Greg Irwin  0:18  

I want to welcome everybody we've been running these month way. My name is Greg Irwin, when I'm one of the partners of BWG, we're teamed up with WWT that's Worldwide Tech, who's been our co host around all of these sessions and Sustainable ETC. So if you haven't already, you know, checkout, Sustainable ETC, I'll put our link here in the chat, follow it and just, you know, keep in touch with the community. Around the way, the way we've set this up are practical steps, and education, and awareness around the practical steps towards arriving at more sustainable enterprise tech, it could be just better hardware, it could be better cooling, it could be better planning. And of course, it's all of the above. Today, as I mentioned, we're going to be speaking with David Smith, CTO at ETT World. And I'm going to take a minute here, David and give a little bit of the high level intro and then I'm gonna ask you to go a little bit further in terms of your experience here. So David is an accomplished technologist with executive roles in r&d, government, commercial and academic academia. He was named one of seven top global futurists in the millennium issue of Business Week, formerly principal designer of Intel Q at the CIA, he directed tech Futures Program at NSA for nine years, an executive at NCC and Cemetech. And he's also part of World Economic Forum. David, we really appreciate you being here with us. Do us a favor at a little bit of color in terms of your background, and really how you arrived at this topic of AI and the overlap around sustainability. 

David Smith  2:03  

Greg, thank you so much. And thank you for the introduction. My background is really diverse and have also been an assistant dean at the University of Texas at Austin business school, and many other different things as part of the MCC Consortium, which was formed to regain the US computer industry from the Japanese fifth generation computer project. We were actually very early in the artificial intelligence game. And we had multiple programs that were directed to that with folks like Doug Lynette and others who some of the publications back then said he was the advanced guy for the future of AI. And like, then he was doing a common sense artificial intelligence, which is something we still cannot do today yet. But I've been blessed to have worked with really visionaries. My whole career. From two of my background mentors were nominated for the President's Technology Award in the United States. Many other folks I was sharing with Greg, one of my favorite speeches I gave for the society design and Process Engineering Sciences, which I'm an elected fellow in. I was wanting to buy primary speakers at their 25th anniversary, and it was the worst speech I've ever given my life. Two of the other speakers were Nobel laureates, one of the other ones was an assistant to Einstein, and when the other was was assistant to Petrie, and then it was me. So it was one of those talks where you're going, Why am I on this stage? But hey, what is going to be happening with this? And we will see that as we go forward through this, I've just been blessed with doing things. I also shared with Greg, some of the places I've gotten to do some really innovative things like testifying before Congress and everyplace else to set up in Q tel, which is modeled after Q in the James Bond movie for the Central Intelligence Agency bring innovation in. I asked. So our own international accelerator, investment committees and boards and stuff and speak a lot around the world in a lot of different topics. I've done over 2000 keynote speeches during my career. So I enjoy doing these and putting them together. And I love some of the comments we're already seeing in chat because they're going to be in part of what I'm going to be sharing with you this morning already. And if it's not feel free to jump in, and let me know and we'll go back and do those. So you know, and I think everybody can see my screen now. We got it. Yes, David. And so we're really going to talk about artificial intelligence and whether it's a friend or foe for sustainability and In ESG, and other applications, this is actually a topic I am on panels very often with with both WF, and the World Talent Economic Forum, which is the one that is optimized for developing nations around the world. I know some of you are very, very advanced. And we're going to have some basics in here, but I'm gonna go pretty fast through those if someone isn't, let me know. And we'll go back and deal with them. And in one of the things when you look at artificial intelligence types, and how the different tools underneath artificial intelligence use versus sustainability, and particularly since sustainability is a global challenge, and not a company challenge, it's not a city challenge, a state challenge. It's not even a national challenge. It's a global challenge, you know, how can we begin to look at different ways to be able to go back and use sustainability. And what I'm going to do with is just take a second give you some of my type of definitions of it. And as I look at AI, you know, it really is me get some of these things out of the way here. It really is applied to event analysis and logic based techniques, including a lot of the advertising, machine learning and others, to interpret events, support and automate decisions. And very importantly, and this is what many people forget, it needs to take action. And I'm going to contrast that today to science in data. Science is actually using AI but using other things as well. It combines math statistics, specialized programs, advanced analytics, AI and machine learning, with subject matter experts to uncover actionable