Brainstorm AI London 2024: From Sports To The Workplace
Dr. Olivier Oullier, Co-founder, Inclusive Brains; Chairman, Institute for Artificial Intelligence by Biotech Dental Interviewer: Ellie Austin, FORTUNE
Category
🤖
TechTranscript
00:00Hello again, everyone. I hope you are enjoying your lunch. So next up, we have our keynote
00:06conversation. Throughout history, sport has served as a catalyst for both technological
00:13innovation and societal change. We touched on this in our sessions yesterday. And the
00:18advent of AI is no different. Today's lunchtime speaker has leveraged robotics and brain-computer
00:25interfaces to enable a quadriplegic individual to drive a Formula 1 car with his mind. It's
00:31incredible. You can look it up on YouTube. And this summer, our speaker will use similar
00:36technology to enable severely disabled individuals to carry the Olympic and Paralympic torches.
00:44We'll discuss the potential of AI to enhance the lives of those with diverse
00:49physical and cognitive abilities and what this might mean for inclusion in the workplace
00:54and beyond. So please join me in welcoming Olivier Oulier, chairman of the Institute for
01:00Artificial Intelligence by Biotech Dental and CEO and co-founder of Inclusive Brains.
01:10Hi, Olivier. Thank you very much. Hi, Olivier. Take a seat. Thank you. Thank you so much for
01:16joining us. My pleasure. You have many titles. You are a neuroscientist. You are an entrepreneur.
01:21You are an investor. And as I mentioned, you run your own company, Inclusive Brains,
01:24which you founded in 2022, I believe. With a fantastic co-founder, Paul Barbast. Genius.
01:29With your co-founder. Now, the goal of Inclusive Brains is to build AI's first
01:36neural foundation model. Can you tell me what that means?
01:40We want to create algorithms and models that enable machines to adapt to who we are
01:47and how we feel, regardless of our needs, abilities and physicality. Machines should serve
01:55us. AI is a fantastic assistive tool, and we want to bring it to the masses. Paul and I,
02:02when we founded Inclusive Brains, had one benchmark in mind. It's the remote control.
02:08The remote control was invented for people with disabilities, for them to interact with technology.
02:14And then it conquered the world. Billions of people have changed the way they interact with
02:19technology, thanks to the remote control. We want to be inspired by this story to bring AI that we
02:27first developed to help people with special abilities and needs, and then leverage it for
02:33inclusion. Because true inclusion means serving and assisting everybody with no discrimination
02:40whatsoever. Most AI systems are trained on language, but I know that you train your systems
02:46on a range of biological functions, whether it's the clench of a jaw, the clench of a hand,
02:52or a heartbeat, a brainwave. Can you talk us through how you do that?
02:55Yes. It's incredible, because look at us. We're interacting, and words matter. But there is so
03:02much more than words in the way we interact and the way we interact with technology.
03:07And there are people who can't use words. So we're training our models with brainwaves,
03:14facial expressions, voice stress, what we call prosody, the intonation,
03:20heartbeats, and lots of biological factors, so that we can leverage generative AI to better
03:26understand and feel how we are, and the technology to adapt to who we are, which I think this is the
03:32key. We have to create interfaces that adapt, that can predict down the road, but first to
03:40adapt to who we are and what makes each of us unique. So to adapt to an individual's needs,
03:46cognitive or physical. Absolutely.
03:49I mentioned in the introduction the project that you're doing this summer with the Olympics and
03:53the Paralympics. It's called Project Prometheus, I believe. Why is the Olympic parade, the torch
03:59parade, an important platform for you to showcase this work?
04:04The entire world is going to watch the Olympics. I'm old enough to have witnessed someone with a
04:10jetpack going out of a coliseum in Los Angeles in 1984. And as a kid, that really sparkled, you know,
04:17my interest in technology and amazing things. And I think we can inspire a lot of kids
04:24to see that the combination of AI with exoskeleton and neuroscience can help people
04:33have concrete, really concrete applications. I think this is why we want sport and its values
04:39to be associated with top technology that we want to bring to the masses.
04:43We're going to demo in a second the exoskeleton. But first of all, I want you to talk us through
04:49how this technology originated, because I believe it came from a chance encounter you
04:53had when you were studying at Harvard. And we've got a photo to illustrate this, I believe.
04:57I'm going to try not to be too emotional because the person you see is my friend Mark Pollock.
