- 5 months ago
AI For Life
Category
🤖
TechTranscript
00:00Sous-titrage Société Radio-Canada
00:30A startup and a company is a living organism.
00:32You have information going in, information going out,
00:35and really, at the end, you have a mind, sometimes a soul,
00:38but it is a living organism.
00:40It represents a certain type of life.
00:42But life is much bigger.
00:44We have 8 billion humans living on Earth,
00:47and we have more than a trillion spaces around.
00:50But what drives life?
00:52How can AI help to understand the mechanism of birth, death,
00:56the ecology, and the environment is what we want to talk about.
01:00You know, today, when you actually get born in a cell,
01:03you don't know what you're going to become.
01:05You don't know when you have to re-implant a baby in a woman's body to do an IVF,
01:10which embryo has to be chosen.
01:12AI can help to understand, to edit the right genes
01:15and bring off some very severe disease,
01:18to be able to cure disease that existed before,
01:20and to be able to improve the way we procreate
01:23and actually create life at scale.
01:27Babies have diseases that can be, you know,
01:30eliminated by artificial intelligence.
01:33It's also like, you know, using CRISPR technologies, for example,
01:36but it also has a lot of development stories
01:37and development steps that can be improved
01:40with artificial intelligence,
01:41predicting which of the kids will have some development issues
01:44and will have a need to have intervention as soon as possible,
01:48predicting and being able to solve.
01:52Finally, AI will help to understand the genome.
01:54You know, we know probably less than 4% of the genome,
01:572% of the cosmos,
01:58only 1% of the viruses.
02:00The genome can be understood way better.
02:03What are the regions that are coding?
02:04What are the regions that are non-coding are being done?
02:06How AI can help understanding how we can change the genome
02:09to have better, you know, health care at scale?
02:13How we can create therapies as well
02:14will be something that AI will bring a lot of value to.
02:17AI will help to treat patients differently
02:20by being able to be able to predict.
02:22This is something from Okina,
02:23who is going to resist to immunotherapy
02:25before we treat them.
02:27How can we bring preventive medicine at scale
02:29is really what AI will bring,
02:31being able to bring these superpowers
02:33to doctors and physicians.
02:34Today, we cannot make this prediction happen.
02:38AI will change the way we do biology,
02:40how we see the data.
02:42Today, you look at a little microscope,
02:44you see things.
02:45AI can see things differently.
02:47And we will be able to augment and accelerate research
02:50to finally understand why people have cancer.
02:52Why do we die, actually?
02:54Nobody knows why death is around.
02:56There's a lot of questions that have to be asked
02:57and where artificial intelligence can solve
03:00and help giving superpowers to doctors and physicians.
03:03The virus, you know virus?
03:05Maybe everything is a virus around.
03:07You know, 1% of the virus infecting humans
03:09are only known today.
03:11But more than that,
03:12maybe every cancer is a virus.
03:13Which will be the next virus
03:15that will infect you or be the next COVID?
03:18Which is the viruses that can explain
03:19why some persons with some genetic backgrounds
03:22do have a cancer today or not?
03:24We have to use artificial intelligence
03:25to map out things
03:26and be able to make the right prediction
03:28of the future.
03:29The proteins will be built
03:31using artificial intelligence,
03:32how to optimize the right protein
03:34for bioengineering,
03:35bringing the food of tomorrow,
03:37of the chemical that can be used today.
03:39AI will help manufacture,
03:41understand, create, and combine.
03:45Finally, for food,
03:46you know,
03:47artificial intelligence will have to understand
03:49how to select,
03:50for example,
03:50the right almonds
03:51to be selected as the green M&M's,
03:53for example,
03:53because everybody agrees
03:54the green M&M's is the best one
03:55of all the M&M's you can eat.
03:57What's the right classifier
03:58to choose the size,
04:00the taste,
04:00and the color
04:01to be able to get there?
04:02At the nature ecosystem,
04:05how can we preserve our earth?
04:06How can we preserve the forest?
