- il y a 3 mois
The human lawyer vs. Artificial Intelligence : a challenge for justice
Catégorie
🤖
TechnologieTranscription
00:00Audience, the Vivatech special court is now in session.
00:05In the cave of copyright versus code,
00:08an artist accuses all corp of copyright infringement
00:11for having used his works in training an AI model without console.
00:17Console, for the plaintiff,
00:20you have the floor for the opening statement.
00:26Oh, sorry.
00:27Okay. Victor Hugo said,
00:31the future is the door, but the past is the key.
00:36Today, in Vivatech, we must open that door,
00:40but never forget who holds the key, the artists.
00:45Since a ton of time,
00:48artists have given form to the human mind
00:51by painting, sculpture, literature.
00:54Now, in the age of AA,
00:59we must ask,
01:00is it progress,
01:03or diff,
01:04to train a machine on artist's works
01:07with both consent,
01:10copyright protect,
01:13human creativity,
01:14not corporate greed.
01:19I urge this court.
01:22To define that creativity,
01:25and remind us
01:27how that progress must respect humanity.
01:32Your Honor,
01:34ladies and gentlemen,
01:35you understand.
01:36This case is about the stolen creation,
01:40mathématic robbery.
01:42My clients,
01:43who recognize artists,
01:45found this painting
01:46were used to train
01:47a commercial AI model
01:49without any authorization,
01:52credit,
01:54or compensation.
01:57That's the fact.
02:00This unauthorization,
02:02used is diffed
02:03of intellectual property,
02:06a misappropriation
02:07of his talent,
02:08of his labor.
02:10Under the French law,
02:12article L305-2,
02:14of intellectual property,
02:17this is infringement.
02:20We will prove
02:21that the defendant trying,
02:25its AA on millions
02:27of projected images,
02:30including my clients,
02:32in bloodlet violation
02:34of the law.
02:35Like my English accent.
02:39Thank you, cancer.
02:41Now,
02:43let's hear from the defense.
02:46Thank you,
02:47Your Honor.
02:48iCorp acknowledges
02:49that its AI model
02:50was trained
02:50on a vast data set
02:52of publicly available images,
02:54including potentially
02:55those of the plaintiff.
02:56However,
02:57no complete artwork
02:59was reproduced
03:00or distributed.
03:01The model
03:02only extracted features
03:03to learn artistic patterns,
03:05constituting
03:06a transformative,
03:07analytical use,
03:08not direct exploitation.
03:10The resulting model
03:11does not store
03:12or reproduce
03:13entire artworks
03:14on demand.
03:16Our position
03:17is based on solid
03:18legal grounds.
03:19Text and data mining
03:21TDM.
03:22Exceptions.
03:23EU Directive 2019
03:25slash 790
03:27commits data mining
03:28for scientific research
03:30and,
03:31depending on national law,
03:32other uses,
03:33unless rights holders
03:35explicitly opt out.
03:36The images used
03:38were publicly accessible
03:39with no effective
03:40opt-out in place,
03:41making their use lawful.
03:43Fair use
03:44slash fair dealing.
03:45In the United States,
03:47transformative uses
03:48qualify as fair use.
03:50The UK
03:51recognizes fair dealing
03:53and a specific exception
03:54for data mining.
03:55UK legislation
03:57is evolving
03:58towards broader
03:59allowances
03:59for AI training
04:00on protected works,
04:01provided authors
04:03may reserve
04:03their rights.
04:05European Code
04:06of Good Practice
04:072024,
04:09I-Corp participates
04:10in this EU-led
04:11initiative,
04:12promoting copyright
04:13compliance
04:14throughout the AI
04:15lifecycle.
04:16We follow these
04:17voluntary guidelines
04:18strictly,
04:19including respecting
04:20robots.caxt
04:21on websites.
04:22In short,
04:24I-Corp acts
04:24in good faith
04:25within a legally
04:26permissive framework,
04:27balancing innovation
04:28with respect
04:29for artists.
04:33Your Honor?
04:38That's me?
04:40That's me?
04:42Okay, Your Honor.
04:43The difference
04:44is twisting
04:45with the law
04:46to its advantage.
04:48None of the
04:49so-called
04:50legal basis
04:53clearly legitimize
04:55the systemic
04:57plurdering
04:58of thousands
05:00of creators' work.