05:03He's an explorer, a technologist. He has a superpower. He brings technology and inventors
05:10that don't talk to each other in order to turn their inventions into actionable solutions to
05:16help the world. What you see is Mark, his team, and I almost 10 years ago to this day
05:24in a basement in UCLA as Mark was using an exoskeleton in order to walk. Mark is paralyzed
05:30from the waist down, and he cannot see. And I was recording his brainwaves in order to understand
05:37how his brain is processing movement following the accident that happened in his life.
05:44And this is really the very beginning 10 years ago of Project Prometheus, the combination of
05:49exoskeleton and neuroscience. Let's try it out. I have to ask you a question and for your
05:56permission first. Would you mind if I use neurotechnology to record your brain activity?
06:01Please go ahead. And do you mind if we show your brainwaves to everyone in the room?
06:07Please go ahead. It has to be done with informed consent of people.
06:12That's very respectful. So this is a brain sensor, wireless brain sensor, that is monitoring
06:21Ellie's electric activity of her brain. Okay. Just sorry to mess with your hair. No, go ahead. I'm
06:29not precious about it. You're not? No. Perfect. This is more important. Exactly. So, and let's
06:38have a look at Ellie's brainwaves. So the one thing that we need to show first is that we're
06:44doing this live. So would you mind moving your head because there are movement sensors. So this
06:50is in real time. And this is how we train our algorithms. The technical term is multi-modal,
06:56meaning that we're using different kinds of information to train generative AI. Brainwaves,
07:02we're also using facial expression, very important. The interface is asking you whether
07:08you allow for the camera to process your facial expressions in real time.
07:16Your heartbeat, lots of things. I could go on for hours as you can imagine, but this is just
07:23to show you that in real time, we can not only monitor what is happening, but leverage artificial
07:29intelligence to measure in real time the level of stress, attention, cognitive load.
07:37Hi, right now. So how are these markers, stress, cognitive load, attention, and arousal,
07:45how do they sit against the baseline of what you would expect?
07:48Well, it all depends on the environment in which you are and how you were before. Well,
07:54we're human beings. We're a bit stressed, even if you and I are quite used to be on stage,
07:59and it changes a lot. Yeah, it's jumping.
08:02The fact that it's jumping, if you have to remember one thing, is the fact that how we stress,
08:07how we are, our personality cannot be captured by a bunch of questions once in a lifetime that
08:12determines who you are in your organization. It changes from a minute to another. It changes from
08:18a day to another, from a person to another. And this is how we can create interfaces that are
08:23tailored to our uniqueness. So when you have this information,
08:27what do you do with it? We can use it in order to prevent accidents,
08:33in order to improve how people learn, in order to inform what the outcome of a decision-making
08:41process can be. But it has to be done respecting all the regulations when it comes to data,
08:47to privacy, and now regulation of artificial intelligence. And we can do things these days
08:54in terms of monitoring in real time that we couldn't do in the past. Generative AI is amazing
08:59for that. But once again, we have to go beyond large language models because there is so much
09:05that language can capture. We're a lot more than words, fortunately. So would you like to try the
09:12exoskeleton? I'd love to. Now, you made it very clear to me beforehand that this is not a
09:17robotic arm. It's an exoskeleton. Absolutely. Why is that distinction so important?
09:21So first, it's designed to help people. So there are two modes we can use it. One is passive help.
09:29There are people who can still move but are losing, step by step, the ability to use their
09:35muscle. They're losing power. So the device is going to assist you when you move. Can you share
09:42with our audience what you're experiencing? Yeah, so I'm only very slightly moving my arm,
09:46but it feels like there's a superpower behind my joint that is doing all the hard work.
09:51So this is also a brain scan that we launched at a fortune event back in 2019 in Guangzhou, China.
10:02It's brain sensing earbuds. And we can leverage them
10:06to do what people would call mind control. In that case, it's control that is done with clenching
10:14because the technology can adapt to what you can do. Some people, like Ellie, can focus and can do
10:20real mind control. One of our participants in the Olympic torch relay, she has a cognitive
10:27disability and she cannot focus enough. So we had to tailor the interface to measure how much
10:32So we had to tailor the interface to measure her facial expressions and also the fact that
10:38she's clenching in order to be doing what we're going to do now, which is me by clenching,
10:45providing you more assistance. Oh, wow. Yeah. That's way stronger.