04:08Understand how the spaces
04:09are going to be,
04:10which spaces can remain,
04:11how to build the right ecosystem
04:12and understand the right interaction
04:14between all the scales
04:16of the environment
04:18and all the scales of life.
04:20This is where we're at.
04:21The human health is,
04:22you know,
04:23the right question
04:24and how we will be able
04:25to make the right intervention
04:26and creating how the future
04:27of the population
04:28will be built
04:30using artificial intelligence,
04:31technologies,
04:32hardware,
04:32and processes.
04:35Maybe becoming a better hearse
04:36to live on
04:37and understanding
04:38the water supplies,
04:39how the environment
04:39will be scaled,
04:40and the human population.
04:41And this is why today
04:42we need a new type
04:43of intelligence
04:44that will combine
04:44the human intelligence
04:46and the synthetic intelligence.
04:47We have to create
04:49the right collective world,
04:50the right collective intelligence
04:52that will be a big step
04:53of human and agents,
04:55AI agents,
04:56to understand ecology,
04:58to understand health,
04:59and to understand environments.
05:00And today I'm super proud
05:02to be with Jean-Philippe Verre
05:03that will talk about AI
05:05and the future of health,
05:06SĂ©bastien Boyer
05:07that will talk about
05:08the future of agriculture,
05:09and Renaud Visage
05:10that will tell us
05:11way more about
05:12the future of environments.
05:14Please applaud them
05:15and we look forward
05:16to hearing of them.
05:17Thank you.
05:23Thank you, Thomas.
05:25So I'm JP.
05:27I'm a scientist.
05:28I've been working
05:28for the last 25 years
05:30or so in the field
05:31of machine learning
05:32slash AI on one hand
05:34and biology medicine
05:36on the other hand.
05:37Since day one,
05:38I was obsessed
05:39by trying to cure cancer,
05:41but as a mathematician,
05:42my only way
05:42to try to contribute
05:44to the field
05:45was to develop
05:46some computational
05:47mathematical models
05:48for that.
05:48So what I want to do
05:50in the next few minutes
05:51is share with you,
05:53you know,
05:54through some slides
05:54what I think
05:56are the most important,
05:57according to me,
05:58domains where AI
05:59is likely to make
06:01important contributions
06:02in health
06:03in the short
06:04and long term.
06:06The first domain,
06:07which is quite visible
06:08already,
06:09is that AI,
06:10together with technologies
06:12that are about
06:13imaging,
06:14genomics,
06:14et cetera,
06:15is going to bring
06:16to doctors
06:17what Thomas called
06:18superpowers.
06:19So imagine you go
06:21to see a doctor
06:21or you or your family.
06:22You would like
06:23your doctor instantly
06:24to be able
06:24to collect information
06:26about yourself,
06:27about your disease,
06:28and make the right decisions.
06:30Identify,
06:31make a good diagnosis,
06:32make a good decision
06:33to treat.
06:34This is what doctors
06:35are trained for.
06:36Now it's pretty clear
06:37that the possibility
06:38these days
06:39to collect
06:40large amounts of data
06:41about you,
06:42so this could be
06:43through images,
06:44through genomics,
06:46through other means,
06:47brings lots of data
06:48and the possibility
06:48to train AI systems
06:50to replicate
06:51or even improve
06:53what doctors are doing.
06:54Thomas gave the example
06:55a bit earlier
06:56of a system,
06:57for example,
06:57that is developed
06:59and already commercialized
07:00by Alkin,
07:01which is an AI company,
07:03to do that,
07:04to help doctors
07:05for diagnosis of cancers,
07:06meaning you have
07:07a suspicion of cancer,
07:09you go to the hospital,
07:10a biopsy is taken,
07:11and up to now
07:13there was a doctor
07:14called a pathologist
07:15that was looking
07:16at the image
07:17in order to make
07:18a diagnosis
07:18and suggest a treatment.
07:20More and more,
07:21this doctor will be helped
07:22by AI systems
07:24which are trained
07:25to recognize the cancer
07:28and more importantly,
07:29to make predictions
07:30about what would be
07:31the right way
07:31to treat the patients.
