This conversation between artificial intelligence industry experts about the role of philanthropy, security and government in helping unlock the potential of AI. This conversation took place at Imagination In Action’s ‘Forging the Future of Business with AI’ Summit in April 2024.
Subscribe to FORBES: https://www.youtube.com/user/Forbes?sub_confirmation=1
Fuel your success with Forbes. Gain unlimited access to premium journalism, including breaking news, groundbreaking in-depth reported stories, daily digests and more. Plus, members get a front-row seat at members-only events with leading thinkers and doers, access to premium video that can help you get ahead, an ad-light experience, early access to select products including NFT drops and more:
https://account.forbes.com/membership/?utm_source=youtube&utm_medium=display&utm_campaign=growth_non-sub_paid_subscribe_ytdescript
Stay Connected
Forbes newsletters: https://newsletters.editorial.forbes.com
Forbes on Facebook: http://fb.com/forbes
Forbes Video on Twitter: http://www.twitter.com/forbes
Forbes Video on Instagram: http://instagram.com/forbes
More From Forbes: http://forbes.com
Forbes covers the intersection of entrepreneurship, wealth, technology, business and lifestyle with a focus on people and success.
Subscribe to FORBES: https://www.youtube.com/user/Forbes?sub_confirmation=1
Fuel your success with Forbes. Gain unlimited access to premium journalism, including breaking news, groundbreaking in-depth reported stories, daily digests and more. Plus, members get a front-row seat at members-only events with leading thinkers and doers, access to premium video that can help you get ahead, an ad-light experience, early access to select products including NFT drops and more:
https://account.forbes.com/membership/?utm_source=youtube&utm_medium=display&utm_campaign=growth_non-sub_paid_subscribe_ytdescript
Stay Connected
Forbes newsletters: https://newsletters.editorial.forbes.com
Forbes on Facebook: http://fb.com/forbes
Forbes Video on Twitter: http://www.twitter.com/forbes
Forbes Video on Instagram: http://instagram.com/forbes
More From Forbes: http://forbes.com
Forbes covers the intersection of entrepreneurship, wealth, technology, business and lifestyle with a focus on people and success.
Category
🤖
TechTranscript
00:00So, maybe what we can do, we can start with a quick round of introductions so we hear
00:09from our panelists on what they do, especially in the realm of AI.
00:14And then I have some questions for them that allows us to participate in a good discourse.
00:20So maybe we can start from that end of the panel and start with Rizwan.
00:23Sure.
00:24Thanks, Sain.
00:25Rizwan Kalfan.
00:26I'm the Chief Digital and Payments Officer at TD Bank.
00:29It's a North American bank, and I've been privileged to lead the digital transformation
00:33for the last 10 years, and looking forward to the AI-first approach at TD.
00:39All right.
00:40Mark.
00:41And I'm Mark Greaves.
00:42I currently run the AI and Advanced Computing Philanthropy for Eric Schmidt.
00:48And this is Schmidt Sciences, a major philanthropy by Eric and Wendy Schmidt.
00:55I'm Anna Kozlowskis, co-founder and CEO of VANA.
00:59We build tools for users to contribute their data to models that they jointly own and govern.
01:04So think a user-owned foundation model created by 100 million people who contribute their
01:10data in a privacy-preserving way and earn as it's used.
01:15I'm Stuart Davis.
01:16I work for the South Australian government and oversee our AI and Health Hub program.
01:20So our healthcare costs are raising at 4% a year, and our tax base isn't so.
01:25So we've got a train crash coming.
01:27So we're looking to lead a transition to preventative and a wellness-focused healthcare system using
01:33AI.
01:34Amazing.
01:35Thank you very much.
01:36So maybe, Mark, we start with you.
01:38You are working on a very interesting initiative.
01:41I have had the opportunity to collaborate with some of your team members.
01:45What's very unique about this program, it is that it's focusing on some of the key fundamental
01:51issues that is surrounding the AI ecosystem and what we need to solve for the next 20,
01:5730 years.
01:59Tell us a little bit more about that, and tell us about some of those challenges and
02:03how we are overcoming them.
02:04I will.
02:05So first of all, thank you so much for allowing me on this panel.
02:11It's from moonshot to application.
02:14You have to start with a moonshot.
02:17That is actually the role of philanthropy.