insights hidden in your organization data. And I hate using the term AI just to be very transparent with people. But I think you've got to really look at the both of them. But the key thing to understand with the topic of AI, is that we come back to here is that AI is neutral. It is intrinsically good and bad at the same time. Yes, AI has biases, the neutral technology, and it reflects back what you have and what can be done there. And as I look at AI, there's a lot of them, I'm just gonna quickly just so for the folks who are not as familiar, AI includes machine learning, deep learning, natural language processing. And already to one of the questions in chat. NLP is one of those that we will talk about, as we look at data centers and others there. There's advanced virtual assistants here like the chat bots and things that have been all the press computer vision generative AI. And I actually include the intelligent Internet of Things, the intelligent internet of medical things, and others, is part of AI because of how we've pushed the intelligence to the edge of the cloud, in how we've pushed intelligence into nodes that are now being able to federate back in real time, in effect change on real time, to IoT is now a big part of where we're seeing AI heading. And, you know, advantages of AI, most of you know, this better way of doing things. Looking at advanced analytics of outcomes, it can interact directly with systems that take action. And it takes in many places, the removal of human intensive calculations out. Now, there's places where human intensive calculations and particularly integration are still important. But it can begin to take care of a lot of that, in AI will ultimately reshape how work is done as the technology replaces tasks, typically performed with employees and change how day to day decisions. We're seeing that in many domains very early. I can tell you here in Austin, when they were building the Tesla factory, I was amazed with the amount of artificial intelligence that was deployed is part of that process. But you know, we mentioned machine transparency in NLP, I want to just give you an example of something that I find really interesting for it. This is a machine NLP, where we translated this sentence the spirit is willing but the freight splashes wheat, from English to Russian and back to Russian. Pretty simple, pretty straightforward. A lot of tools for that correct. But what we see is this is the translation. The spirit is willing but the flesh is weak to the vodka is strong, but the meat is rotten. And this is one of the things as we begin to use AI in Peru. particularly using it with multiple languages, we need to understand that it's not perfect, that there's places in it where we need to handle it in different ways, and particularly when I look at it in sustainability, and, and ESG type of governance, because there are still places like, for example, in the last cop, which was beginning to extend that dramatically and to develop nations that we came out of COP, that there's things that don't translate into many of the languages into me and machine languages. So where are you going to have these errors as we work through this, and the different things that can be done by AI ethics is a place where particularly again, with sustainability, we're pushing the edge of the envelope a little bit, you know, the First Age of AI ethics focused on what AI might be able to do. And super intelligence was one of those which we're still not at yet. The second way, address the ethics of the consequences of machine learning. But where we are today is we're in the third wave of AI ethics, which must play Sustainable Development at its core. And there's a growing movement all around the world towards AI for good uses, for example, AI for good. And it's going towards the AI sustainability goals from COP and AI sustainability, to movement to foster change, and in the entire lifecycle of the products, from idea generation training, retraining, information governance, and even towards its recovering, we're going to talk more about that in a minute or two. As such, is focused on more than AI application, it has to address for sustainability, the entire social technical system of AI. So again, you have to take the basic AI that we know as technologists, and begin to look at it broader when you think about it as the whole ecosystem of AI. And also, sustainable AI is a field of research that applies to the technology as well, as has been noted in chat, chat to the hardware, powering it, the methods used to train it, the actual processing of data, and then also the applications of it. And we really need to understand that sustainable ai, ai deals not exclusively with the implementation or use, but it should address the entire lifecycle of AI, inherit how it can be used, you know, if you look at ESG, the environmental, social and governance, it's your impact on the world, your contribution and how to conduct yourself. You can see here that there are places where artificial intelligence can be applied across all of these, and in how it's used. I kind of like the comments at the bottom, it's your commitment to future generation, how returns are shared and your license to operate. And it particularly with the topic today, I really think that is very important. And there and where the comment was just made that how do we begin to, to talk about the move from ethics to ethical behavior, it's a really difficult one, because, as I said earlier, the the problem of sustainability is a global problem. And just because I'm very good with using the tools and applications, and I have the companies and the others comply within my region, if the countries around