07:33And what we observe
07:34is that not only
07:35can the AI help
07:37the doctor go faster,
07:38but also it can do things
07:40that the doctors
07:41cannot see yet
07:42because the systems
07:43can be trained
07:44on more images
07:45and find subtle correlations
07:47between what you see
07:48in the biopsy
07:49and what you see
07:50in the hospital
07:51when you treat someone.
07:53So that would be
07:54the first big domain
07:55which is augmenting doctors.
07:57Now there's a second domain
07:58where obviously AI
08:00is already starting
08:01but will increasingly
08:02make big contributions
08:03is to give good doctors
08:06to everyone.
08:07Today it's fair to say
08:09that if you have a cancer,
08:11it's better to live
08:12in a city
08:13or in a country
08:14where you have good doctors,
08:16good medical systems
08:17and it's not the case everywhere.
08:19AI brings the possibility
08:21to put on your smartphone
08:22maybe
08:23or through some local
08:24networks of computers
08:25some expertise,
08:27some know-how
08:28that is not everywhere yet
08:30but that can be exported everywhere.
08:32Just to give one example,
08:34a few years back
08:34I was privileged to work
08:36with doctors without border
08:38I was at Google at the time
08:39where they developed
08:40some application
08:41that is a smartphone-based application
08:44to help doctors
08:45in any country
08:46make the right diagnosis
08:48for antibiotic resistance.
08:50That's typically a problem
08:51that we know how to solve
08:53in Paris
08:54in the Pasteur Institute
08:55but which is harder to solve
08:57when you're in a country
08:58with limited doctors.
08:59And AI is a way
09:00to bring that
09:01to some machines
09:01that then are dispatched.
09:05The third domain
09:06where AI
09:08is clearly already
09:09bringing novelties
09:10is in bringing new treatments
09:11because it's good
09:12to have a good doctor
09:13or to have it everywhere
09:14but we're still lacking
09:16treatments for many people.
09:18There's lots of
09:18unmet medical needs.
09:20People still die of cancer.
09:22People still suffer
09:23from any diseases
09:24and clearly AI
09:25is deployed
09:26and used these days
09:27in many companies
09:29including at Aukin
09:30to help find
09:31new candidate treatments
09:33or to help
09:34the pharmaceutical industry
09:36develop them faster
09:38and for cheaper.
09:39Just for those of you
09:40who are not familiar
09:41there is a statistics
09:42that is an awful statistics
09:43of the industry
09:44as a whole in pharma
09:45which is that
09:47when a drug
09:47is developed
09:48the success rate
09:49between the time
09:50the drug is given
09:51to a patient
09:52to test it
09:53and it is validated
09:55to be given
09:55to everyone
09:56the success rate
09:57is about 10%
09:58meaning 90%
10:00of the drugs
10:01that are developed
10:01by the pharma industry
10:03end up not working
10:04or being toxic
10:05or being stopped.
10:06What we want to do
10:08with AI
10:08is obviously
10:09change that 10%
10:11to maybe 20, 30
10:12maybe one day 50%
10:14and when you think of it
10:15this means
10:15more drugs
10:17more treatments
10:18cheaper and faster
10:19right
10:20and this is clearly
10:21things where AI
10:22is helping a lot.
10:24Finally
10:25I will mention
10:26that there is
10:28an even broader field
10:29that AI
10:29is starting to help with
10:31is really to
10:32understand life
10:33and understand biology.
