Player is loading...
DemoSquare: Democratize democracy with AI
Dr. Victor Kristof, Co-founder & CEO of DemoSquare
Wednesday, Feb. 21, 2024 · 10 a.m. · 12m 13s
Few aspects impact businesses more than regulation. However, monitoring and reporting on public affairs is a tedious and costly process for companies and associations. DemoSquare develops state-of-the-art AI algorithms to collect, monitor, analyze, and report on legislative activities, so companies of all sizes can effectively defend their business position. In a transparency effort to reinforce trust in democracy, our platform is publicly available to NGOs, journalists, academics, and citizens. We also collaborate with Le Temps and the NZZ to provide real-time predictions of elections and popular votes in Switzerland.
Embed
Transcriptions
Note: this content has been automatically generated.
00:00:00
okay thank you very much like it for the introduction and spun in concert because it always telling
00:00:05
my um p. g. and i organised for the apply machine learning days at e. p. f. l.
00:00:10
the a. i. n. democracy track and so i'm super happy to be here today on the order of this uh
00:00:16
current events and um to see that it's still relevant to talk about that in the markets
00:00:22
but it is a clear mind equivalency of the most corridor together with michael
00:00:26
ingenue happen and we both have a peach tea machining from e. p. f. l.
00:00:32
um yeah and we all lounging did not correlate is it u. p. f. l. spinoff
00:00:36
we are building and a high platform for um accessing and money during the laws and regulations
00:00:43
i will try to give you a bit of an overview of what's
00:00:46
problem we ought wrestling for companies from your startup right so i'm trying to
00:00:51
just companies and the the problem we also being i'm going to show
00:00:55
you what your doing and the vision that we have on the long term
00:00:58
um and then connect that um to the markets democracies complex i'm not writing you
00:01:06
anything you and this is an illustration of this was and i'm the marketing process
00:01:13
and for new law proposal i'm starting from a
00:01:16
preliminary project i'm drafted by the um um federal conceal
00:01:21
going to consultations committee work parliamentary discussions
00:01:25
parameter evil which potentially refer on them
00:01:28
um and then implementation and enforcement actually if you replace the logo of the of
00:01:33
the book and the federal council chair by the flag of any country in the well
00:01:38
you would probably gets kind of the same um uh
00:01:40
she masking that i'm an illustration of the of the process
00:01:45
and every step of this process if your company or an organisation
00:01:50
provides potential risks but also potential uh opportunities because
00:01:56
when you think about it every business every organisation is
00:01:59
affected by registration turns out that companies allocate substantial
00:02:05
budgets to navigate through regular to be landscape regulatory changes
00:02:11
and and and actually every browning can thing of can
00:02:16
you go cop expo and kitchen about is nicely logo microsoft
00:02:20
all of these companies they have a big effort stings so that's people that are
00:02:25
tracking following the um the political developments to try to make sense of new regulations or being
00:02:32
currently i'm created an end is the time consuming task it's very changes to do
00:02:38
it usually guided by instinct and intuition of these um
00:02:42
of these people you probably get pristine very often under
00:02:46
task is to report to the executive and and management
00:02:50
of the company to brief them about the latest developments
00:02:53
we also often and conduct lobbying and advocacy activities
00:02:58
and they also i'm really from the uh the compliance and legal team in
00:03:02
the company so it's a very very cute in in a in a company
00:03:06
and that's usually operates in a very uh district i get a polygraph or
00:03:11
steamed fail in their activity and this can results in many risks for company
00:03:16
for example last year and global regulators have find more than
00:03:19
ten billions and complain import companies in the in the world
00:03:23
it's not only about finds its leads to compliance
00:03:26
risk to reputation risks to finance risk of course
00:03:30
operation risking you can think of you were activities that have been paused in many regions of the world because
00:03:36
um and it made him do you do you missed some i
00:03:39
record your changes and we're not a adapt the way it works today
00:03:45
you that this people they spend a lot of time abroad and
00:03:48
through complex excel sheets uh long p. of document or hundred pages long
00:03:53
complex government websites and men and make a
00:03:59
no no more tightly or super it's in terms of contents but it they are often very
00:04:03
difficult to to access and to and to to read in terms of user experience solicitation is difficult
00:04:10
to access it's difficult to monitor difficult to and that's what we aim to do at the
00:04:17
most current is to make to the marketing process
00:04:20
more transparent and more efficient to um and advice
00:04:24
for that will building in a high powered legislative intelligence tool and
00:04:29
where the goal is to collect data coming from
00:04:32
many different sources parliamentary affairs news and articles social media
00:04:36
and regular tribute to reaches combined that's into the
00:04:40
platform through layers of artificial intelligence algorithms that we're building
00:04:44
and then to get access to the data into the
00:04:47
output of our algorithms through a modern user interface so if
00:04:54
you connect to do more square um you'll have access
00:04:57
to this kind of a platform or you can search for
00:05:01
um i'm a topic it's interesting to you for example the box trusts
00:05:06
um and this brought through all this with parliamentary affairs at the moment
00:05:11
and then you can have access to details about low proposals for example here
00:05:16
it's um upon thoroughfare submitted by um some you've been now ah
00:05:20
who is and consider it from the cotton for from social is group
00:05:24
about um artificial intelligence and it's expensive it's either trust that we have in switzerland
00:05:30
um you get access about two or what what state is it is in an apartment