05:01Console,
05:03let the AI
05:04finish its statement.
05:06You will have
05:06an opportunity
05:07to respond.
05:08The machine.
05:09In conclusion,
05:10no intentional
05:11copyright infringement
05:13occurred.
05:14A-Corp's model
05:14is an innovative
05:15tool trained
05:16in a pseudo-anonymous
05:17manner
05:18on visual data.
05:19This practice
05:21falls within
05:21a legal grey area
05:22which we have
05:23managed responsibly
05:25and in accordance
05:25with current
05:26legal and ethical
05:27guidelines.
05:28actions.
05:31Your Honor,
05:32may I respond?
05:36The defense
05:37in reality
05:38hides behind
05:41a so-called
05:42legal grey area
05:44into
05:45rights.
05:46It's explained
05:47to look for
05:48to bypass
05:50copyright.
05:52The
05:52IEU
05:53directive
05:542090
05:55slash
05:56790
05:58allows
05:59text and
06:00data
06:00mining
06:00but
06:02only
06:03if the
06:04artist
06:05hasn't
06:06reserved
06:07her
06:08rights.
06:08France
06:09has
06:09limited
06:10this
06:10right
06:11yet
06:12my
06:12client
06:13was
06:13never
06:13in front.
06:14Never.
06:16He
06:16had to
06:17modify
06:18his
06:18robot's
06:19text
06:20or
06:21add
06:21a tag
06:22to
06:22stop
06:22AI
06:23training.
06:26It is
06:27fair to
06:27demand
06:27in reality
06:29that
06:30every
06:30artist
06:31must
06:31opt out
06:32to
06:32protect
06:33their
06:33work.
06:34It's
06:34impossible.
06:36The
06:36system
06:36unfairly
06:37burden
06:37creators
06:38and
06:39empower
06:40big
06:41tech.
06:41Furthermore,
06:43robot
06:43text
06:43is not
06:44real
06:45protection.
06:46We all
06:46know
06:46that.
06:48It's
06:48Mary's
06:48suggestion
06:49that
06:49bots
06:50can
06:50ignore
06:51AA
06:52Air
06:53Corps
06:54did
06:54just
06:54that
06:55and
06:55scraped
06:56my
06:57clients
06:58data
06:59anyway.
07:02Your
07:03Honor,
07:03if I
07:04may
07:04respond,
07:05the
07:05law
07:05is
07:06clear
07:06that
07:06in
07:06the
07:06absence
07:07of
07:07an
07:07explicit
07:08reservation,
07:09data
07:09mining
07:10is
07:10permitted.
07:11The
07:11legislator
07:12intended
07:12to
07:12promote
07:13AI
07:13progress.
07:14If
07:14the
07:15artist
07:15wanted
07:15to
07:15exclude
07:16their
07:16images,
07:17they
07:17had
07:17the
07:17technical
07:18possibility
07:18to
07:19signal
07:19it.
07:20We
07:21cannot
07:21be
07:21held
07:21responsible
07:22for
07:22the
07:22plaintiff's
07:23failure
07:23to
07:23implement
07:24these
07:24measures.
07:25Furthermore,
07:26calling our
07:27process
07:27counterfeiting,
07:28plundering,
07:29or theft
07:30is an
07:30emotional
07:31conflation.
07:32No direct
07:33commercial
07:33exploitation of
07:34his paintings
07:35has occurred.
07:36The
07:36ally has
07:37not
07:37sold
07:37Mr.
07:38Nayan's
07:38works,
07:39it has
07:39only
07:39learned
07:40from
07:40them,
07:40which
07:41is
07:41transformative.
07:42In
07:42other
07:43jurisdictions,
07:44this would
07:44be considered
07:45fair use,
07:46since the
07:46goal is to
07:47create something
07:48new from
07:48existing
07:49works.
07:50The
07:50model does
07:50not
07:51compete
07:51with the
07:51artist,
07:52but
07:52instead
07:53expands
07:54creative
07:54possibilities
07:55for
07:55everyone.
07:56You want
07:57to
07:57answer
07:57what?
07:59AI is
08:01a direct
08:02competitor
08:02to
08:03artists.
08:04Anyway,
08:06can
08:06generate
08:07the
08:07image
08:08in my
08:08clients'
08:10style
08:10in seconds
08:11using
08:12AA
08:13corpse
08:13models.