me Do not do it, then I had the big challenges with I'm only a small part of the solution that requires really global action to make it work and be able to use it. And you know, it's it's easy to lie, the comments easy to talk, but not so easy to act inertia is a tough cookie to overcome. I agree. One of the things that I do is we talk about this, particularly in the world talent, Economic Forum, is there is a former UN official who lives in Jordan. And he often participates on those. And he rightfully reminds me very often, that it's really not the government officials. It's not the executives of large companies that have the greatest impact on sustainability. It actually is the common citizens, because most of them do not see the impact of climate change. Most of them do not see the impacts of sustainability. But until they begin to have a voice at the table, and can influence the companies and can influence the governments. It's going to be hard to get this into a global a course of action. And Dr. Leif is so good about reminding us as we talk about this topic. And it's one that fits directly into the question in the chat about how a nourisher is a tough cookie, right, it's going to take education, it's going to take broad based education, to be able to really have a true impact to sustainability in these areas, and to be able to make those differences. But the other thing that's does is there's the concept now being talked a lot about the circular economy. And it really is where the raw materials to design the production, the distribution, consumption, use, reuse, repair, then collecting that when it's no longer usable, and recycling it, and then even doing things with the residual waste, that cannot be directly recycled back into a product. This is a big part of the movements towards sustainability. And as we look at ESG, this is where many folks around the world are saying that to truly meet ESG targets, most companies are going to need to begin to adapt a version of the circular economy is they look at their products as they look at their services, is they look at how it comes up across the ecosystem. And you know, if I look at environmental social governance, you know, according to IDC, with IP spending on AI, hardware and software, and how fastest increasing compound annual growth rate of 24%. They're going to be a very important part of the investment technology in any company. And Blackrock CEO. No analysis firm now has a gore co Gore of investing with environmental sustainability. Goldman Sachs has now made sustainable finance core to its business. So you're seeing major companies step in line with it. At the Davis where Economic Forum, the International Business Council announced metrics and frameworks were reporting that include this, so we're getting some frameworks in place that can work with us around the world. But we're seeing some real interesting places like in agriculture, AI can transform products production by monitoring, managing environmental conditions and waste, and particularly help reduce fertilizer, pesticides, and water while improving crop years. So a direct impact on sustainability. As we go began to go and look at energy, it's really easy to see how artificial intelligence can be used as we go across the grid to meet the demand and supply particularly of renewable energy. The other thing I want to say and this is one of the big changes we're seeing in the energy area, is we're staying the movement away from large geographical grids to more and more micro grids and mini grids, where we can optimize it on the local basis how we can look at alternative sources. And where we can bring more and more intelligence and we have less waste than transmitting over those long distance transmission lines. In transportation, we know it can help enable more, particularly in supply chain logistics, and others and even more autonomous driving. AI is the key along with machine learning, Vision recognition, others were making autonomous driving happen. Water Resource, you see if they're the same in manufacturing, you can reduce waste and energy use. And you know, robotics give us better precision. Plus robots will work 24 hours a day. And you can see more uses of it. So as we talk about sustainable ai, ai, there's one where it's this sustainable AI, about the data centers and things like have been asked in that chat. But also there's AI for sustainability with its applications. So there's a whole bunch of places where the sustainable AI has both of those paths going forward. And just to give you an example of the first one. This was some data going back and looking at only a neuro linguistic Pro, I mean natural language processing and natural language processing compared to how much its co2 emission benchmarks that you see on the left one passenger going from New York City to San Francisco with that level, an average which human and American language we are larger consumption, US car manufacturing. And then over there you see how to train natural language processing and the number of co2 units it uses across the pound. And it really is very interesting to see that it is that now, that's getting better. We're seeing new types of processors coming out. We're seeing optimizations to the learning process, and things which will be reducing that considerably. One of the biggest things I think that will help us reduce that is when we see more biological computing began to hit the commercial marketplace, that will dramatically pull those figures down. But that is one of the big challenges as you look at AI on the using the technology itself side. And it goes many different places, you have all the missions or the products, the operation if

