring again then actually will lose public trust and on that matter of trust I think the idea of safety cases which we've worked
• 131:00 - 131:30 on
• 131:30 - 132:00 e
• 132:00 - 132:30 e
• 132:30 - 133:00 e
• 133:00 - 133:30 e
• 133:30 - 134:00 e
• 134:00 - 134:30 e
• 134:30 - 135:00 e
• 135:00 - 135:30 e
• 135:30 - 136:00 e for