02:19One of the things we do is try and put the smartest people, the hardest problems, and
02:25the most risk together to try and make progress.
02:29And so if you think of smart people and hard problems, that's the key to this program,
02:34which we are running with Hussain, called AI 2050.
02:39And so the key question here, or the key framing question, is, okay, it's 2050, it's 26 years
02:48from now, and AI has turned out to be hugely beneficial.
02:54So what happened?
02:55Right?
02:56What did we have to start working on now to get right?
03:00And so AI 2050 is all about that question.
03:03And so one of the interesting things that we had to think about when we wanted to address
03:07that is, how do we structure a moonshot?
03:11Right?
03:12How do you actually build one?
03:13And so we came up with a couple of principles.
03:17Very briefly, one principle is we wanted to get the very smartest people on this.
03:22And so we primarily fund young faculty in universities all around the world.
03:29We have a complex nomination process to get to them.
03:34And we give them relatively unrestricted funds, because we can't really predict how this might
03:40work out.
03:41So we want the smart people to follow the rabbit, not us to prescribe what they do.
03:47But the second thing we have to do is inspire them, right?
03:50Because the smartest people can work on lots of problems.
03:52So how do you inspire them?
03:54And the way we did that was building this question about how we get to 2050, and focusing
04:00on the benefits of AI, not really the risks so much, but really how can AI help us?
04:06And that turns out to be enormously inspiring.
04:09And the second thing we do is we don't just ask them to be free around what they might
04:16get.
04:17We ask them to put their projects or their work into categories.
04:20And these are categories around what kinds of problems you might have to solve.
04:25And so one category is, I think, fairly technical.
04:29How do you get AI to be more capable, more general, more trustworthy, solve the hallucination
04:35problem, solve the robustness problem, hit some of the safety problems where the AIs
04:39just generate crap.
04:41Just try and get all this stuff, which can be fairly technical in there.
04:45And we have a group of people who are really inspired by that.
04:49But the second thing is, how do we think about AI and the great challenges of our time?
04:53The science challenges, the engineering and design challenges, the education challenges.
05:00How do we sort of apply AI in these domains, and how do we think about that?
05:07And I'll also put into there what I'll call the economic and access challenges.
05:14Right now, a lot of AI is controlled by a fairly small number of companies, primarily
05:18on the United States West Coast.
05:21That's not really sustainable, right?
05:23So how do we think about access and participation in the growth of AI?
05:27This turns out to be enormously inspiring for sets of people.
05:31And then how do we think about deploying it responsibly in society?
05:35We have a lot of societies in this world.
05:38And if we want AI to spread across the world, how do we think about all of that?
05:42How do we think about the fact that in Morocco, it is against the law to criticize the king?
05:47In the United States, it is expected to criticize the president.
05:52How do we think about that, and how do you build AIs that work in different societal
05:56contexts in the different ways that people have chosen to organize themselves?
06:01And kind of finally, I think about social systems.
06:04We're a very pluralistic world.
06:07In this country, we're very pluralistic societies in the West.
06:10How do we think about governance?
06:12How do we think about how governments can use AI and bring them to their citizens, right?
06:19So that their citizens grow up to be really first class participants in the world, aided
06:25by this technology that's as revolutionary as electricity.
06:29So that's how we think about it.
06:31Those are the problems we're trying to work on.
06:33Happy to talk to anybody afterwards.
06:35Thank you so much.
06:36I'm sure we have lots of faculty members who want to come and talk to you.
06:39Erna, maybe we come to you next.
06:43In our conversation, I was very inspired because one thing that really stood out was how you
06:49emphasize on data, data strategy, and the centrality of the data and its ownership.
06:55And you consider it even more important than AI itself.
06:58So tell us a little bit more about that, your view, and what is your vision for VANA?
07:05Yeah.
07:06So I think with all of these AI models, ultimately, they are just a product of the data that they've
07:11learned from, right?
07:12You can think of them almost as just like a being that learns from the data that it
07:15reads.
07:16And today, it's mostly the public internet, right?
07:18So it's just sort of watching all the YouTube videos, reading Reddit, et cetera.
07:23But what you really want is very high quality data to teach these models to make them much
07:27better.
07:28And so how do you create incentive structures such that everyone can help to teach the AI
07:35and contribute their data towards it?