Greg Irwin  21:02  

yes, I'm sorry, kind of pausing for a moment here. Would you mind taking us back to that last slide? What I find most interesting about this slide, it is staggering in terms of the amount of co2 emissions from training an AI model,

David Smith  21:21  

or natural or natural language processing. 

Greg Irwin  21:25  

But even more interesting is I don't think this has gotten really much discussion. I know I've been involved in AI discussions I've been involved in, in energy efficiency. And we've talked about AI. I've never seen this before. And I'm wondering, is it well understood? Is there a real initiative taking place to it, because when you look at this, even if you get a 10x, improvement, and efficiency on training, the argument here is that, you know, we're overlooking the environmental costs, or the ESG impact in this race to deploy these models, because of all the outstanding things that we say they can do.

David Smith  22:06  

It's in a data center, it's not the manufacturing. So it's not truly federated. It's data center centric. It's not federated across multiple data centers. 

Unknown Speaker  22:17

What I'm wondering is, is built into this, the externalities of of building a data center? No, nor is this just the marginal cost. It's the Marty, it's the marginal model. 

David Smith  22:25

It's the model, it's the model, it's the cost of doing the training of the model, it's not part of is not the data center operations itself. But ChatGPT is not a natural language. Sensing application is a generational AI application. So this is more where they're using some of the more advanced natural language processing programs. So like in healthcare, one of the NLP programs we use a lot is called c takes that was developed by Johns Hopkins, MD Anderson and Cleveland Clinic and Mayo Clinic's, and it is a power hog as you train it. And it takes a lot of information to train it, you know, just to use that example, in healthcare. Since I think most of us on the call are in North America, the word provider in health care, you know, the word provider has over a dozen different meanings, and they're all very different. So if I'm trying to do natural language processing, what happens when I hit the word provider? How do I train the NLP to go look behind look ahead, look at context is to decide which meaning the word provider has. And that's just an easy example. There's many of those. So that's why when you're looking at things like natural language processing, why they training requirements for it is so high. It's a lot of computation, a whole lot of computation. You hit the key right there, and in all sorts of different complexity. It's like when I was doing some work in the intelligence community and in DOD, one of the agency there get it's somewhere between two to three Yoda bytes of data every day. And they try to go ahead and run machine learning NLP, across all of those Yoda bytes of data, which is well beyond what you see in a normal commercial application. And the stakes on that are very high. That's one of the things we learned during 911 was the states of not doing a full job of Looking at it all, is very high. So it actually goes back to our worst you're seeing in the commercial sector too. The other thing, though, is, as you mentioned, the edge of cloud. And I'm gonna go back to my comment about many of micro grids, what we're seeing as we look at processing at the edge of cloud, that's where many of microgrids have come in. And so things like sustainable power generation from being able to go and take, for example, the the leftover crops on a field and cow manure and everything and put those into bio plant to generate the power to feed that computation is a green process, not only a green process, it's also an economic development process, because it creates jobs that were not there, oh, and the leftover from the biomass used to generate the power becomes fertilizer to go back and enrich the land the soil from where it came. So that's part of trying to look at this whole circular economy, even in the micro sale of like a biomass plant to power data centers that can do this type of activity to and as I'm looking at ESG, those are some of the hidden nuggets of wisdom that we need to begin to look at from an ESG perspective, because I can tell you, most of those are buried way down in the detail of the corporate reporting. FPGAs are good technology, but the broader range of morphological computing, yeah, and SVGA, or GP or whatever. So there, but the real place you're gonna see fundamental level of advances is where we in true biological computing becomes commercial. Know where we've seen biological computing used. The results have been tremendous in this. But there's just not a good company commercializing it outside of government programs today that are classified. To I'm on the board of a company in Athens that's using generational AI and things then working across 92 different languages as part of theirs. And it's amazing seeing what's happening in that area. Man on the other side, I'm fascinating. I'd love to get more details about your project. If you go to our ETT website and look at our company bio data AI. It is an open innovation, intelligent platform that is tying everything to its nine applications to platform, but your type of application needs the breadth of that and the breadth of intelligence tools, and things. So that's kind of what we're doing the platform, I'd love to see if there's places where we can work together some down the road. So Greg, thank you for letting him talk. But we now have a electric power powered airplane. So it's even a lot better. But let me just say one other thing to echo your point, though, is it's not that it's the last 1% so hard. It's the new constraints we're putting into it, of echoing out all types of bias, and others like that. So those new requirements are expanding the algorithms that it has to train up on dramatically, which is also contributing to that last 1%. As we're beginning to look at what is the absolute requirement, we have a company called universal tutor, which teaches social emotional learning with 3d, conversational AI avatars. So it's not recorded speech, but it's generated in real time response. And it's read that way. We've adapted the philosophy on there. 95% 97% is adequate. As we're doing that, and I think you're going to see more of that, do you actually need 100%? Because that last bit to get to the 100% is a difficult one. Applications don't require it where others do.