10:35When you think
10:36of biology
10:37these days
10:38there is lots
10:38of technologies
10:39as I mentioned
10:41images
10:42I mentioned
10:43genomics
10:44these are just a few
10:45there is proteomics
10:46there is lots of way
10:47to collect data
10:48and so what we have seen
10:49in AI
10:50is that
10:50if you do the parallel
10:52between this domain
10:53and for example
10:54natural language processing
10:56or computer vision
10:58what we know
10:59to do today
10:59is if you have
11:00enough data
11:01like enough texts
11:02written by humans
11:04then some AI system
11:05can be trained
11:06just on the raw text
11:08to learn the language
11:09right
11:10I don't say
11:11it's able to think
11:12but when you use it
11:13every day
11:13with large language models
11:15you realize
11:16that the model
11:16has learned something
11:17about human language
11:19now think
11:20you do the same
11:21not on human language
11:22but on biology
11:23you take biological data
11:25you take DNA
11:26that is collected
11:27in the environment
11:28you take images
11:29of tissues
11:31in cancers
11:31and if you have
11:32enough of them
11:33and feed a neural network
11:35or some AI system
11:36with them
11:36could the AI
11:38learn the language
11:39of biology
11:40and biology
11:41is really a science
11:42where we don't know
11:43the equations yet
11:44it's not like physics
11:45in biology
11:45we know lots of facts
11:47we have some understanding
11:48but maybe now
11:49is the time
11:49where AI
11:50will help
11:51create a new science
11:52of biology
11:52this is challenging
11:54this requires deep thinking
11:55there's lots of research
11:56going on
11:56we just created
11:58a new company
11:58named Bioptimus
11:59that is focusing on that
12:01long way to go
12:02but I think this is really
12:03where the field is going
12:04using AI
12:05to better understand biology
12:07and ultimately
12:08if you better understand biology
12:09this is when you will see
12:10a lot of new treatments
12:12a lot of new applications
12:13in life sciences
12:15so I'm quite optimistic
12:16about AI in biology
12:17and I hope my next speakers
12:19co-speakers
12:21will shed light
12:22on different applications
12:23of it
12:29thank you JP
12:31hi everyone
12:31good afternoon
12:32I'm Sebastian Boyer
12:33co-founder and chairman
12:34of a company called
12:35FarmWise
12:36at FarmWise
12:37we essentially use AI
12:39to make robots
12:40that help farmers
12:41grow better
12:42and healthier crops
12:44so I'm excited
12:45to tell you today
12:46about why
12:47I think AI
12:48is going to revolutionize
12:49agriculture
12:50maybe there are just
12:51two big reasons
12:53the first one is
12:54that agriculture
12:56as a whole
12:56globally
12:57is going to be facing
12:58tremendous amount
13:00of challenges
13:00over the next 20 years
13:02the first of which
13:04is that
13:04the population
13:05agriculture needs to feed
13:07is growing rapidly
13:08about 20%
13:09between today
13:10and 2050
13:11that's in part
13:12due to population growth
13:14but also in part
13:15due to
13:16really good news
13:17that a lot of people
13:18are getting out of poverty
13:20that's really good news
13:21for the world
13:22very challenging
13:23for the agriculture industry
13:24we need to feed
13:26again 20% more people
13:27by 2050
13:29but the real challenge
13:30in doing this
13:31is that
13:32we need to do
13:33how agriculture
13:34even society as a whole
13:36need to find ways
13:37to produce food
13:3820% more food
13:39with drastically
13:41fewer resources
13:42and that's the big challenge
13:45land
13:46the amount of land
13:47that is available
13:48for agriculture
13:49to grow food
13:50is not growing at all
13:51the projections
13:52are pretty clear
13:53we already use
13:5450% of viable land
13:56today on earth
13:57to grow food
13:59we can't grow
14:00the land
14:00on which we grow crops
14:02the second challenge
14:03is that
14:05fresh water
14:06is getting
14:07more and more scarce
14:08today agriculture
14:10globally
14:10uses 70%
14:12of all fresh water
14:14so that's already a lot
14:15we cannot grow this
14:17actually
14:17with climate change