00:05:35
i wasn't it is votes when it's going to be discussed in the committee when a plenary discuss in the plenary session we also
00:05:44
building productive um models for example we have a an algorithm for
00:05:49
predicting the likelihood of a regulation to be adopted by the parliaments
00:05:54
um and this gives very important information to public affairs teams because
00:05:58
a day wants to allocate resources the best in the best possible whale
00:06:02
way and it's also a matter of quantifying discover new regulation to impact the
00:06:08
um activities of business is one of the task of these changes also to i'm analysing
00:06:15
reports on on on on the given topic for example here um a. i. security in switzerland
00:06:21
and so we're building in a text editor that's very uh inspired
00:06:25
by the notes taking up notion for those of you who know what
00:06:28
where the idea is that you can um build the documents and pulling in information coming
00:06:33
from external sources for example here each block is h. block here is a parliamentary affair
00:06:40
and you can reference that directly to documents and create a compelling a report
00:06:45
and we also building in in using and large language models to that and these people they can
00:06:52
query directly all the point refers to regulations and get and says i'm
00:06:57
in pull that in directly into the documents our vision for the um coming
00:07:05
here is is to build a cool governments data platform and that we were
00:07:11
collecting process this data in order to improve transparency into the marketing process is
00:07:16
into reinforce interest in public institutions what we want to build is a dashboard
00:07:21
of democracy um that will collect not only data about politics but of all
00:07:26
about little bit out for example laws and our court rulings but also business
00:07:30
related data for example coming from the news coming from company but just it's
00:07:34
and we have started doing that in switzerland and we want
00:07:37
to expand that's to your parents uh you can institutions because
00:07:41
all the regulations that come out of the uh of of
00:07:43
brussels have a huge impact everywhere in and about including switzer
00:07:49
and and then we want to build this uh data
00:07:52
with you where and um um managers could have access to
00:07:56
and the global political polls on given and trends in topics and get alerts when
00:08:02
under the new change and also um gets more problem um predictions about upcoming regulations
00:08:11
the most square and in meeting ah holes for the market dies democracy it's doom or doom also you square
00:08:18
and maybe we have a research i guess it's also um her business strategy
00:08:24
to make part of the platform accessible publicly and for free to
00:08:28
anyone and not only to companies that pay has subscription to the platform
00:08:33
for us it's very important that also journalists and citizens angelos can have
00:08:38
access to at least part of the data and that we are collecting
00:08:42
so actually this is available you can connect and the most courts each now when you have
00:08:47
you'll be able to look through all the affairs and i have
00:08:50
access to um most of the probably do that or we are collecting
00:08:54
you can have access about also parliamentarians themselves for
00:08:58
example here the value in cancer treatment and will
00:09:01
uh with the time to be activities the leaders votes
00:09:04
and their connections to interest groups and and lobbying groups
00:09:09
um and also direct their activity in the press we also
00:09:14
have a nice visualisation for for results and that's not always
00:09:19
accessible on the swiss comment website some of the votes are only accessible to p. d. f.
00:09:24
documents that will scraping all the data from the p. of document and making them accessible through user
00:09:29
friendly interface like this one in this with the
00:09:36
same um ah push in the mornings we uh
00:09:40
built and tracking system unproductive system for the federal
00:09:44
elections and in switzerland in october twenty twenty three
00:09:49
and we had a system i was able to make predictions in real time on
00:09:52
the day of the vote so as soon as the um a ballot offices closed
00:09:56
we uh and the first municipalities published a result we were able to
00:10:00
predicted the um results for the for the party results for that's six major places in switzerland
00:10:07
so here each box um is that what do you do a
00:10:10
dog a line here with the margin of error is our production
00:10:15
and a light align these the an intermediate intermediate recounts meets
00:10:20
the results that were available at a given time so you
00:10:23
see that our predictions they are very stable uh doing a
00:10:26
whole day and the intermediary count uh goes a bit everywhere
00:10:32
and on average we were able to have less than one percent of uh
00:10:36
ever with of various productions with the earliest result that were available and was super
00:10:42
happy because uh we're partners beads to um um big media is with an looks
00:10:48
on on on the and the and the dead in the eyes is german part
00:10:52
who use our predictions and the tracking system on the day of the election and they uh were
00:10:58
using that on the front page that you have a sort of visibility as a very uh very nice
00:11:04
and we're planning to do that again for the popular vote 'em
00:11:07
on march third for now the um um ah initiative autumn initiatives and
00:11:13
referendums and it will be again using our predictions we are in a
00:11:20
playful spinoff um the technologies grandad on our research of jeremy and myself
00:11:26
my research was about mostly um a model in political processes i've developed
00:11:32
the first system for predicting votes in switzerland but also in germany and us
00:11:36
and another system for and predicting d. they just if i men men
00:11:41
so the law uh changes proposed by parliament turns in the u. parliaments
00:11:47
engineers research was about building i'm i'm very domes for information retrieval
00:11:52
specifically applied to the media industry so he was was really
00:11:55
i didn't the section of these two research projects in that
00:11:58
lay the foundations of the of the um um algorithm that prompted