07:37I think one really important point here is decentralization.
07:42Because if you imagine a world where just one company is trying to determine what truth
07:46is and figure out, hey, what is the right answer to this question?
07:50It's kind of an impossible question, right?
07:52Everyone has a different perspective.
07:54And how do you make sure that AI actually represents many different people?
08:00It's that you have a decentralized AI ecosystem with many different AIs that represent many
08:05different people.
08:07Decentralization, specifically, I got interested in because I was obsessed with central banks.
08:12I had a picture of Janet Yellen hanging in my high school bedroom.
08:16I was obsessed.
08:18And then came to MIT, learned about decentralized central banks, and started mining Ethereum
08:23here.
08:24And so with VANA, what we're doing is applying those same tools of decentralization that
08:29have really worked for currency, like Bitcoin and Ethereum, to AI and data, right?
08:34How do you put that power in the hands of the people and ensure that it remains decentralized
08:39and well distributed?
08:41That's amazing.
08:42I'm sure that you and Rizwan will have a lot to come to discuss.
08:46So Rizwan, as a segue, maybe we come to you.
08:49And of course, we have had an opportunity to collaborate together for years.
08:53And one thing that stands out for TD and its public knowledge, that it has been the bank
08:57that is always in the forefront of digital and data innovation.
09:01It became the number one personalized bank globally.
09:05So tell us a little bit more about that.
09:08You are working for a complex bank, which is regulated, but you continue to bring innovation
09:15to the market, both in terms of your AI and data strategy.
09:20What are the lessons learned?
09:21And perhaps what are some of the things that our attendees can have as takeaways?
09:27Yeah, thanks.
09:28I mean, you come to this place and you always learn something new.
09:31I learned a young person could have a central banker as a role model.
09:35So, you know, I've been privileged to kind of lead the digital transformation.
09:43In every 15 years, you have a disruptive technology that impacts society across industry.
09:50Each of us changes our behaviors, our preferences, our expectations.
09:55The last time that happened was the smartphone, and it changed our lives.
10:00When I reflect back on the lessons, I think a lot of the lessons there could be applied
10:06going forward as we embrace the new wave of AI.
10:10Mark talked about talent.
10:12So think about a bank trying to attract the best AI scientists.
10:17That's a difficult proposition, but we have to do that.
10:20You have three choices.
10:21You can do it through acquisition, which is what we did at TD.
10:25We acquired a leading predictive AI startup, and that startup within the bank has now grown
10:325x as far as talent is concerned.
10:37You can obviously do it organically, which for an incumbent organization, it's really
10:41tough to do.
10:43Or you can actually partner with the industry, but you've got to make choices here.
10:48So talent is going to be key.
10:50The second thing is you need an operating model.
10:52If I reflect back on the mobile transformation, we embraced Agile, Agile at scale.
10:58Created pods, cross-functional teams coming together.
11:01So what is the operating model going forward where you can infuse AI mindset in everything
11:07we do from facing our customers to the back office operations technology?
11:13The third thing I would say is the right architecture.
11:18On the mobile side, we work very closely with startups like Flybits to abstract our legacy
11:25environment and create new innovations in the front end, customer and colleague facing
11:31experiences or capabilities, when it comes to personalization, when it comes to basically
11:37serving the holistic needs of our customers.
11:42In this context, what is the right architecture for an AI-first organization?
11:47The way we think about it is a platform.
11:49I hear a lot about different components of the platform.
11:52We hear about data, which is super important.
11:54Obviously, the value of data is exponentially increased, both within the firewall and outside
11:59the firewall.
12:00But people talk about data separate from compute, separate from models and middleware and application
12:06layer.
12:07We're thinking about a platform that kind of brings all these components together that
12:10can scale as a resiliency, because a lot is going to change.
12:14We are just in the nascent phase of this transformation, and so having the right architecture
12:19is important.
12:21Delivering capability is one thing, at least for us.
12:24Our mission is to enrich the lives of our customers, our colleagues, and in our communities.
12:30It's not a capability game.
12:31It's about delivering experiences.
12:33We double down on human-centered design, and so it'll be interesting to see how human-centered
12:39design evolves in an AI world, because previously, it was a lot more about visual design, making
12:46it easy, intuitive, fast, but now it's going to be about taking personalization and engagement
12:52to a whole new level.
