Player is loading...
Morning panel
Wednesday, Feb. 21, 2024 · 12:15 p.m. · 46m 43s
— Prof. Lonneke van der Plas, Group Leader at Idiap, Computation, Cognition & Language — Sara Ibrahim, Online Editor & Journalist for the public service SWI swissinfo.ch, the international unit of the Swiss Broadcasting Corporation — Dr Kerstin Johansson Baker, Head of CNAI Unit, Swiss Federal Statistical Office — Prof. Andreas Vlachos, University Cambridge — Dr. Victor Kristof, Co-founder & CEO of DemoSquare — Julian Eisenschlos, AI Research @ Google DeepMind — Dr Ashley Thornton, Geneva Graduate Institute
Embed
Transcriptions
Note: this content has been automatically generated.
00:00:00
slide the available and i will ask all of the speakers actually
00:00:03
if they can provide me the slides because yes it's been a very
00:00:06
dense formats short thoughts and a lot of speakers i think have done
00:00:10
a lot of references so you can look them up and and go
00:00:13
much deeper into things okay then move up let's move on to
00:00:17
the the questions and so um do we start with our actually it's
00:00:21
the longest one long as they go so uh we have to refresh
00:00:25
our memories a little bit but the question for you a is and
00:00:29
about this code of ethics that you were talking about some people wondering how
00:00:33
do we actually go about creating such a thing we should be involved how do
00:00:38
you have any clear ideas and how this could actually work yeah um thanks for
00:00:43
the question for short the more people uh are involved the better use as a
00:00:48
usually when we create regulations and codes um for um journalism
00:00:54
we have two more than lie it's really a mediocre anything
00:00:59
small to be go once they have to work to get there to understand
00:01:05
then created twenty five year volumes first thing and uh a second thing for sure
00:01:11
um technical people um so techniques is to uh
00:01:15
have a to be involved as well to understand
00:01:19
our problem is what does it mean uh to work on quantity in journalism
00:01:24
and um program the very high systems to
00:01:28
wear these uh goals um in and one very
00:01:32
important stakeholder is for sure the c. v. society ease the audience that we have it's our target
00:01:38
so they have to be involved i know that there are
00:01:41
some projects already for instance the reporter some uh some point yeah
00:01:45
uh_huh um n. g. o. e.s working to where
00:01:49
it's the says involving see the society in having conversations
00:01:53
on how uh we can i'm me a more transparent or use it all the high
00:01:58
in journalism because this concerns them a lot uh yeah in in the first place so that
00:02:05
all this because there's that are are concerned uh should be involved
00:02:09
so there's also uh an awareness aspect right i think that the the the civil society should stand
00:02:15
up and say we want is because that this is meant this will happen right it won't happen
00:02:20
otherwise i think at the question i had like you saying we have together all the media
00:02:25
institutions do you think that there's a willingness for all of them or do you think that some
00:02:28
will say well actually let me just do my thing i don't really care about this and they will
00:02:34
a step away no i think uh the reason i
00:02:37
we'll uh because we understood the importance of having your rooms
00:02:42
intel sounds in being more transparent um to what's our
00:02:45
readers seen have been more transparent uh about our process is
00:02:50
and team building common values uh i i'm kind of charter
00:02:55
and that can uh strengths and aware aware journalism and
00:02:59
we're systems and instead of we can you need so we
00:03:02
needed we still with the you way id regulation how important it
00:03:06
ease but still these regulations not really interesting journalist and consumers
00:03:11
and so we are we are happy need we are lacking bass
00:03:15
and uh we want to um feel this gap way
00:03:20
and i think that indeed for a i specialise there's
00:03:22
a role for us to play also to advise you on the safety of certain tools and uh what are the
00:03:28
the threats and what i did the risks of certain tools that you might not be wear off and i
00:03:32
think that we can together come up with you know what tools we should use and which ones not for which
00:03:38
uh like girls or for for what they yeah yeah this is a
00:03:42
a very important point because many journalists to have no idea about a i
00:03:46
i'm on science and tech journalists but um any other colleagues they
00:03:50
work on something else the armenian plastic user all
00:03:53
the i but they maybe don't uh identified the challenges
00:03:58
for a guy in journalism so we need uh to collaborate
00:04:01
more with the academia with the mm experts uh for shore
00:04:06
great thank you very much so now move on to julian actually because i have a very
00:04:12
varied that like yeah easy question i think for you what is
00:04:15
the name of the model type using negative and positive examples for
00:04:19
table finding maybe just to you know somebody who's there if we're waiting
00:04:24
to hear that model type was the only member or not maybe not
00:04:28
but that's something you can look up on the slide yes but
00:04:32
so our model recorded the table you will also be the our body
00:04:36
was it with a valuable would be a contrastive learning model because it
00:04:40
was comparing the possibility of examples to show yeah what is it good
00:04:46
table to find and uh i don't but able to find to verify the potential clean
00:04:52
great that so that was only one question don't think you're you're done now i
00:04:57