Greg Irwin  29:21  

Let's let's let's go back to David here, David, you prepared a fabulous presentation for us. And I ran us off out into left field with some with some great interesting comments from from the group here. Before you know in our in our remaining time here, are there any additional comments you want to direct us to in your 

David Smith  29:42  

Let me let me jump to just a couple of slides and do that real quickly. And I'm going to jump really quickly from through some of them. Now. This is one that is the cut these couple of slides here are directly on ESG and this is a survey Out of EY. That was just taken end of last year of looking at that the EA of ESG really continues to dominate the investors allocations. And you can begin to see that in the numbers. But we also see that active funds are still the most way that most investors are trying to look at ESG. And because of that, that they're active funds. And looking at those allocations, the use of artificial intelligence is one that we can begin to train to be more deferential to these type of allocations. Most AI that I've seen used, and sustainability and ESG have been very broad and flat. And if you look at the people actively using them, they are not blocked, broad and flat. And we need to understand that. Also, if you look at the allocations, the environmental allocation is well above the social applic applications and the governance at allocations in the clients with what they want to do, despite almost 40% recognized social issues are being totally overlooked, during there, but the one on the right is the one that I think is really interesting to me, is while accessing ESG data remains a challenge, look at the different places with measurements management approach transparency, engagement, and are the ratings consistent? You know, the transparency and consistency are two are the big things that AI is struggling with when you look at ESG is that there's just so many variables on this, that it's hard to compare apples to apples, because there really is not as you looked at a lot of the application through it. So some of the big areas that need to be fixed in AI for ESG is error reduction, how do we get greater efficiency, and we need to look at some of the raw materials, because that is still very difficult to do, as you look at it. And you know, I see the comment over here and let me pull this from back up and expand it. You know, quantum computing is great, Frederick when I look at Mass computation, but when I look at artificial intelligence, it's more I see a greater application, for example, deep neural networks today is being a driver for doing this type of training in AI, then I see quantum computing, if I'm trying to compute weather models, quantum computing, it's great. If I'm trying to understand all the nuances in the multiple threads of AI, I go to something like deep neural networks, or I go to the future of biological computing and where it's being run right now is being the greater pressures there is what we're also seeing as we've gone through all these industrial AI, it's and from industry, one industry to industry industry for where we are going to industry, five, which is where cyber physical cognitive systems come on place, you're gonna see green manufacturing, or ST, cultural collaboration. And those are all the things that you is working towards that. And Thomas, your comment is still lacks programming languages. at South by Southwest last week, there's a lot of emphasis on quantum computing. And we've saw major announcements there of advances in programming languages and OS this to the list. Just remember though, the world today is totally digital, always owned, instrumented, and interconnected. You know, if you think I like to ask people to name a device that cost over $100 uses power and doesn't have a computer chip in it. So we're talking about a world that has tremendous opportunities for us to do things in it. And if you look at the data, you know, data has been doubling every year for the last two years, every year every two years for since going way back. But here's what's happening is what we see going forward is that by 2024, about 1.7 megabytes of new information will be created. Let me get this right, every second for every human being on the