Setting International Rules on the Use of AI

good morning thank you very much for having me it's a great pleasure to be here in Abu Dhabi Finance week especially in this track about sustainability and talking about AI regulation and governance and it is important because we are all aware that AI is a very disruptive technology and what does disruption mean for me disruption can be described in many any ways but I like to describe disruption as the end of the status quo something that we thought it was stable is not stable anymore and we need to think about new instruments to achieve the stability and to encourage Innovation while keeping I mean fundamental values uh in place so the question about artificial intelligence regulation has been around since the early days of AI emergency and there are many voices that say we should not regulate AI why because this is going to hamper Innovation and this is not true and in fact we have regulated Technologies since the very early days of Industrial Revolution sometimes we have regulated working conditions sometimes we have regulated environmental emissions we have regulated safety why these time should be different so the important thing is that we have had already a lot of initiatives around AI governance and regulation in the last years in fact you see here that there were some in initiatives starting in 20199 but it's clear before and after Chad GPT moment where all of a sudden a lot of conferences a lot of discussion about ethics ethic guidelines governance and regulation took place in the world specifically after CH GPT moment we just had six in one year but in the first half of this year last year there were already seven the most important Landmark regulations have been the executive order by Biden harrris Administration already one year ago the will not be follow up by uh Trump Administration the G7 Hiroshima Summit that came to International guidelines and ethical guidance or R gener AI this is the charter of digital rights that I led when I was the former state Secretary in Spain I was the state uh Secretary of uh digitalization artificial intelligence and this one is me leading with my team the European I as was the leading negotiator of the European a act that it is in place and it is one regulation 27 countries in the world so there are a lot of initiatives at the national level on ethical guidelines on regulation and sometimes people try to misc confuse these three words which are interrelated but are not the same ethics governance and regulation ethics and I would say does everybody agree that we need an ethical development of AI I found never somebody saying the opposite ethic it means what are the principles and values I need to guide companies Developers implementers and also users including governance in this use of AI without knowing that is bringing Innovation but it's not for example affecting human rights fundamental rights safety or health governance means what are the instruments to ensure that companies and governance are behaving ethically because ethic ethics is subjective ethics can change from country to Country from culture to culture from a CO to another Co so while ethics is nice to have it doesn't imply any accountability it doesn't imply any rights we do need instruments governance instruments to set up the right accountability mechanisms to ensure that both companies and users and governance use AI in an ethical way and regulation is one of these instruments but it's not the only one we can use regulation we can use Market incentives we can use transparency and monitoring mechanisms and that was the question that we had at the global level in the United Nations so Secretary General Antonio Gutierrez posed his question do we need a global governance of AI on top of the existing efforts at corporate level uh good practices or even a regulation do we need for AI the same thing that we had to set up with atomic energy do we need an international agency of AI in the same way we developed an international agency of atomic energy and to answer that question he gathered 39 experts am among uh a call for more than 2,000 experts I had the privilege to be the co-chair of this uh important lineup of experts as you see in the first rle second place is Minister of AI of United Arab Emirates Sultan Omar Alama who is member of my board and has been a great contributor not only in smart regulation but also in the application of in healthare so this group of people which is very diverse we are representing 33 uh countries in the world diverse in terms of gender and Academia background in uh public uh leaders also private sector have been working during one year from October 22 October 23 and we just issue an inter report in December but the most important part is the last report which is a result of our work that was presented the United Nation general assembly last September and that report was called governing AI for Humanity so in the report what we first already detected in in the report is that there is a need of global governance why because there is a global governance deficit on AI This Global governance deficit is in three areas one lack of inclusiveness you've seen all those initiatives that I showed you in the beginning there are only seven countries in the world that that participate in all these conferences only seven there are 118 countries in the world that participate in none oh surprisingly most of these countries come from the global South why because everything we know about AI everything that he's been developed by AI has only been developed from data from the global North from computing power from the global North and from Talent from the global North so there's a lack of inclusiveness a lack of representation of data data that allows us not to be biased by gender by Faith by cultural heritage even a country like UAE which is a very powerful uh capable country according to the tortoise Index this is the tortoise index of AI which is the top 50 countries in the world capable of developing AI UAE stands for position number 25 and it only participates in two or five sorry in two of these initiatives that I mentioned before only in two so there's clear a lack of inclusiveness and representation of diversity in the current models of AI therefore we need an inclusive approach inclusive approach that is good for equity for Effectiveness and efficiency and we need to network the existing organizations in the same direction and enable local systems and local environments for new entrepreneurship that have curated data representative data and also Talent from local countries that find the solutions to the problems they have because we have the tendency in the global North to use technology as a new way of techn colonialism I develop the solution I devel the technology and I think that the rest of the world are going just to be the consumers of those Solutions no we need to provide the tools that every country in their own needs find the solutions to their own problems and for that we need data talent and compute the second important governance Gap at the global level is lack of coordination look at how many different just in the UN system how many initiatives within the UN system exist around AI of course there is a subject matter expertise I need to keep who better than uh wh to assess the impact of AI in healthcare who better than wio to assess the impact of AI inal property but all these efforts cannot be in silos we need a global coordination because AI is a general purpose technology that requires a holistic approach and because this is not just a technological problem this is a technosocial problem because it's going