have another one uh and that is a maybe a difficult one really difficult and
00:05:02
i'm not even sure that it's something that you can answer but since it was
00:05:05
the asking also for big attack and you're you're somebody to closes the that um
00:05:11
we're all missing a lot of different ways in which a ice
00:05:15
trying to help make them about the processes more transparent in your case
00:05:19
uh helping tooth but counting misinformation out of that and both of your talks
00:05:24
uh now the question is how do we reach people with this how do
00:05:28
we make sure that all these great tools are actually use i so so
00:05:31
that you were opening it up for people just but i do people actually
00:05:35
and especially the people that we think we needed most are they going to check
00:05:40
i use your tools and could big pack maybe help us
00:05:44
to um make sure that these tools are actually something that
00:05:47
this kind of offer to people on them daily basis instead of you know you have to go to a specific uh
00:05:53
project works the that side of some new project i try as i
00:05:57
install some code that you can do it you have some ideas about that
00:06:03
this is very much outside of work but yeah i think so little
00:06:06
put a qualified first i think this is only partial utility problem we
00:06:13
the the different tools that expose information i think there was a question most or some goodwill or
00:06:19
you know it's ah yeah have it chose to
00:06:22
try to expose these are also quite a variety claims
00:06:26
ah that the mall might be making like to chapel merry making too
00:06:31
make sure anybody using it can understand that it's possibly
00:06:36
no not faithful to its source information on the one hand
00:06:43
but uh yeah i think we really need to work closely with pro people and we
00:06:48
other experts in this feels to understand how to
00:06:53
yeah which may yeah to reachable better idea from flow point of your models
00:06:58
this is already happening that all the troubles of the phone companies are trying to
00:07:04
is and especially i can see where it's entirely they're really trying to uh put a focus
00:07:09
on retrieving information from reputable sources and trying to
00:07:14
evaluate the extent that we can't but it's not
00:07:18
yeah okay thanks a lot as anybody else maybe from the dallas does anybody have an
00:07:23
idea off i thought about this source as seen the issue of by being able to
00:07:28
reach people i that yeah they exactly the community to it it was
00:07:32
actually question i'm trusting you address what do you think about that is there
00:07:40
um
00:07:41
sure uh okay are you how what what we tour of so i
00:07:45
don't know all problems or sport of what it means what route but the
00:07:50
ah one thing that i like about it with the drop and then
00:07:53
you could not clark should probably work with which i think is actually a
00:07:57
great ritual show ah i wish to watch more movies cause of a
00:08:01
four oh oh what yeah which is actually one of the more with logo
00:08:07
congress should show full of things of the actually the spatial for socially technology and
00:08:12
the the companies and community or not although since we're out with uh would like
00:08:17
to mention user of the third party fox second column about facial close well of
00:08:23
course it was on a machine shop today but we will for a while oh
00:08:27
worse also the the we we shot responding fox signal when there's
00:08:31
huge to fox working one thing so tall one portion of the
00:08:35
world to usually work what do we do to launch all those
00:08:38
buttons on which we were you should be used to internal buffer issue
00:08:43
uh_huh yeah that's right also through funding it exactly there's a lot of the research that we do
00:08:48
is actually funded by big tech maybe some when people might not know this but uh there's actually
00:08:53
fact checking research uh other kinds of research where we really trying to help
00:08:57
the democratic processes are actually of the fund it so that's another way yeah
00:09:01
are you the age of theirs when you are watching that
00:09:06
um uh uh we were truck short while we're sort of the
00:09:10
corporate control contraption bond control for few years uh the actual very
00:09:15
well supported by the big industry you know induction loop racial marshall
00:09:20
well i'll treat tour all contributed what will
00:09:23
we ensure look how you look mole funding organisation
00:09:29
will show a book what we've got really do want to be come regard to single one vermont
00:09:35
small all you know what was what was important
00:09:38
upper welcome it's important to distribute your support to see
00:09:43
usual you don't become we're going to be in of them individually so uh let's all the rubber top job
00:09:48
coloured lots of just it was to get the big that you're supporting a a a research so that would be
00:09:55
really good for party with residual uh at the same time we want it to look reliable big text for
00:10:01
lot clues but we all those shows will forcible will pass the new they have something to add to that right
00:10:09
yes although i did say that i was not talking in the name
00:10:14
of the corporation as a regulator i do believe any they are uh um
00:10:19
welcome this work on going to regulate platforms and the probably the
00:10:24
question of whether that should make data available to researchers et cetera
00:10:30
and then that really as a person or common pleas disclaimer uh
00:10:34
you can now have all the tools you once in a given
00:10:37
searches on menu once if you don't change the business model uh it's not
00:10:42
really gonna change any incentives and wanted
00:10:46
to play sorry okay the business portal of