14:19this is becoming
14:19harder and harder
14:20to do
14:21so we need to start
14:22learning to grow food
14:24with drastically less water
14:26that's the second challenge
14:28the third challenge
14:30is that
14:31more and more people
14:33are finding
14:33better
14:34easier
14:35more enjoyable
14:36jobs
14:37outside of agriculture
14:39agriculture jobs
14:41and again
14:41I've spent
14:42the better part
14:42of the past
14:4310 years
14:44on field
14:45with farmers
14:45so I've seen
14:46a lot of these jobs
14:47first hand
14:48they're very
14:49very challenging
14:50they're some
14:50of the toughest jobs
14:51that any economy
14:53has
14:54you are essentially
14:55under the sun
14:56working
14:57down the field
14:5810 hours a day
14:59so it's really good news
15:00that fewer and fewer people
15:01find it attractive
15:02to work in agriculture
15:03but it's also
15:04a big big challenge
15:06and labour shortages
15:07are the number one
15:08problems
15:09that a lot of
15:10countries
15:10are facing
15:11when it comes to
15:12continuing on their
15:14agriculture systems
15:15today
15:15that's the third
15:16challenge
15:17and the fourth challenge
15:19is related to
15:20obviously
15:20climate change
15:21you may not know this
15:23but agriculture alone
15:25is responsible
15:26for 20%
15:27of greenhouse
15:28gas emissions
15:29globally
15:2920%
15:30one-fifth
15:31that's huge
15:32right
15:33this is in part
15:34due to
15:35cattle ranching
15:36so about
15:375%
15:38and the rest
15:39is mostly due to
15:40to one single
15:41process
15:42one single
15:43industrial process
15:44that we use
15:44to make
15:45fertilizers
15:47so that's
15:48these are
15:48really big challenges
15:49we need to find
15:50a way to produce
15:51food with
15:52less water
15:53less labour
15:54and drastically
15:54less carbon emissions
15:57so that's the first
15:58reason why I'm excited
15:59about AI for agriculture
16:00the second reason
16:01is that
16:01I think it can work
16:02I think AI can actually
16:04help us reach those goals
16:05and the reason for that
16:07is that AI
16:08is particularly
16:08well-studied
16:09as you obviously know
16:10to take huge amount
16:12of data
16:12and find patterns
16:13to make processes
16:15more efficient
16:16that's exactly
16:17what agriculture
16:18needs
16:18agriculture
16:19is based
16:21on tremendous
16:22amount of data
16:22weather data
16:23is obviously critical
16:24to make farming decisions
16:26genetics
16:27that plays into
16:28the growth
16:30of every seed
16:31and a whole set
16:32of variables
16:33from soil data
16:35to microbes
16:38that are surrounding
16:39each plant
16:39these are variables
16:41that play a big role
16:42in the way
16:43plants grow
16:45so I want to show you
16:46three examples
16:47of how AI
16:47is completely used today
16:49to make agriculture
16:51drastically more efficient
16:52on the resources
16:53that are getting
16:53more and more scarce
16:55the first is
16:56the ability
16:57to make better decisions
16:58here's a picture
16:59of a
17:00pretty modern
17:01sensor
17:02that you can put
17:03in the ground
17:03to measure things
17:04like temperature
17:05soil moisture
17:05microbes
17:07and using that
17:08information
17:08and correlating
17:09that information
17:10with for instance
17:11weather data
17:12from publicly
17:14available sources
17:15of data
17:15and from
17:17satellite imagery
17:19you can start
17:20to find patterns
17:21and correlations
17:22between these variables
17:24and the outcome
17:26in terms of yield
17:26these patterns
17:28are helping today
17:29farmers make better
17:30decisions
17:31on when to water
17:32and when to fertilize
17:33that alone
17:35has the potential
17:36to drastically reduce
17:37the amount of water
17:38and the amount of fertilizer
17:39that we use
17:39and therefore
17:40get us closer
17:41to the goals
17:42that I mentioned earlier
17:45the second big use case
17:48in my opinion
17:49for AI
17:49in agriculture
17:50is genetic engineering
17:52the ability
17:53for
17:55bioscience companies
17:56to find ways
17:57to edit
17:58the genome
17:59of plants
18:00so that
18:00they consume less water
18:02or they are able