to affect society it's going to impact the future of work it's going to impact the environment it's going to impact special collectivities like women like children and also other uh minorities and there's a third important governance Gap lack of transparency and when there is no transparency we cannot ask for accountability why we have no transparency in AI models because they have been mostly developed by the private sector first of all you see that the highest number of granted patents is from China 80% 20% by the US and the rest of the world accounts for less than 2% but if we see the number of foundation models you see that 100 more than 100 are just developed by the industry but only 28 are from the Academia why is it so why has been AI for the first time ever the most disruptive technology we have ever known has not been developed by Academia because Academia didn't have the talent no because the Academia didn't have the funding no Stanford Max plan institutes CERN they do have the talent they do have the money what they don't have is data and the reason why this technology has been developed in the same data monopolies is because they had access to data which is not transparent and they have become black boxes because they are black boxes we cannot ask for transparency and without transparency how are we going to up that mainstream because the only return of investment we can make from the huge amount of money that we are pouring in dat infrastructure and AI development can only be paid back if we do adopt AI massively along all the industries and for that we need transparency on the models and if you see here also the high level of concentration is United States but Arab Emirates is appearing here as one of the top three among the top models so you're doing a really really very good from this ecosystem but we need transparency to ask for accountability bear in mind that I know you see it's not a little bit blur but the four five the the top five are Google open AI meta Microsoft and de mine Google is De mine openi is Microsoft so that reduce the number of 5 to three so this is really concentrated in very very few hands and very very few powers and because this is so opaque there are a lot of of risks that are preventing us to adopt AI with trust we performed during the UN advisory body a risk po analysis all over the world and these are the top concerns of people and experts on AI concerns about information Integrity The Fakes and also uh fake news also in abuse of power in bias also in safety and you know there are a lot of safety institutes that are going to prevent the misuse of Technology but this is not enough there are long-term issues around safety and health even mental health but there are really present issues about fundamental and human rights but important thing of this is that not all the countries in the world have the same perception of those risks and if you look at this just highlighting the most relevant differences because we have here Africa and Latin America while proba would say the Global Northern more concerned on fundamental rights and safety the global South is concerned especially on the use of AI for armed conflicts and the impact in the labor force look at that that is Latin America the top green one the impact of AI to harms to labor from adoption of AI so when we talk about sustainability of technology in this case on AI we should not only think about environmental sustain ability we need to think about economic sustainability social sustainability and environmental sustainability and these things these three things are interwined this is an important uh report that we do also together with ILO the international labor organization assessing the impact of AI in labor the future of work and this impact is not even it depends on the type of work especially the number of of wars that be mostly affected will be clerical jobs by the way those who are mostly taken care by women so the impact of hyper Automation and this intermediation of AI will not affect evenly men and women or certain positions than others but there are three situations here and when we're going to evaluate the impact of v in labor we need to understand these three different scenarios we need to understand in our company in our process what which are the tasks that can be automated fully automated hyper automated and therefore there will be a people replacement so those tasks that will replace and automate people what we need to do rescaling because people do not have the skills but in those areas when we can use technology to augment our human brain or find alternative strategies for the same problems like is happening today in Pharma Discovery and research where uh things like Alpha fault are bring absolutely news perspective of pharmaceutical development which is really dropping than the cost there we don't need to replace people we need to augment them then we need to do upskilling and cross Skilling so it's not so simple but it's possible because we are not going this is not going to happen overnight we're not going to go to bed one day and wake up in the morning if this thing have changed there is a transition period that we can monitor and we should monitor last but not least let's talk of the elephant in the room and the elephant in the room is the lack of sustainability of thei large language models which are not only consuming already a projected figure 3% of the total energy in the world but they are consuming Huges amounts of water and doing emissions for example CH gp3 chpt3 J 4 is already improved that but J gpt3 is is emitting the same number of CO2 that five cars along all their lifetime five cars along their lifetime is that sustainable it is not just make a comparison chp3 processes one trillion of tokens exactly the same amount of neuron connections that our brain consumes and you know what do we need to make our brain work every day one sandwich and for glasses of water so there's a lot of room for improvement for green algorithms algorithms that green and sustainable by Design not developing the technology and then as the public sector to deploy the infrastructure we need to have energy for free of course the impact of this the impact of data centers which are not sustainable are changing energy policies as well because the battle for AI today is the Battle of the control of energy and we're going to see a nuclear Renaissance that is we're going to see of course as small nuclear reactors close to Data Centers but that will not be enough because this consumptions only from the training phase what happens with the inference phase so just to finish these are the clear recommendations and the instruments we propose for a global governance of AI mostly three important ones that are already in place an international scientific panel for AI in the simp pl we have buil the ipcc panel for climate change to have evidence scientific evidence and transparency on those models on the risk and also on the opportunities this is already going the only difference is that we don't expect to report every six years but to report every six months the second big instrument is an a policy dialogue that stands for Global cooperation and multi-stakeholder platforms to discuss of international standards and risk taxonomy and the three important instrument is going to be a global fund for AI a Global Financial effort with public and private sector not only to invest in this local assistance we need in facilitating computing power to the global South but also investing in AI for sustainable development NOS only these we're going to be sure to make AI work for the good of humanity thank you very much