00:10:52
increasing age meant uh why yeah uh yeah
00:10:56
information what information gets shown shaded et cetera
00:11:00
i mean uh yeah but we want to actually that anything to add to that actually
00:11:07
i don't know really adding anything new to that comment that um i
00:11:13
it is problematic the way that the business models are designed
00:11:16
around user engagement um and you know getting people to the
00:11:23
getting people to come come p. hamas to have the most emotive pose so that they get more like they get more views
00:11:29
um and of course the social media companies get more revenue every time
00:11:34
people will when people are on their platforms any stain gauge the seymour at um and
00:11:41
i mean i guess it's not the nature of these i mean that's how they're
00:11:44
making their money um and silence things yeah i wonder how how i i i'm yeah
00:11:53
i have no idea how they would change um their business model but i do think in some ways that
00:11:59
if there are laws or regulations that are put in place we
00:12:02
should make it in their best interest in terms of legal liability to
00:12:09
at least understand you know how their platforms are being
00:12:13
used and some of the negative consequences of the waiters that
00:12:17
forms are being is and h. did he ultimately held responsible
00:12:20
for mitigating on those issues i don't know if anybody read
00:12:24
the recent articles about the the air canada chap by that gave a refund
00:12:31
um when it wasn't supposed to get every five air canada i'm really sorry that
00:12:36
um we can't get a refund because that's against our policy and eventually i judge said well wait a minute
00:12:42
your chapter are which is supposed to be in a part of your customer service unit now uh_huh gave the refund
00:12:49
so the refinance stance you you have to you know refund the customer is so um
00:12:55
you know i think maybe that's the first step right like they're they're chap that kind of a role
00:13:01
and air canada pay the price yeah and yeah but sunday
00:13:05
will be uh some courses for people to fall on how to
00:13:08
get ten bucks to tell them together we fight like body
00:13:12
shake it a yeah exactly and and the argument that correct information
00:13:17
our website so if you had on your research in compared with the chat battle idea what's on the website you would know
00:13:24
yeah that was an entirely accurate and the judge said you know that's just a bunch of fully like
00:13:30
you can't have the chat by saying one thing and then your written
00:13:33
materials for customer service on on line saying the other because what's the point
00:13:38
you could have just gone to the website and gotten the correct information in the first instance
00:13:43
yeah but instead you interactive chat i instead of a human gave
00:13:46
you the wrong information and it's like yeah it's kind of um
00:13:53
mind numbing yeah and um one thing i remember from your talking also saying that
00:14:00
i am i thinking of a way to address this problem with the business model
00:14:04
education right i mean happy people to understand what the business models are behind
00:14:09
ah could there be maybe think that that's that's probably our our last resource but
00:14:14
also maybe a very good with or stiff or issues that are very hard to
00:14:19
the software that i mean what i yeah i think depressing point about it is you leave you know
00:14:25
technology has been evolving for quite some time we've known that there has been
00:14:29
election interference in these kinds of issues since twenty sixteen yeah with doctrine so
00:14:35
why hasn't anything been done already now everybody's panicking because we're coming up to the
00:14:41
you know twenty twenty four is a year with many elections in many countries um but
00:14:46
we could have tried to address some of these issues in a meaningful way earlier and
00:14:52
all that said i don't think education is a is a wasted effort in fact quite the contrary i think it's
00:14:58
one of the best things that can be done to help people understand
00:15:03
more concretely the the technologies that they're interacting with ah how the technologies
00:15:11
could be a in a and inadvertently playing to human crises and i mean there are lots of things
00:15:17
in life that end up doing that so people should be just we're that that's kind of the side effective
00:15:22
of some of the social media platforms um and i've states really you know
00:15:31
governments can situations they really need to take this seriously i know in the us there on me
00:15:36
i think it's kansas is the only state that has and me and it worries social media literacy course for kids
00:15:41
and and you know there are some other states that have
00:15:45
media literacy courses now in various forms which is a good thing
00:15:50
right and it needs to be much more widespread uh_huh so that people
00:15:56
actually have a chance to make a decision for themselves like okay you know this is how
00:16:03
these platforms are interacting with me and that might influence how i interact with them uh_huh without the knowledge
00:16:10
knowledge is power they have you know they're just kind of bystanders in this whole thing that's happening
00:16:16
yeah so regarding education we don't really have an educational specialist this time in the panel but that's
00:16:22
yeah i i guess media has something to say about it maybe also you'll have an oval
00:16:28
there you see will you can you say something about education and government on the wall or
00:16:34
sun mon not education for kids that usually cancer now on top of that
00:16:39
so they had a it but yes within government we which share in u. k.