10:52Wow. And so this could be installed at someone's workstation, for example,
10:56if they needed help at work. It could be in their home. All different use cases.
11:00Absolutely. And this is the goal is to provide assistance to everyone. I'm going to free you.
11:07Thank you. I appreciate that. Thank you very much.
11:10And yeah, remove my headpiece. And so, Olivier, how accessible is this technology right now,
11:16both in terms of cost and availability? Let's go. Let's sit down. Yeah.
11:22It's a few hundred euros for the brain sensors, a few thousand for the exoskeleton. And what we do
11:29with our technology is we're hardware agnostic. We're developing the AI that can basically be used
11:36in any connected device or environment that has sensors. Because we're training the model with
11:43all this information, the brainwaves, the facial muscles, the facial muscles, the brainwaves,
11:49the brainwaves, the facial muscles, heartbeats. Generative AI will be able, down the road,
11:57to recreate the signals, and we will need less and less wearables.
12:03The goal with that is really to have machines to adapt to us and how we feel. There are moments
12:09where I'm too distracted for some task, especially if I'm working on a brain surgery.
12:18Caution, word of caution, I'm not a brain surgeon. You do not want me to operate on your brain. But
12:23we did work with surgeons. Orthopedic surgeons.
12:26Absolutely, in Abu Dhabi that are specialized in the knees of your favorite football players.
12:33And we monitored in real time how their brains interact. So, two surgeons, it was a world
12:38premiere, to understand the difference of how they process information when they are performing
12:44routine check, routine operations, and when they are dealing with the unknown.
12:49And what did those results tell you, and how is that going to change the world of surgery
12:54if we throw it ahead? We can optimize the whole day of work.
13:00Surgeons, like most caregivers, they have operations after operations after operations,
13:06and they're just like every one of us. At some point, you know, you're mentally tired,
13:11it's harder to focus. So, we can better time care delivery in order to minimize errors and
13:18accidents. Another example is what we're doing in dentistry. A lot of us are afraid of dentists.
13:25So, we can monitor in real time the level of stress of a patient with the earbuds in order
13:30to make sure that we wait until the patient calms down in order to perform whatever intervention
13:37we have to do. And little details like that can change everything in care delivery, but also
13:43on a construction site or for knowledge workers. So, it really is about enhancing people's
13:48experience regardless of their cognitive or physical abilities. This can be integrated
13:53into so many different sectors of life. That's your ultimate goal for this technology?
13:56Yes, as long as we do it in a responsible, fair, ethical, and legal way. This is why people have
14:04to be informed. What we did is I did not record your data. Your data is not stored.
14:11We can have aggregated data about a group of people without having information about the
14:16individual that is anonymized, so that we can learn that a group of people works better at
14:21certain times of the day or can perform better. The idea is really to leverage this technology
14:28for good in order to improve safety, whether we're talking about physical or mental health,
14:35in order to improve decision-making, avoiding accidents. There are so many operations and
14:40accidents today that are happening just because we do not have a good relation with technology.
14:46Whomever buys a computer, whichever workstation you have in your office, we have to adapt to the
14:53work environment. We have to make the effort. It doesn't make sense. It would be just like giving
14:59clothes to go on a construction site to someone who's 1.7 meters. You would give clothes of someone
15:08who's two meters. The helmet is too big. The shoes are too big. You don't do that when it comes to
15:12physical safety. We have to do it about psychological safety as well. We customize. This is like haute
15:20couture, but for cognition and physicality. On that note, I wouldn't normally comment on
15:26your fashion, but you've told me that you're wearing very significant trainers. Can you talk
15:30me through these? Yes, because we're talking a lot about high technology, AI, exoskeleton,
15:37but low-tech can help a lot. This is a partnership between a famous brand, Sportswear, and Cafes,
15:46another brand that employs only people with cognitive disability. Explain to people who
15:53can't see at the back what's going on in your shoes. My shoes, one is yellow. The other one
15:58is black. This low-tech can help people who have some cognitive disability remember which side is
16:05which. It's just as simple as that. It's about designing environments that are inclusive,
16:12work environments, everyday life environments, as long as we leverage AI or low-tech for good.
16:20Olivier, thank you so much for coming. Thank you for bringing your technology,
16:24and good luck this summer at the Olympics and Paralympics. Thank you very much.