planet. So 1.7 megabytes, every second for every human on the planet. And think about how that's gonna go into our AI models and how it's going to be used plus we're in these very complex organization. Where is it? cross all these medias? Yes, we're getting smarter algorithms, faster computer powering things. But if we looked at COVID, when it hit traditional analytics died on COVID, because many of the models that looked heavily on historical data failed. And the pandemic changed a lot of things, rendering a lot of historical data useless as we look at going forward. And particularly as we see AI, that's going to be so much more in the future, in the Internet of Things is hitting all the verticals, as was commented earlier. And everything is smart today, and it's going, but we also see status and demands, this is out of Deloitte 78% of investors think. And I think companies should make investments that address ESG. But companies only think 55% of that, that they need to do more on the short term profitability. And when you look across ESG, there's a lot of evidence that you begin to see that between finance leaders in investors, they're close, but there's still a lot of disconnect. In looking at sustainability. And I I kind of like this analogy. The graphics is another beer, please, hell, I'm sorry, Dave, I can't do that the scales and the mayor say you're getting too fat. We can do all of this with AI today. But the question is, is do we want to use all of what we're trying to do with AIS for the type of applications we're trying to do? Or is it better to focus on and look at the ones that can help greater wisdom, sustainability, and at use a Pareto chart and prioritize those, and the other ones that take a lot of the extra energy, a lot of the extra data? Do we have to deal with all those in one time, if you look at the sustainability goals out of COP, you'll see it goes across many, many different areas. And some are real suited for AI, some are not, and what's there, and, you know, these are the goals. But each one of them, you can see here, where I just tried to put together a list of how AI could be used, and some of them and for each of those sustainability goals. And we can see that there's application there. But it's a lot of work. And one of the ways to do this, as I was telling him earlier, is through the use of public private partnerships, because I see p threes as a way to change the global playing field. Because a p3 can cross geographical boundaries really well. And we see them happening all the time. 

Greg Irwin  37:52  

we're at our hour. So I think I think we're going to wrap us up here. If you want to make a closing, a closing comment, I'm going to, you know, just close off with some logistics. And

David Smith  38:03  

let me go down to one slide down here. So what we're seeing to address the AI and ESG challenges, we need to be smart companies should craft and AI data strategy based on what's important there. It should be connected it to next generation models. And it needs to be talent led, a eyes are a tool. And it's a tool that can be used, my definition of technology is technology is not a thing. It's the application of knowledge to achieve result. And that's really what we need to do as we work across AI and its use and sustainability and ESG. So Keith, thank you for allowing me to go through this.

Greg Irwin  38:48  

David, thank you so much for taking time and speaking with us. Folks, I want to thank you all for joining, please take us up on one of our core mandates, which is to connect the community. So LinkedIn is great. I encourage you all to connect over LinkedIn. If you need some help just reply to Jake or me here at BWG. And a big thanks to our friends at WWT Worldwide Tech, Don and everybody for helping us put this together. They are the brainchild and the and the sponsors here for s
Sustainable ETC, the Sustainable Enterprise Tech Council. And we appreciate all their support. If you're doing work around sustainability specific technologies and you'd like to bounce some ideas off of off of somebody I would start with Don and his team at WWT at but of course connect with everybody across our community. With that we're going to wrap things up. David, thank you again. And thank you all. Everybody. Have a great day. Thank you