00:16:45
its capacity building call it what you want but i think that needs to
00:16:49
happen as well because i think that's very important with human factor automation bias and
00:16:56
maybe it's not for opinion making but people using new tools and
00:17:01
it since it's straight vacancy mutual and they have to be great
00:17:05
because they were so expensive and then modern and data like to
00:17:08
steak and a oh there's no way i'm gonna question this i think
00:17:13
it needs to be you train in the education train equally we want 'em up scaling uh_huh
00:17:20
condition different names but uh yes i think
00:17:23
it's absolutely wheeler updates network is there any project
00:17:27
that addresses this are you know and it was definitely planned as one of the main pillows uh
00:17:32
uh it's like in it and there is there is there are uh is
00:17:36
is more with the data sign strategy if the configuration uh_huh and and yes
00:17:41
sits on going it's there it's kind okay good great i have another question
00:17:46
for you about your methods of collaboration i was wondering you do most of your
00:17:52
all day i uh stuff that you do in the network is it all in
00:17:55
house with a an in house he i specially so do you collaborate with uh
00:17:59
in situations outside and and if you collaborate what is the method for collaboration
00:18:07
so just to disclaimer this see the project is stuff that the c. n.
00:18:12
a. m. and maintains is you know
00:18:14
voluntary submission by people who are are uh
00:18:20
we'll have projects on the way your so i have absolutely
00:18:24
no control of the project so now i hear it uh in
00:18:27
case i have also if i have any i project i could submit it and it would be less that is that oh
00:18:33
for the moment we get concentrate on a federal employee but still yeah yes this is a great it
00:18:39
great plans to to go beyond but you know for the moment suspect we collaborate we signed up
00:18:44
a known status in it where sit to plug into academic and
00:18:48
that it's that that's the that's more the way we do private
00:18:51
yeah but so so oh i am in contact with pride project leads but uh mm
00:18:58
and yes we do know that does a lot of interaction with
00:19:01
their academic institutions external partners sometimes it's completely in house sometimes it's
00:19:07
a hybrid would probably have less projects are completely boating because that's hasn't been that my
00:19:13
primary focus of this list it's been really more to get a synergies in people's it
00:19:19
uh expertise so the mom quite a few
00:19:23
of them are a largely in house but
00:19:28
it's kind of a bias of at least one thing okay great ah yes yes there yet and yeah yeah yeah
00:19:34
i wanted to add something about medication because for sure we
00:19:37
me yeah we have um a big role to play there
00:19:41
in terms of uh shaping um the critical mindset uh of the society
00:19:46
um and that's why we have a big responsibility in how we frame
00:19:52
uh uh where um articles in which kind of articles we push a
00:19:57
true and i think one um important point and one that is also scary
00:20:03
is that we are more more aligning ourselves to the
00:20:07
model in core etched by u. d. yeah i cease themselves
00:20:11
pushing more associates in such an at least a articles and contents
00:20:16
um and fabricating vancouver for instance balance and you want article
00:20:22
and um for sure disease i way to increase reach
00:20:26
and to have a people are more around on our platform
00:20:29
for the creek abating that uh it's um b. d. e.
00:20:33
m. we are dependent uh on e. uh in some ways
00:20:37
but i think it is also shortsighted view because people i hope
00:20:42
uh that are getting a already frustrated and we'll get frustrated with
00:20:47
seeing the word just in black and white in this way e.
00:20:51
so we really have to uh find down under a more creative way
00:20:55
to to reach a dear heart in their minds in the future
00:21:00
but i think that what i had a question also when you victor
00:21:04
uh and so you were talking about uh the tools that
00:21:07
you developed for looking at these democratic processes and making it more
00:21:11
understandable and transparent for people uh you're talking about federal processes and then
00:21:17
you is talking about going to new levels and there's a question that's that's
00:21:21
how what kind to know that must be very complicated for people to
00:21:24
follow could you do are you planning to do something welcome to squeeze federalism
00:21:30
laurel twenty six different ways of accessing cantaloupe parliamentary did
00:21:34
hans differently and i were room road map at some point
00:21:38
um maybe after we've have reasonable first uh funding round because it's a real time consuming
00:21:45
um and i actually need to unit yeah it's i mean i don't think i'm
00:21:49
many companies they don't provide undercut the p. r. to have very easy access to data
00:21:55
yeah so we might start with a lot bigger concerns at some point um
00:21:59
but many um companies and organisations we've been talking
00:22:03
to one quarter using on the platform mention that
00:22:07
um you did our actually very important to
00:22:10
them because what's happening in brussels completion a lot
00:22:14
what's happening in britain afterwards yeah suits also agreed we for them to be able to anticipate
00:22:20
nope no in an extra three five years ah sort some so
00:22:25
hence the ah yeah you did recollection yeah and then into
00:22:29
a thinking also of business models that we talked about before
00:22:32
you were explaining to us that part of your business model is also that you want to do you know open up the
00:22:37
tools and uh i think that's also an interesting aspect so how does that relate to discuss if it had been added server