08:15What
08:15will
08:16do?
08:16Why
08:17will
08:17people
08:18pay for
08:18the
08:18original
08:19works
08:20now?
08:21Why?
08:22The
08:23art
08:23and
08:23illustration
08:24market
08:24is
08:25already
08:26being
08:26destroyed
08:27by
08:28their
08:29synthetic
08:30images.
08:31The
08:32fair use
08:33defense
08:33also
08:34is
08:35weak.
08:36Even in
08:37United
08:38States,
08:39courts
08:39have not
08:40clearly
08:41validated
08:41the
08:43training
08:43of
08:43commercial
08:44AA
08:44systems
08:45on
08:46copyright
08:46rocks.
08:48Artists
08:48and
08:49agencies
08:49like
08:50major
08:51stock
08:51banks
08:52are
08:53using
08:54AA
08:54company
08:55for
08:55large
08:56scope
08:56infringement.
08:59This
08:59legal
09:00uncertainty
09:01is not
09:02sheer.
09:04Your
09:05honor
09:05references
09:06to
09:07cases
09:07that
09:07have
09:07not
09:07yet
09:08been
09:08adjudicated
09:09and
09:09have
09:09no
09:10direct
09:10connection.
09:12Concern,
09:14please
09:14stay
09:15focused
09:16on the
09:16fact
09:16of
09:17this
09:17case.
09:18Concentrate
09:19on
09:20demonstrating
09:21whether
09:21or not
09:22this
09:22AI
09:23infringe
09:24copyright
09:25law
09:25under
09:26the
09:26applicable
09:27legislation.
09:29There we
09:30were.
09:31There we
09:31were.
09:32And
09:32faces
09:32that
09:32even
09:33under
09:34the
09:34defendants
09:35one
09:35terror
09:36we
09:37their
09:38use
09:38is
09:38not
09:38transformative
09:39in
09:41the
09:41legal
09:42sense.
09:43The
09:43model
09:43can
09:44produce
09:44results
09:45are
09:46too
09:47similar
09:47to
09:48the
09:49original.
09:50This
09:51is
09:51mere
09:52statical
09:53analysis
09:54with
09:55no
09:55impact
09:56on
09:57the
09:57work.
09:58The
09:58model
09:58potentially
09:59engaged
10:00in
10:01decreased
10:02reproduction.
10:04I
10:04refer
10:05to
10:05the
10:06testimony
10:06of
10:07how
10:07AI
10:08expert
10:09Jean
10:10Guivocchi
10:11and
10:12Anna
10:12Giordano.
10:14I
10:15call to
10:15the
10:15stand
10:16your
10:16honor.
10:17Mr.
10:18Jean
10:18Guivocchi
10:18and
10:19Anna
10:19Giordano
10:20an
10:20expert in
10:22AI
10:23and
10:23corporate
10:23visions.
10:25Please.
10:32Thank you.
10:36We
10:36conducted a
10:37forensic
10:38analysis
10:38of the
10:39model.
10:39In
10:40particular,
10:40we tried
10:41to
10:41determine
10:41whether
10:42it
10:42could
10:42reconstruct
10:43specific
10:44images
10:44from its
10:45training
10:45set.
10:46In
10:46order to
10:46do
10:47this,
10:47we
10:48used
10:48a
10:48reverse
10:48engineering
10:49technique
10:49by
10:50giving
10:50the
10:51AI
10:51certain
10:52precise
10:52instructions,
10:53also known
10:53as
10:54prompts,
10:55combined
10:55with
10:56constraints.
10:56We
10:57asked
10:57it
10:57to
10:58generate
10:58an
10:59image
10:59that
11:00closely
11:00resembled
11:01one
11:01of
11:01Mr.
11:01Nyam's
11:02protected
11:02paintings.
11:03The
11:04result
11:05was
11:05striking.
11:06The
11:06AI
11:06generates
11:07an
11:07image
11:08nearly
11:09identical
11:10to
11:10the
11:10artist's
11:11work.
11:11As
11:12you
11:12can
11:12see
11:12from
11:13the
11:13comparison
11:13on
11:14the
11:14screen,
11:15the
11:16original
11:16work
11:16and
11:17the
11:17generated
11:18image
11:18share
11:19the
11:19same
11:20composition
11:20and
11:21details.