18:02to retain
18:03more water
18:04they are able
18:05to retain
18:05more fertilizer
18:06and therefore
18:07decreasing the need
18:08for us
18:09to use
18:09these resources
18:12one fun fact
18:14is that
18:14the genome
18:15of wheat
18:16like a simple crop
18:18like the most common crop
18:19one of the most common crops
18:21on earth
18:21wheat
18:22the genome
18:23of that plant
18:23is five times bigger
18:25sorry guys
18:26than the human genome
18:28there is a tremendous amount
18:30of information
18:30in that genome
18:31that companies today
18:33thanks to AI
18:34are starting
18:34to be able to use
18:35they are using AI
18:37and CRISPR
18:37to understand
18:39which genes
18:40to change
18:41in order to reach
18:42the outcomes
18:43that we want
18:43these outcomes
18:44again
18:44are things like
18:45better water efficiency
18:47better fertilizer efficiencies
18:50so that is
18:51incredibly exciting
18:52with the same
18:53with the same resources
18:54and better
18:56genetic engineering
18:57we can reach
18:58or we can get closer
18:59to reaching
19:00some of these goals
19:02so that's the second use case
19:03and the third use case
19:04is the general trends
19:09of building better systems
19:10better machines
19:12more precise machines
19:13on the farm
19:15some of these machines
19:17for instance
19:18this one
19:19uses computer vision
19:21for instance
19:21to recognize plants
19:22and essentially give
19:24to your regular tractor
19:26or your regular farming
19:27implement
19:27that sits in your garage
19:29when you're a farmer
19:31obviously
19:31to give these machines
19:33eyes and brains
19:34so that they can make
19:36better decisions
19:36better decisions
19:38in terms of
19:38which plant
19:40to act on
19:41how much fertilizer
19:42to apply
19:43how much herbicides
19:46or pesticides
19:46to apply
19:47and the overall
19:49idea of these systems
19:50is essentially
19:51to enable farmers
19:52to start making decisions
19:54not only at the field level
19:56like this is the case today
19:57but more and more
19:59at the plant level
20:01and that's really exciting
20:02because once you start making decisions
20:04at the plant level
20:05then you can drastically increase
20:07the efficiency
20:07in how you use
20:09the resources that I mentioned
20:10water
20:10fertilizer
20:11and land
20:13so all of them put together
20:15and these are just
20:15really the tip of the iceberg
20:17like this is
20:18these are early examples
20:19of how AI today
20:20is having a big impact
20:22on agriculture
20:22but I'm convinced
20:23that this is only the beginning
20:28this is probably not
20:30how the future
20:31of farming looks like
20:32but I'm convinced
20:34that farming
20:35is going to
20:36go through
20:37a huge period
20:39over the next 20 years
20:40of extremely fast
20:42paced innovation
20:43we mentioned some of them
20:45again it's only the beginning
20:46I'm very excited
20:47for engineers
20:49and innovators
20:50in that field
20:50because I think
20:51the potential is huge
20:52it's not only a potential
20:54but it's a necessity
20:56these resources
20:57that I mentioned earlier
20:58they're going away
21:00no matter what we say
21:01no matter what we think
21:02they're going away
21:03and the population is growing
21:05so these technologies
21:07are not a luxury
21:09like they're a necessity
21:10and there is no question
21:11at least in my mind
21:12that they're going to happen
21:13over the next couple of years
21:14that's why I'm excited
21:15about AI for agriculture
21:17thank you
21:25hi everyone
21:26my name is Renaud Visage
21:28in a previous life
21:30I co-founded Eventbrite
21:31the event ticketing platform
21:33that you might have used
21:34in the past
21:35I'm now a founding partner
21:37at Slate Venture Capital
21:38we're a new fund
21:39focusing on climate tech
21:40we invest in Series A and B
21:43in two great companies
21:44and a lot of them
21:45have AI at their core
21:49there's been a lot of data
21:50available in the last 20 years
21:52that is extremely useful
21:54for a climate tech
21:55and the first true application
21:58for AI
21:59is around climate adaptation
22:03I'll pass on this one