00:22:44
good question so for us it's really core two distributions are
00:22:48
we have to pulp and paper important to do tonight we have
00:22:52
it's always i'm i'm hard to just decide where to put the limit what we provide
00:22:58
free what to keep behind to do the digits start or what is that you know
00:23:06
not not not back i think what is you don't know what will provide for free in what do we uh keep behind a lot in in what
00:23:13
and so what we sell to to to companies were using our platform um
00:23:18
but yeah it's really also way for us to differentiate also from the competition
00:23:24
um and we also um and partnering with me design we really strongly believe
00:23:28
in in in powerful synergies with the media and so does also something we
00:23:32
look to push for hum um so yeah i'm i'm also i mean we
00:23:37
come from in or academic background so what you've got researches in the room for
00:23:42
interested in some of of the data that we have we're happy to open a discussion i think we're good doing are very
00:23:49
t. just an important job of cleaning collecting are
00:23:53
getting in normandy normalising disneyland so you're happy to have
00:23:56
a discussion with our anyone who's interested in that sounds like a a nice invitation a lot of people he
00:24:03
um yeah i still have a question for uh andreas and that is about um how the sources art
00:24:09
shows that like you know and the sources for finding evidence i maybe actually it's also what union actually back
00:24:14
how do you decide what an isn't that uh by it and our
00:24:19
people happy with that you know i actually when i was talking about
00:24:23
a organising this event i have met some people on the train and they were telling me uh you know but fact checking the they
00:24:28
can also be on the the wrong people and we we trust them
00:24:32
you know i thought that yeah what what is the work to school
00:24:37
for the first part of what little cabal who says we could be you
00:24:42
know we could be you know we get i you you serve which encyclopedias
00:24:47
also but one tool mobile just gets were written by people in different languages
00:24:53
and it's coming to put to use showing some so so soul because mistakes
00:24:58
resourceful shoulder shockley people think quietly to also could be used are paid
00:25:02
for crude but companies oh what it's not locked on a whim sticks
00:25:08
it even contradictions within sort of boswell you uh when we do the
00:25:13
digits on top of lot and will be the loosens general would talk about
00:25:17
cringing only supported we should it by you have bins uh that's because we don't want
00:25:23
to make its output on the evidence itself even though we could we could be use
00:25:27
actually a good starting point o. show last august the actually i i you know i
00:25:32
think uh we could be your username actually more when you think about shapes and sizes we
00:25:37
the cost of creating would you pay you use quite minimal
00:25:40
compared to any kind of um social between some chooses the
00:25:45
more successful law school in conversational author which i think we can better probably should how can
00:25:50
improve other uh form you know for certain parties because we could be just give not on off
00:25:56
uh we actually but when people uh a. p. i. for the short
00:26:00
of it is but that's not but to be more precise would've died
00:26:05
only the fox circular suppose would serve there's you just uh than we actually
00:26:11
how are still do to fall into the web of your phone
00:26:14
usually forgot um uh that's probably a subway a two wheel too
00:26:22
uh_huh yep yeah again i have specifically work on that aspect but
00:26:29
yeah this search algorithm dot google uses to showcase their souls
00:26:34
has some rhyme kenyan border tries to look at
00:26:38
different things like you know some quite the indexes and
00:26:42
different uh um network of claimants and typically yeah it kind of
00:26:49
what downstream that's gonna try to leverage that and use
00:26:53
this also longer for stoppages the i think that is
00:26:58
also personal opinion the longterm vision should be that a
00:27:04
system should be able to aggregate the and maybe even organise
00:27:09
the information and referral sources uh and i would be lately a person
00:27:15
should be able to customise okay trust discovers a uh maybe a um
00:27:19
i i camera bothersome from here not so much about it was coming from this other please
00:27:24
a separate that because you know they get noted
00:27:27
for only know specifically the logic problem really one
00:27:34
okay great i think yeah with that have uh about fifteen minutes so uh uh one thing i'd like to do now
00:27:40
is that we've talked about it was several talks that focused
00:27:43
on threats the others that focused on solutions and on opportunities
00:27:47
so i'd like to ask the panel to kind of choose their
00:27:51
you know where they are leaning against like more thinking of a i
00:27:55
has actually more opportunities for a new remote with the or has more
00:27:59
threats i'm i'm very curious to see for something like i guess i
00:28:02
know where they were big so let's say that let's change the uh the like the uh that's at all the people that think they're more
00:28:09
than threats on that side and the ones that think that there are more opportunities on this side and then let's see how we can uh
00:28:17
convince each other so where do you wanna go to the flat side or to the war in the middle
00:28:24
it's will lose yeah i would love to see what's what's true
00:28:34
it is very hard but for that uh let's start maybe with um actually
00:28:37
i think you talk with him main yeah i i think both there and
00:28:41
actually i think you both at quite a lot of emphasis on the threats