12:54And then finally, you need a mature innovation ecosystem, because you want to be able to
13:00kind of have your own colleagues involved, so from colleague ideation, incubation, acceleration
13:07all the way to partnerships with startups, fintechs, to having a patent portfolio.
13:13You need to have an ecosystem, innovation ecosystem, that's robust, that kind of can
13:18embrace this change in a way that the organization can truly take advantage of it.
13:25I say the things that, early learnings right now, mobile took years.
13:31For us to get the first million customers, the first 10 million customers, it took years.
13:35What we're seeing with this new wave of AI, it's moving so fast.
13:39So speed is going to be a challenge for a large organization to navigate through.
13:44I'd say a double down on the comments around data.
13:50The data ecosystems that are being formed are critical.
13:54It was a lot about harnessing the power of data within your firewall.
13:58Now it's about connecting data to different ecosystems, potentially across industry.
14:04And how do you ensure that as you're actually leveraging this data, you do it in a way
14:09where you maintain the trust with our customers.
14:12Trust is the fundamental relationship we have with our customers.
14:16And as we pursue AI first mindset, we want to not only maintain the trust, but enhance
14:22the trust.
14:23That's how we are thinking about taking the learnings and applying it going forward.
14:28Thank you.
14:29Thank you very much, Rizwan.
14:30Stuart, you're working on a global problem, which I think exists all around the world,
14:36which is what's the future of healthcare and where can digital and especially AI can come
14:43and that intersection may help us to really manage our healthcare systems better from
14:49patient experience to drug discovery, to access to healthcare, to access to specialists.
14:57Your initiative in Australia is pretty unique.
14:59Tell us a little bit about that and what can we learn from it?
15:03Yeah, absolutely.
15:04Thank you.
15:05There's some really interesting synergies with what's been discussed so far.
15:09So some really nice parallels.
15:12So our ambition is to drive a system that's focused on prevention and wellness.
15:17As a state, we've got a public healthcare system and have data sets that are quite unique.
15:23So we're the only state in Australia with a statewide electronic medical record.
15:27Other states are hospital by hospital.
15:30We also have genomic samples that are really valuable.
15:33South Australia is quite isolated, so we're quite unique in the value of those data assets.
15:38But then they're held by the state as well.
15:39So there's a responsibility to the community, the health consumer and the patient in terms
15:44of how they're realised and driven to unlock the value.
15:50That value needs to come from partnerships with the industry, research and the health
15:54organisations themselves.
15:56So how do you have an organisational structure that allows that?
16:00How do you ensure that it's patient driven?
16:01And so health consumers are in the conversation the entire way along to guide the solution.
16:09Those conversations are had with really clear language, easy to understand and are structured
16:15in a way where you're not having preconceived ideas about security or what AI is or the
16:21risks.
16:22So they're not getting in the way.
16:24But then what's the architecture that allows that?
16:27If you've got a public health care system that's allowing data to be passed to a private
16:31industry organisation to draw profits, then the community is not going to stand for it.
16:36So if you've got a structure that allows a trusted execution environment within the health
16:43care system itself, where they're controlling governance, access and ethics, and you're
16:48just passing the insight and the value that goes back to those organisations to understand
16:54disease, validate models and develop new drugs, then there's a way there where all
16:57of these pieces can fit together in a way that can transform health care.
17:02That's amazing.
17:03Yeah.
17:04So I will share my takeaways from your great comments and then I have some specific questions
17:11for you and we want to keep everyone on time for the rest of the programme.
17:16I think the key takeaways from your comments is that we need to have an ecology understanding
17:26of AI and really come up with a framework allowing us to really look at it as what really
17:32came from the defence system, system of systems.
17:35They all need to come together in a coherent ecology, whether it's governance, data sharing,
17:41ethics, transparency.
17:44I personally do not believe in explainability in AI because if you know how many of these
17:49networks work, it's very difficult to explain how they work.
17:52You can audit them, you can bring more transparency and associate risks, but it's very difficult
18:00to explain exactly how a decision is being made.
18:03So that's why I would say auditability and transparency.
18:08And I think if there are frameworks like Amazing Initiatives like you that really allows great
18:13minds in the world to collaborate and think about them, we will have a framework in which
18:18we can really build that trust medium.