22:06we used to make decisions
22:07based on what happened
22:08in the past
22:09so this time is over
22:11things are changing really fast
22:14temperatures are rising
22:16so you can't just trust
22:17the stone anymore
22:18for making decisions
22:20that impact your business
22:22climate adaptation
22:23is something
22:23all businesses
22:24will have to face
22:25especially the ones
22:26with physical assets
22:28there
22:29there are a lot of modeling
22:30that is possible today
22:32thanks to computational advances
22:35supercomputers
22:35and AI modeling
22:37that were not possible before
22:39is the site
22:40that is next to a river
22:42going to be flooded
22:42in 20 years
22:43when the oceans have melted
22:46and we're two meters higher
22:48will there be landslides
22:50are these going to create
22:51more potential for forest fires
22:54all this is potentially calculatable
22:57by or can be calculated
22:59by AI today
23:00we invested in climate tech
23:02for example
23:03that does exactly that
23:04it helps banks
23:06real estate developers
23:07any company that has physical asset
23:10understand under different climate scenarios
23:14what are the implications
23:16in the physical world
23:17for these properties
23:23as I'm speaking about
23:24computational power
23:26there's also a very short time risk
23:28that AI will make things worse
23:30we can't ignore that
23:32it requires immense computing power
23:35and energy
23:36to make these models
23:38be trained on the right data
23:40do iterations
23:41and then do inference
23:42when you want to answer
23:43specific questions
23:44for these models
23:45again
23:46I'm a tech optimist
23:47I think
23:47we have some of the brightest minds
23:50trying to figure out
23:51what the next generation of chips
23:52will look like
23:53and yes
23:54the software AI developers
23:56are pushing the boundaries
23:57we as consumers
23:59are embracing it
24:00and using it on a daily basis now
24:04but I think we'll see
24:06great advances
24:07in the amount of energy needed
24:09to run these AI models
24:11in the future
24:12maybe not in five years
24:14but in ten years
24:15definitely
24:16and it has a real impact today
24:1730% of Microsoft's emissions
24:21increased by 30%
24:22since 2020
24:25mostly due to the AI needs
24:27of the world
24:28so this is real
24:29and present danger
24:30but hopefully we'll overcome
24:32the next big field
24:34of application
24:35is the energy grid
24:36this is a picture
24:38of my beloved San Francisco
24:39where I spent so many years
24:41the grid there
24:42was built in the 60s and 70s
24:44and you can see
24:45that it shows its age
24:47we're adding
24:48a lot of new sources of energy
24:51last year alone
24:52we added 50% more renewable energy
24:55that existed in 2022
24:57it's been decentralized
24:59on every day
25:01we're adding
25:02solar rooftops
25:04on commercial buildings
25:05on residential buildings
25:07our cars
25:07on becoming battery storage
25:09the complexity of the grids
25:11and running it
25:12I think is already beyond
25:14the capabilities
25:15of the human mind
25:17or the individuals
25:18that are part of these energy grids
25:21the emissions related to losses
25:23in the grid are huge
25:25it's one gigaton a year
25:26it's several percent
25:28of our global CO2 emissions
25:31so there I think
25:34it's the type of applications
25:36where AI shines
25:37there's multiple parties
25:39having different interests
25:40there's trading
25:41there's powering
25:43there's storage
25:44there's consumption
25:45there's residential
25:46there's businesses
25:47it's insolvable
25:49without someone
25:51something that's smarter
25:52than the average person
25:53to decipher
25:55and there's a lot of data
25:56like all these have meters
25:58so we understand
25:59what consumes what
26:00what produces what
26:02at what time
26:02and at what price
26:04a very rich field
26:05if you're looking for
26:06new business ideas
26:09then you have
26:10the urban environment
26:11and the way
26:13the things that consume energy
26:15I mean it's
26:17again
26:17built over time
26:19you see all these
26:20AC units