00:28:45
and i was just wondering whether after using the other talks that was there any
00:28:49
aspect in those thoughts when you thought like okay maybe that could actually help me with those that's it i've seen too
00:28:55
help us move together uh you know and and make maybe the the best
00:29:00
of a i you know in a way that we collaborate if you see something
00:29:04
ah the hope but i certainly think that there's hope happens because as you know we're all sitting here
00:29:12
together right and having these interdisciplinary conversations about the the
00:29:17
risks and and also about some of the benefits and
00:29:20
you just mention right now potentially user customisation a customisation in terms of the search search results you see but
00:29:26
what happens when someone decides that you know they
00:29:29
only want search results from fox news and right are
00:29:34
they're customising their you know search results to be exactly what it is
00:29:38
that they want to see um which is you know a very unfortunate
00:29:43
human cognitive bias confirmation bias um and i so i don't know
00:29:50
what the answer is but i think that it's really really important
00:29:53
that we have these these discussions to try to see um
00:29:59
is there a way forward where we can you know leverage the the really
00:30:06
powerful abilities of this technology and the benefits that i
00:30:11
can bring to society uh_huh um in terms of dissent
00:30:15
knowledge ah in terms of you know things like using a i'd help with
00:30:20
cancer diagnoses and you know the other
00:30:23
medical and sneers and and also you know
00:30:29
perhaps we can brainstorm togethers and some ways that you know okay if we do
00:30:33
implements user custom customisation features i know that you know open a i applaud post
00:30:40
last year had suggested that they wanted to have a uh like a values toggle on their check g.
00:30:46
p. t. output so that you could you know within reason somehow um say well i want you know
00:30:54
i want my results to be more like this i want my results to be more like that i mean
00:30:58
as a social scientist to me that is an absolute
00:31:00
terrible idea um because again you you're just i guess enabling
00:31:08
cognitive biases just continue reproducing reproducing reproducing um and uh you
00:31:13
know this is a natural human things it's not it's not
00:31:16
appealing um but in that sense i think it still comes
00:31:21
down to education you know it is really helpful to now
00:31:25
yeah and that's how the the human mind works on that stuff is always gonna be
00:31:32
stronger is gonna weigh more heavily in your mind and the good stuff and that's you know and evolution early
00:31:37
um important adaptation because it was much more important for you to spot the sabre tooth tiger there
00:31:44
no your grandmother or something um so uh once you are
00:31:49
aware these kinds of things i think that that yeah can also
00:31:53
hell so you know the question is what's the best way
00:31:58
to have these kinds of interdisciplinary um conversations and discussions about
00:32:05
the way that the technology is going to evolve what are some of the new
00:32:10
tools and features that might come down the pike
00:32:14
and and what are the possible benefits but also the
00:32:17
negative consequences of implying these kinds of for you know
00:32:23
i'm implementing kinds of features uh_huh so i guess there
00:32:28
it would be i mean they're the best intention sometimes it seems and also sometimes i
00:32:33
have to fit ice ice what's the movie the social them uh and i also saw that
00:32:37
actually often it's very hard tons for us as a i specialise to see what will
00:32:43
be the consequences of the tools that we develop whereas you might may have a better
00:32:47
feel for that because you have a a different background and you could help us maybe
00:32:51
see that and uh also for these things that we sometimes implemented i think
00:32:55
we're going to see the sole something are actually not selling yeah of course
00:33:00
you can yeah yeah i mean there are always going to be bad actors out there also we're going to get vantage of these um
00:33:06
loopholes and you know opportunities to do do really bad
00:33:09
thing so also from a national security perspective i think that
00:33:13
um that's that's important thing as well because you you don't want
00:33:18
tons of by stay i generated material being just like
00:33:23
palm into social media and so that people are
00:33:26
being constantly bombarded with half actively charged incorrect political information
00:33:31
and you know they won't be able to parse what's true from its files and you kind of
00:33:36
lose touch with with reality you know if you get that biting rubble call saying you shouldn't about
00:33:41
um it i was into it sounded exactly like that and so
00:33:46
you have to have a i suppose on minimum level of um
00:33:51
is education word i don't know i'm i'm and and it's and and minimal awareness of how
00:33:57
you know your your rights as a voter to know that that was you know who that is
00:34:04
wells yeah we just wanted to ask you some for a cruise or your opinion because
00:34:10
i understand the that off bottles and being
00:34:15
isolated remote opinion but i'm curious uh if you
00:34:19
think that would be different from how it used to work where a person would watch one
00:34:23
couple t. v.'s really got what newspapers my conjecture maybe is that there used to be other
00:34:29
forms of community the lord's prayer and the more that would expose you to make that maybe you're
00:34:35
so show circles in real life but it sure uh_huh so
00:34:41
you know absolutely things used to be different right and um i think the problem today
00:34:49
has to do with the this p. this p. and the just