18:21My view is, I remember, go back 20, 25 years ago, it was unfathomable that we could have
18:28multi-point HD video conferencing from point to point in the world.
18:32We now take it for granted.
18:35So I think we have created a very powerful communications medium around the world, but
18:40now we need to go and build a transactional and a trust medium on top of that and kind
18:46of removing the technical jargons, whether it's edge computing or decentralization or
18:53blockchain or AI.
18:55These can all come very nicely together to form that trust medium and AI will be a key
19:01component of that.
19:03I think my takeaway from Euro's one is I remember like 2008 when Steve Jobs announced the Apple
19:10Store, it really created an ecosystem.
19:14But I could be an entrepreneur with an iPhone and an SDK in my basement.
19:18I could build a game or an app and I could make millions of dollars.
19:21But if I'm an AI entrepreneur right now, I can do it.
19:25I can build an algorithm, but without data, I don't have anything.
19:28And you have no idea.
19:29I have an opportunity to co-teach a course here with Ramesh that you will hear from later
19:34today called AI Venture Studio.
19:37You have no idea how many frustrated entrepreneurs we meet that are like, I have this amazing
19:41thing, but no one gives me the data.
19:43Well, of course, they should not give you the data because you also need to understand
19:47the portability, the privacy, the contracts, similar to what you mentioned.
19:53And I think that's really key that you highlighted that decentralization doesn't mean Bitcoin.
19:59Decentralization can do so many things with the ownership of data and create confidence
20:05and comfort for people to share in exchange of value.
20:10And going back to governance, I want to have a kill switch.
20:12I want to de-link my data.
20:14I want to see how my data is training these models.
20:18And I think another thing that I think is the convergence of your talks, like you mentioned
20:21genomics, we have a program that is building an intergenerational health network using
20:28genome data to connect people in their 20s to people in their later stages of their life
20:33to learn from each other.
20:35What are the things I did?
20:36What are the nutritional things that I did?
20:38Now, imagine what data we are using, perhaps one of the most sensitive, critical data that
20:44you can ever access.
20:46But if you can use that and keep it encrypted, local, on a chipset that you can own, then
20:52you can unlock a lot of interesting things in the healthcare industry.
20:57So maybe I go back to Mark.
21:03Of all the things that you're seeing, what do you think is going to be one of the key
21:07impediments that if we solve, suddenly we can unleash the impact of AI on a societal
21:16way?
21:17I mean, you're working, the associated programs that you have are really addressing key challenges.
21:22But if you can pick one and say, if we solve this, we are going to have a big leap, what
21:26could that be?
21:27Oh, gee.
21:28You know, which one of my babies is the grittiest?
21:35Maybe I will pick two.
21:36Sure.
21:37Okay.
21:38Yeah.
21:39So one is, actually, there was a wonderful editorial in the New York Times about this
21:44just a couple of days ago.
21:46AI is currently almost unmeasurable, right?
21:50If you think of AI as a science instrument, it is uncalibrated.
21:55So how do we deal with that?
21:57This weaves into these ideas of trust, of how we think about what conclusions are that
22:03it comes to, what kinds of systems we put it into.
22:06It all rolls down to what are the right metrics for AI and how do we trust them?
22:11This is not a solved problem.
22:13There are some people who are thinking about it.
22:15There's a lot of ideas out there.
22:16We're funding a lot of great work in there.
22:18But how you actually measure AI, what are the quantities of interest, right?
22:24And how you get better and better along some kind of trajectory.
22:27That turns out to be an incredible, I think, technical issue.
22:31I think the main other impediment, I'll say, is a socio-technical one.
22:37We all know that AI is going to change employment, it's going to change relationships, it's going
22:44to change the nature of work.
22:47How do we think about that and how do we manage that transition so that at the end of the
22:54day, we have a marvelously beneficial technology in society and while we're in the transition,
23:01because AI is not going to be evenly distributed, right?
23:06And so while we're in the middle, how do we ensure that we are continuously providing
23:12the benefits of AI as broadly as possible?
23:16And so we think a lot about AI and economics.
23:19We have some wonderful work in there.
23:22It's still early days, but hopefully our grantees will be able to provide the information that
23:31everybody else here on this panel needs to wisely guide AI.
23:35That's amazing.
