Player is loading...

Embed

Embed code

Transcriptions

Note: this content has been automatically generated.
00:00:00
mm good asked them listen jasmine uh my name is a lot of and that and i will talk about ross
00:00:07
which is any operating system for roberts was artificial
00:00:12
a little cheryl vision of future and our place in this future but first let me talk about my passion
00:00:19
and my passion is wine or our wine especially suisse fine
00:00:25
half of it going here in valet because it's just fantastic it is
00:00:30
what i've paid my forever it's it's have fantastic powerful taste
00:00:34
with different notes and that one of the reasons for that is
00:00:37
because they scrape finds allocated on mountain sides and um
00:00:42
these games sap taste more potential complexity but also makes
00:00:47
complicated the life of wine maker because many operations
00:00:50
are not easy to do on the mountains and there's an example is a chemical treatment so um
00:00:57
each grapevine needs to be treated to civil times through season and it is
00:01:02
toxic operation and enters for human so it's ideal place for machine
00:01:07
and got in valley you can see helicopters do this job it is
00:01:10
quite costly and uh the problem is of massive pollution of
00:01:15
environment with this kind of treatment so ah when you speak to
00:01:20
'em wine makers we see that uh actually they need only
00:01:25
upper part of plants to be treated precisely of course helicopter could
00:01:30
not do these and the promising solution could be drawn
00:01:34
you can imagine the drawn flying for rules and percent it three these uh plans
00:01:41
but uh it will we we just need to develop there were nice
00:01:46
and um feasible way to control such a drum so we have a three i imagine that
00:01:53
you can teach the strong and you can lower from you and so drawn tries to identify the park on the plant
00:02:01
that you can't that that must be treated you just come from the action and it
00:02:06
treats and it becomes more and more intelligent and after suit opens it just
00:02:12
goes itself into this task itself so sounds good um
00:02:17
i think yes but ah mm maybe these machine
00:02:20
can be quite costly one and that it'll be a big investments for small one years and that
00:02:25
he didn't wanna have a hundreds of small one yards and though they don't have enough budget for such thing so
00:02:33
let's consider you don't not need to buy this machine it'll you only can
00:02:38
take it for and for a couple of days and they only for this
00:02:42
service that's like you pay salary for in a in a plea so
00:02:50
wind treatment whiny arts treatment is only one example
00:02:54
off how self learning robot can feel feet into the real business and you can imagine
00:03:02
many different panes of different businesses that can be addressed
00:03:07
in this way making life easier to people
00:03:12
and so this is always in a world full of intelligent machines that lower from people
00:03:19
and work for them and this is what the ropes about it
00:03:24
so the problem address is actually a huge amount of knowledge and skills which he
00:03:31
made to have you just can not be problem to robert because this programming
00:03:36
becomes a bottleneck and it's just impossible to program all this stuff and
00:03:43
um these actually declines or
00:03:47
that blocks many companies very small and medium companies to go to this market uh with the
00:03:53
new products and for example there are many drone companies who could do this job
00:03:57
but for the moment they can for example parrot ah they can't so they all are
00:04:02
a potential customers because what we do we offered to replace for this programming
00:04:07
with the teaching and of course it implies that self that future robot will
00:04:15
need to be self learning and people based on another kind of software
00:04:19
it's all this is what we do we do a
00:04:23
new operating system for robots powered with artificial intelligence
00:04:28
can stay there for the visual their chances that the neural networks i think many of you know
00:04:33
this exactly and the neural networks it's just that well detailed low level model of fire
00:04:40
one of the neural cortex which is are part of human brain these models
00:04:45
can solve many different tasks but what are never once and each
00:04:50
plastic what requires enormous computing power and huge data sets
00:04:55
our approach is different uh we mortal that human prey on
00:04:59
higher level and we found a way to avoid programming
00:05:04
all of these billions of single neurons and we just focus on main functions
00:05:09
on learning and remembering and our model works thousand times faster and
00:05:17
robert can then in real time more of this this
00:05:21
algorithm is comparable to learn more and more complex
00:05:27
behaviour day after day and uh as human intelligence these
00:05:33
official vengeance is capable of the requirement different skills
00:05:38
for different applications and this is our main competitive advantage and this universality
00:05:45
actually gives us an ability to go to that all segments of robotic market
00:05:50
which is that already peak and it you see it grows ten times in years
00:05:54
six years and so um of course we have many competitors and um
00:06:01
general that evaluated too too big parts first part is there
00:06:06
frameworks for robots which is is right now like rose or would
00:06:10
be that's um this is a simple frameworks that connects
00:06:16
uh that that might provides a uh an interface to robert sensors and actuators for governments to
00:06:23
and and uh and they don't have any learning it's you need any intelligence another big
00:06:28
part is um the neural network providers like in video for example and they provide some
00:06:34
solutions for these um for some specific areas like a a computer revision or um
00:06:41
uh it's speech recognition but this is not about there
00:06:46
behaviour but probably behaviour and none of them can do loading in real time
00:06:51
so this is a real specific thing for us and with that
00:06:55
we actually have that for the moment quite modest goal we we achieve
00:06:59
to um uh we talk to take a three to five percent
00:07:04
of these market to sort of these three to five
00:07:08
percent of market into solvent twenty five um
00:07:12
and um uh which will we have already around eighty million
00:07:17
dollars so where what is writable come from actually uh
00:07:23
in in in the centre of our business model is it robust foundation uh
00:07:28
which owns the operating system and licensee to manufacture of robot for free
00:07:35
or manufacture um gives a um build a robot then uh provided to the user
00:07:42
uh as a series and user pays robot the salary uh which airs
00:07:47
um uh which of them set up for the for the job that's introvert
00:07:53
tries to do the job better and better to or no money