00:34:55
sheer mass of information that's coming at people from all sides um and
00:35:03
you know you i watch people on on public transport we're just going through
00:35:07
the tutor uh the to talk radio's going through on her going through um
00:35:12
if that's the way that you're getting their information and unfortunately a lot of
00:35:17
younger people and you know a certain age group um that is the way that they're that they're getting their
00:35:23
information i'm not getting any more from here times um
00:35:27
for the washington post or you know other major international newspapers
00:35:31
this is the this is the quick way of doing it it's really
00:35:35
engaging it's actually kind of fine um and you know the problem is there
00:35:42
there's no there there right it's it's really superficial the
00:35:46
information that you're getting in like a small sound by
00:35:49
and that's because you know you can remember the sound bite right and if there was something funnier
00:35:54
unloaded about the video you're gonna remember it even more than that the problem is it may not be true
00:36:01
sandy has something to say they're about you know bringing
00:36:04
information and a few changes strategies that has a file
00:36:08
for for this like to dress like kind of meets more young people that actually preferred this kind of media
00:36:16
for sure we have to be present there so
00:36:19
we decided to open even attic talk account and
00:36:23
to try that ah then is going very well actually even
00:36:26
though our target audience are not mainly young people but i think
00:36:30
it should be in the future because the young uh
00:36:33
a generation is our where next uh readers um so
00:36:39
we have to be present there we we have to try at least to to compete with that system
00:36:45
but um i think it's important also to go
00:36:48
out of of each to to propose something different because
00:36:53
and as i said um the purpose of this platforms is just you
00:36:57
engage you so much that you will stay there as much as possible
00:37:02
and uh the way they are doing that without
00:37:04
a return send and trying to reading your preferences
00:37:08
uh it's very subside and that's the difference with the real world i think because interviewer where there is
00:37:14
much more about are or were uh um our one at this you shouldn't
00:37:19
in place and why when you we opened the internet we don't even recognise
00:37:24
how it is influencing our own decisions even
00:37:28
net and net flicks a platform for instance
00:37:32
but um regarding your first question if i you want to see more clearly
00:37:37
scored the opportunity of of course as i you i highlighted i'm very worried
00:37:42
but i want to stay in the opportunity assigned because i think it's important
00:37:47
to highlight solutions in journalism different to
00:37:50
serve presents a research uh showed that people
00:37:54
our tirade to to hear just about problems for sure we have to put them in front of
00:38:00
the recent that's very importance that but the next step is to find solution that we can propose
00:38:06
and we have to move to work to what's that uh also because
00:38:10
not just the bad guys are developing a i also good guys uh like
00:38:14
you were like professor block was that i hope there will be yeah system
00:38:18
lie okay your is uh in place a one day for us journalists you
00:38:23
hello eyes a spot a misinformation fake uh news and also for the the
00:38:29
people for the audience in even for the social media platforms where struggling now uh_huh
00:38:35
okay yeah i guess that and oh okay that's true too though since
00:38:40
ritual those systems of the will probable so uh we rely also deposed
00:38:44
what work could be either select group of us will also let us
00:38:50
actually i think uh the the g. for me is that when the people
00:38:55
jerusalem to show it to people do to foster
00:38:59
credentials been yeah that's true oh that's actually what
00:39:04
i was thinking wouldn't talk about would be given also in a very personal job will not screw
00:39:09
successful were so usually live web use up with what we we don't want to subject
00:39:14
to view a sock on little should to buy would be you or some other such
00:39:19
might possible but actually show clubs point that formal so well that's what we want to
00:39:24
the social but but open that uh the ludicrous while so but but it's just too much
00:39:32
that is a very interesting uh oppose and i was also thinking of
00:39:35
your link to the multi link war because i think that i mean
00:39:40
the way that uh i anyway but the p. s. which in english about i don't know chinese affairs it's very
00:39:45
different as when you read it in chinese right so there's a lot of fat i think it's very interesting to see
00:39:51
from different countries reporting about the same events in a different way and the
00:39:55
same for the media articles i guess you could have a much more varied
00:39:59
way of looking at things and people could inform themselves i i really like this idea um
00:40:04
yeah so so why don't you go uh oh oh shit so we could because engine because
00:40:09
it's multilingual digital you pop the purpose but it was it was very interesting is that you oh
00:40:15
the more are is our particular topical we keep it you're the user who sure which takes
00:40:21
me get are getting picks foster jewel the so you know that it's easiest to uh the
00:40:28
just produce information why would you peed your control resource works you
00:40:32
putting things in the lower the option of spokane pages are the species
00:40:37
so we're actually usually low of pursuing show you some just was wired powering the the jewish actually sure what you could do could be
00:40:45