23:36So human AI systems, you can always use AI to build decision-making systems, but we can
23:41also use AI to build decision support systems.
23:44So we will make the final decision, but we can have the ability to consult personalized
23:50models and large models to make decisions.
23:53Very quickly, Riz, when you think about what AI is doing to change industries, like after
23:59the industrial revolution, we divided industries into sectors.
24:03You're in finance and you're in healthcare and you're in retail.
24:07What is the future of a financial institution?
24:09Is it continuing to be an organization that just manages our money or it can become a
24:14trust hub to connect us to essentially our living a better life?
24:19Look, I mean, you know, when you look at Americans, you know, the financial fluency continues
24:25to be really low.
24:28And even though we've got, you know, we put the bank in everybody's pocket through a mobile
24:32device, you know, people are not that interested in banking.
24:36They're not.
24:37Who wakes up in the morning, says, gee, I really feel like a credit card or I really
24:41feel like a mortgage.
24:42No, you say, I feel like going on a vacation or I want to, you know, renovate my family
24:48home.
24:49Right.
24:50It's about lives.
24:51And I think the opportunity here is that, you know, we're going to harness the power
24:55of data to understand people's lives, their objectives, their goals, their aspirations.
25:02The advantage that we have, at least at our bank, is we, you know, as I mentioned, our
25:08relationship with our customers is fundamentally based on trust.
25:12And so if we can harness the power of data across their lives and be able to, you know,
25:18maintain, if not enhance their trust, we'll be able to, you know, serve them more holistically.
25:24That's a huge opportunity.
25:26That's effectively, you know, how do you like, you know, more recently, I'll tell you that
25:30we're just launching, you know, as an example, serving small business customers.
25:35Today, a small business customer is running, you know, a restaurant or a cafe.
25:41They've got to deal with disjointed experiences.
25:44Banking is different from payments.
25:45It's different from invoicing.
25:46It's different from taxation.
25:48It's different from AR, PR, payroll, etc.
25:51We just, we are in the process of launching, you know, an integrated, you know, solution
25:56where all of these come together.
25:59And it makes their lives easier.
26:01Why?
26:02Because they focus on what's important to them, running their business.
26:06I see, you know, the next wave of AI giving us an opportunity, regardless of what the
26:11customer is, retail, wealth, insurance, small business, midsize, commercial, institutional,
26:17for banks to be able to collect this data, maintain this trust, and then serve customers
26:21more holistically.
26:22Hopefully you can do some of that with Flybit.
26:26So very quickly, I have been given the notice.
26:28Tell us about your vision for Banner.
26:30It's a fascinating thing.
26:32I read the website.
26:33You know, what's next?
26:34Yeah, I think in the theme of moonshots, like if you think far out and you have a perfect
26:40almost AI clone of yourself that you can deploy autonomously, right, there are 10 AI Annas,
26:45they're like out there working in the world, producing economic value, collaborating with
26:49others.
26:50How do you make sure that like, I'm the one who owns AI Anna, right, or you're the one
26:54who owns the AI version of yourself?
26:58Because as we start to see like huge economic shifts that are created by AI, right, where
27:03there's actually a lot of displacement of labor, et cetera, having you be the one who
27:08owns the AI that gets created from your data ensures that, I don't know, a big company
27:13doesn't take your data, train an AI, and kind of replace you with it.
27:17And so the big vision is really making sure that we have this healthy, decentralized ecology
27:23of AI models where people are in control and benefit from the technology rather than
27:28it being too concentrated.
27:31Thank you.
27:32Stuart, you're one of our greatest collaborators at the MIT Connection Science.
27:36We would love to have you here again next year.
27:39And if you think about next year, what should we expect?
27:42What do you want to do with this great initiative that you have in Australia?
27:47In terms of where I'd like to be next year?
27:50I think the critical piece with such a grand moonshot is where do you start?
27:55So if we could be within 12 months to have an end-to-end sliver that demonstrates benefit,
28:01that improves health outcomes within one domain, say diabetes, for example, that's driven by
28:07the consumers in that space, and it's making a change to the world, then that's one step
28:12that then you can build on.
28:14Brilliant.
28:15So thank you very much for joining our panel and visiting us at the Media Lab.
28:20You have a great day ahead of you, and we are all looking forward to stay connected.
28:24Thank you.