00:07:56
i like like a human and robust foundation charger small fees from every payments into
00:08:03
this uh we do this with the transaction fees and these fees at boeing
00:08:08
to the pro to developers uh for their contribution to the operating
00:08:12
system and to another type of country because the teachers
00:08:16
who actually teacher robot some specific skills that are uh needed
00:08:22
and uh well one of the system looks like
00:08:25
the robot shares its celery with all its creators and
00:08:30
contributors so this is it actual sure recording
00:08:35
um we have a bar ambitious project and we know these and uh we have over ambitious team so my name
00:08:41
is a good got offended that i'm interpret over the fifteen years uh in i. t. in healthcare and um
00:08:49
i have a next experience of failures and successes uh my uh_huh
00:08:54
i overall they raise their own twenty five million dollars for all of
00:08:57
my projects interactions with salt and my passion is to create
00:09:02
um new businesses business opportunities and changeable to better
00:09:07
uh sedition ski is uh our science you
00:09:10
are a a he's uh he has a twenty five years of it really and um
00:09:17
uh_huh she was well with um he peeled what he found it one of the
00:09:23
first russian companies machine doing and uh sold it single years ago to
00:09:28
be corporation and currently he is a science director of the centre of artificial
00:09:34
intelligence in one of the leading russian universities uh m. i. p. t.
00:09:39
so he specialist to create new recording pain to understand the human mine
00:09:45
uh we have archive you sound or a a famous
00:09:47
yeah interpreter ah who was former uh investment manager
00:09:52
in these to global quite famous investment fund uh don't invest of the phase will enable barked feature
00:09:59
and now he's ahead of official intelligence and biggest russian didn't come to the communication company and he is a
00:10:06
main passion is to create the disruptive products that uh demanded biker consumers uh but okay we love
00:10:13
he's the only guy in our project that most fashion one but the only one who has
00:10:17
no clue about project itself and packet challenges also always challenges as of some questions
00:10:24
uh it help us improve uh what we do and he's uh doing operations
00:10:28
right b. management and we have uh some other people who are not
00:10:31
present here in software management and the and i'm a financial overlay we have
00:10:37
a great team uh and also have a serial um several um
00:10:43
advisers and we're looking for me once but uh how we or in spite of all these expertise we
00:10:49
get that to the t. we understand that project of such scale just can not be down
00:10:54
by a small t. so we're going that's why but we're
00:10:58
creating yeah rob was foundation it'll be nonprofit foundation and
00:11:03
we will engage open community to join to contribute um to
00:11:08
support and develop this project and um so um
00:11:13
everyone can join a and we'll be happy if you this so how we
00:11:17
wanna make it we have a three main pillars in our action plan
00:11:21
fund raising a product development and the sales and marketing and uh
00:11:25
we're going to make it creates you all the useful to
00:11:29
get money to the start developing in p. p. um of next
00:11:33
year and work with the early adopters um and base
00:11:38
case study was early adopters uh will make some publications and confront pops
00:11:44
which allow us to raise more interest and that uh actually run the main find rising company i seal
00:11:50
and we aim to to to to find ways around forty
00:11:54
million dollars and uh this will allow us to launch
00:11:58
uh the operating system to us and to this point we will uh have around conduit of active customers
00:12:04
will to work with them before and the after aggressive marketing campaign we're going to have a
00:12:11
thought of customers and so this is a this is our plan and uh in terms
00:12:16
of financial so we estimate the overall investments in the project around forty one median
00:12:22
a dollars and um we expect to get profitable in five years without transactions
00:12:28
and uh it was a change for us to demonstrate actually the type
00:12:32
because um we have a quite um a universal thing so uh we decided to
00:12:38
to demonstrate a prototype and as a prototype a task which was um
00:12:44
moved which was um proposed was basically my in that could you know from what yep uh
00:12:49
we we might we could go and they he said okay you have some demonstrations but
00:12:54
uh he was a benchmark task you you need to solve and um there's so a benchmark
00:12:59
is uh is call more mountain car you see this small car between two mountains
00:13:04
and uh uh the goal is to get to the flight of the wrong moment right multi
00:13:09
but the problem of the car is that it does not have enough what our power to get there
00:13:14
a duty so it needs to to get a clue clever solution and so we um at this
00:13:20
in a standoff environment in open a. g. so we connect this open age into our
00:13:26
the brain prototype and you'll see the um real time demonstration of how it works um
00:13:34
actually i'm so this is environment this is uh connect to the server and first bring does not control
00:13:41
of the car it is it's working from switch uh it just collecting rent the movements of car
00:13:47
to know some information about the world the noisy that car stops and at that moment silver begins to
00:13:53
think over there the the the information and uh finding it's it's it's a in a strategy
00:14:00
and you can see that uh our model takes root few amount of data it is lesson and
00:14:06
i mean it's you know one single computer how that went usual net neural networks work
00:14:12
on multiprocessor environment it's huge amounts of data and no silver is
00:14:17
controlling the car and you see that um the stages
00:14:20
of courses to swing from one side to not as uh it will take i think a couple moments to these
00:14:29
i actually yeah it's uh uh uh yeah so ah we we
00:14:38
of course understand that is a very simple uh up in
00:14:42
example where it easy robot but this is the first step and
00:14:46
you know that every journey to the cell than miles
00:14:50
begins with this for step so uh this is at the stage where we don't know but we think it will be actually exciting
00:14:57
julie absolutely exciting and um uh and speak in our and i
00:15:02
sent out of time but i i i want to
00:15:05
think um i see for the uh for this meeting three weeks
00:15:09
because i think we made a good progress absolute great prose
00:15:12
with our project and we were what we try to be more concrete and more precise about what we're doing and finally
00:15:19
i see made as to develop this prototype if not i see maybe a we we should not do this
00:15:24
and a just war with our mathematical models um that so