you you could but if you like limbaugh there's there's
00:40:48
holes problem so nobody the problem solution racial or worse
00:40:52
well you how sure people engaged yeah sure want to to show
00:40:56
a small and uh what else but we don't like to also oh
00:41:00
it's not even last long watches that will be on
00:41:02
the represented actually are bought all show even the call change
00:41:08
sure what though uh you know you what oscar either something struggled to
00:41:11
useful stop would work with a lot you get working for the uh also
00:41:16
so if you were to put those rubber isn't big capers league was
00:41:20
bought language was mostly mostly true you're straight up you get some was appalling
00:41:25
oh are welcome males shuttle launch was what a little bizarre what michael should look
00:41:30
was usual for up control only those
00:41:34
are good control mall shopping unforeseen where's thinker
00:41:37
we'll we'll because the person or world and the locking
00:41:41
will will be lots of former your records or not
00:41:45
great yeah guess then maybe to the close we can give it to guess since you had something mm thank you
00:41:52
yeah i think uh a big at the end of the day you brace you know which you ditties it's very much it
00:41:58
sorry yeah but jess take it's solvable uh it's not good or
00:42:03
bad per se what's probably quite different with the eyes that the impact
00:42:08
are of a an exponential nature and that's very different we human princeton
00:42:13
deal with exponential curve very well and it works at several levels and
00:42:19
on the one hand yes some if it's can be absolutely deleterious and we have twos talk about then and on the other
00:42:25
hand it's not necessarily that bad you know that doomsday scenario which
00:42:29
actually leads to regulatory capture right it's how bad we need to
00:42:33
you did not do nothing regulated completely whereas it's it's it what
00:42:38
most of them are tools if they have messy defects but it's
00:42:42
probably quite context dependent today that did the topic is that context dependent so i think
00:42:48
uh maybe once we zoom out a bit of the hype and uh you
00:42:53
getting grounded because that's gonna be annexed to there's gonna be a next take
00:42:57
because of computing is gonna do cooped up if you or something uh i think
00:43:03
and let's not forget yeah we live in a more consistent with these things
00:43:07
should be subject to the bite the what impact one these tools to have
00:43:11
i'm not yeah in a lot of things can be dealt with by tactic solutions but everything
00:43:19
okay but if i think we ah you have lots of course that it that i'm some someone mentioned it
00:43:27
to him something about related to the consequences of the tools that we're building an and just wanted to see
00:43:33
something we have this productive system for parliamentary votes and and you know it it
00:43:43
has very high accuracy right little to put the the votes at the parliamentarian level
00:43:49
uh_huh uh_huh and the first time we've uh made a prototype for that um so
00:43:55
it's in front of our eyes to be how nice it was a bit frightening
00:43:59
because what does that say you on the state of democracy for able to predict the rules we search high and um
00:44:06
and accuracy and yeah i think it's uh we we i mean something ongoing discussion that we
00:44:11
have internally with my co founder how about the consequences of having such a system in place and
00:44:17
to whom i'm i'm i'm removing my starter founder hats
00:44:23
what are the consequences of something that's to to
00:44:25
um to private companies and to to lobbyists because um
00:44:30
yeah and uh many of 'em were uses door
00:44:33
knob you spend your trying to influence the political process
00:44:38
some young that's um but it's a very interesting discussion but we haven't we haven't
00:44:42
found a solution yet so but still i think it's important to ask this kind of questions
00:44:47
yeah we've had a discussion with an ethics um experts on
00:44:51
pure fell and just set that's as long as any um
00:44:58
stakeholder in your condition can have access to the same tool to simply
00:45:03
ground can be like a big multinational company or can be small and you
00:45:07
then you remove this responsibility from from your shoulders right and
00:45:12
everyone can play with the simple did whatever they want to use it however they want
00:45:17
so that's uh that's where we are at the moment but still that's an ongoing discussion just wanted
00:45:21
to say yeah but that that's very interesting i can just say see how difficult it is but also
00:45:27
in order to be able to fund your being opening
00:45:30
opening up this democratic processes for people to to see
00:45:34
for free i mean you have to also get some revenue from someone else right and uh yeah
00:45:39
yeah yeah exactly exactly and but but to be to be around us so we have like a special
00:45:44
discount for and you're was for example when i'm on our current customers we have companies and we have also
00:45:50
but listening and you wouldn't know the difference the interest of handicapped people rooms with them
00:45:55
we of course will discount because we don't have some revenue um but also last thought of company i used
00:46:04
them on from yeah yeah i see what it's nice that you make this difference for uh for the different uh
00:46:09
types of clients quote so i think that'll close the panel here
00:46:14
and we go for a nine inch uh there was some questions also
00:46:17
about multimodal a fact checking so talk about the face and i would
00:46:22
like to say please go to the post of battle it this is
00:46:25
we can stop it uh so uh and go to all of the posters the people here also to
00:46:30
present to you so uh i hope you have a chance uh uh doing much to to visit them