26:21they don't talk to each other
26:23they all
26:23and come
26:25backed by operators
26:27who have
26:27have very limited knowledge
26:28so how do we bring
26:30intelligence to that
26:31how do we account
26:32for the weather forecasts
26:33how do we account
26:34for energy prices
26:35to decide when
26:36the AC is running
26:37or not
26:38very relevant
26:39for large office buildings
26:40that have
26:41centralized power
26:42a lot of
26:44connectivity
26:45of all these devices
26:47and that's
26:48another field
26:49I think
26:49where AI will shine
26:51and will overcome
26:52all the limitations
26:53that we have
26:54to manage
26:55this
26:55very highly
26:58consuming
26:58in energy
26:59sector
27:01and then you have
27:05construction
27:06concrete
27:07and steel
27:08alone
27:09accounts for
27:09almost 15%
27:11of our global emissions
27:12it's huge
27:14there I think
27:17AI can play
27:18at different levels
27:19it can help
27:20with the manufacturing
27:21of concrete
27:22but you can also
27:23find ways
27:24to use less concrete
27:25for example
27:26a slab
27:26does it need to be
27:27fully concrete
27:28or can we find
27:29different shapes
27:29different additives
27:31different ways
27:32of building it
27:33or pre-building it
27:34for example
27:35that have
27:35lead to the same
27:37physical characteristics
27:39but use less
27:40of the material
27:41same for steel
27:42can it be replaced
27:43by wooden shapes
27:45and the complexity
27:46of this R&D
27:47I think
27:48can only be solved
27:48by trying
27:49hundreds of thousands
27:51of permutations
27:52so many materials
27:53you could use
27:54so many shapes
27:55you could try
27:55so many dimensions
27:56you can create
27:57this into
27:59and there
27:59the complexity
28:00again is enormous
28:01and will not be solved
28:02by humans
28:03doing things
28:04one by one
28:08shipping and
28:09transportation
28:09about 20%
28:11of our global emissions
28:11and as you can see
28:14there's a lot of room
28:15for improvements
28:17talking to startups
28:18I realized
28:19ships for example
28:20go full speed
28:22to their port
28:23and then they wait
28:23for two weeks
28:24for a slot to open
28:25for them to unload
28:27so inefficient
28:27if they were driving
28:29or riding the boats
28:30for 30% of the speed
28:33or 40% of the speed
28:34they would save
28:3540, 50, 60%
28:37of the energy
28:37they consume
28:40autonomous vehicles
28:41like the revolution
28:42is underway
28:43huge consumption
28:44of AI models
28:45and the need
28:46for representing
28:47and connecting
28:48this physical world
28:49with the devices
28:51that will allow
28:53transportation
28:53in the future
28:56robots
28:57have many uses
28:58I hope this one
28:59will not become
29:00the mainstream use
29:01for them
29:01but they become
29:02the hands
29:03and the eyes
29:04of the AI
29:05and I think
29:06there's a lot
29:06of potential there
29:07especially in industry
29:08where we can automate
29:10reduce waste
29:11find better ways
29:12to build things
29:13automated lab
29:14for example
29:15R&D is a big piece
29:16of exploration
29:18for finding
29:19alternative
29:20more sustainable
29:20sources of material
29:21for example
29:22what if the AI
29:24created
29:24all the
29:25possible permutation
29:26that ought
29:27to be tried
29:28and then
29:28we had an
29:29automated lab
29:30that was controlled
29:31by the AI
29:31to try all
29:32the different
29:33tests
29:34and do that
29:35much more efficiently
29:36than the humans
29:37and I think
29:38our time is out
29:40but the last one
29:41is industry
29:41and there
29:42there's
29:44a shitload
29:45of possibilities
29:46for AI
29:47to optimize
29:47there's a lot
29:49of data
29:49again
29:49a lot of possibilities
29:50and hopefully
29:51the AI
29:52won't team up
29:53with the robots
29:54to decide
29:55that the humans
29:56are unnecessary
29:57in the equation
29:58thank you everyone
29:59we hope to have
30:00you shown
30:01a snapshot
30:01of what AI
30:02for life can do
30:03wish you a very nice
30:04end of VivaTech
30:05and please come to see us
30:06if you have any questions
30:07bye bye
30:07thank you everyone
30:10is this Remo
Comments