Share this talk: 


Conference program

Jérôme Vasamillet introduction
Jérôme Vasamillet, Program Director InnoPeaks
17 Sept. 2018 · 2:03 p.m.
119 views
BetterSkillz Pitch
Madan Raj Rajagopal
17 Sept. 2018 · 2:17 p.m.
BetterSkillz short demo
Madan Raj Rajagopal
17 Sept. 2018 · 2:30 p.m.
BetterSkillz Q&A
Madan Raj Rajagopal
17 Sept. 2018 · 2:30 p.m.
VERISWISS Pitch
Javier Gutiérrez Fernández
17 Sept. 2018 · 2:42 p.m.
VERISWISS Q&A
17 Sept. 2018 · 2:58 p.m.
MoneyID Pitch
Lucas Oliviera dos Santos
17 Sept. 2018 · 3:05 p.m.
MoneyID Q&A
Lucas Oliviera dos Santos
17 Sept. 2018 · 3:19 p.m.
Robos Pitch
Igor Pivovarov
17 Sept. 2018 · 3:30 p.m.
Robos Q&A
Igor Pivovarov
17 Sept. 2018 · 3:45 p.m.
Volt-Face Pitch
Nader Erfani
17 Sept. 2018 · 4:32 p.m.
Volt-Face Q&A
Nader Erfani
17 Sept. 2018 · 4:47 p.m.
LeukoCounter Pitch
Gabriel Minoru
17 Sept. 2018 · 4:53 p.m.
LeukoCounter Q&A
Gabriel Minoru
17 Sept. 2018 · 5:03 p.m.
TellTheHotel Pitch
Aldo Polledro
17 Sept. 2018 · 5:12 p.m.
TellTheHotel Q&A
Aldo Polledro
17 Sept. 2018 · 5:27 p.m.
Nicolas Loeillot, ceremony awards
Nicolas Loeillot, Groupe Mutuel CIO
17 Sept. 2018 · 6:28 p.m.
François Foglia, ceremony awards
François Foglia, Idiap Research Institute
17 Sept. 2018 · 6:34 p.m.

Recommended talks

Keynote: Get Ready for Change!
Michael Jackson, Global conference speaker, author & broadcaster at theothermichaeljackson.com
23 June 2015 · 9:19 a.m.
221 views