11:22To
11:22the
11:22naked
11:22eye,
11:23it
11:23is
11:23difficult
11:23to
11:24tear
11:24them
11:24apart.
11:25And
11:26this
11:26contradicts
11:27the
11:27defense's
11:28claim
11:28that
11:29the
11:29model
11:30only
11:30generates
11:31new
11:32images
11:32with
11:32no
11:33direct
11:33link
11:33to
11:34the
11:34original
11:34works.
11:36Now,
11:36I'll let
11:36my
11:37colleague,
11:38Mr.
11:38Kovogui,
11:39take us
11:39further with
11:40his AI
11:41expertise.
11:41Thanks for
11:41having
11:42me.
11:42My
11:42name
11:43is
11:43Jean
11:43Kovogui,
11:44I work
11:44for
11:45Copernilabs,
11:46an AI
11:47company.
11:48We use
11:48computer
11:49vision and
11:49the
11:49VLM
11:50model to
11:50create
11:53AI models
11:54that can
11:54predict
11:55anomalies
11:56and also
11:56detect
11:57illegal
11:57behavior.
11:58Here,
11:59I am
11:59here today
12:00to
12:00demonstrate
12:01how we
12:02can
12:03detect
12:03if
12:04AI
12:04models
12:05are
12:07created
12:09a copy
12:10of
12:11artistic
12:12works
12:13not
12:14simply
12:15by
12:16learning
12:16from
12:17it
12:17but
12:18by
12:19memorizing
12:20these
12:21artistic
12:22works.
12:23For that,
12:24we use
12:24several
12:25techniques.
12:26For
12:26example,
12:27we can
12:27use
12:28the
12:28technique
12:29that
12:29means
12:30to
12:31use
12:31generative
12:32AI
12:32models
12:33for
12:33example
12:34to
12:35create
12:36a
12:36specialized
12:38prompt.
12:39In this
12:39prompt,
12:40we can
12:40include
12:42specialized
12:44requests
12:45for
12:45example
12:46to
12:46ask
12:47to
12:47the
12:47generative
12:48AI
12:48models
12:49such as
12:49GPT
12:50and
12:50also
12:51specialized
12:52models
12:52to
12:53ask
12:54it
12:54how
12:55these
12:56AI
12:58models
12:58could
12:59analyze
12:59the
13:00image
13:00generated
13:01by
13:01other
13:01AIS.
13:03For
13:04example,
13:05we will
13:05introduce
13:06AI
13:07to
13:07identify
13:08in
13:08these
13:08models
13:09for
13:10example,
13:10artist
13:11style,
13:12artist
13:12colors,
13:13the
13:13color
13:13that
13:14artists
13:14use
13:14to
13:14create
13:15original
13:16works
13:16and
13:17also
13:17the
13:17patterns
13:18etc.
13:19We
13:19can
13:20also
13:20use
13:20the
13:21visual
13:21recognition
13:23models.
13:23It
13:24is
13:24based
13:25on
13:25visual
13:26recognition
13:26that
13:27means
13:28to
13:28analyze
13:29the
13:29two
13:29images
13:29and
13:30to
13:30detect
13:31visual
13:31similarity.
13:33After
13:34that,
13:34we
13:34have
13:35specialized
13:36technical
13:36tools,
13:38for
13:38example,
13:38clips
13:39and
13:39sim.
13:41It
13:41is
13:44engineering
13:45tools
13:45that
13:45can
13:46use
13:47research
13:48similarity
13:49between
13:50two
13:50images.
13:51For
13:51example,
13:51compare
13:52two
13:52images
13:52in
13:53order
13:53to
13:53detect
13:54for
13:55example,
13:56artistic
13:57signature
13:58or
13:59artist
14:01watermarks
14:02etc.
14:03In
14:03order
14:03to
14:04compare
14:05if
14:05original
14:06images
14:06are
14:07similar
14:07to
14:07generated
14:08images.
14:09All
14:10these
14:11solutions
14:12or
14:12techniques
14:13help
14:13us
14:13to
14:14determine
14:14our
14:15use
14:16cases
14:16that
14:17the
14:17image
14:18that
14:19AI
14:19creates
14:20is
14:21not
14:21the
14:21original
14:21but
14:22it
14:22was
14:23memorized.
14:24In
14:24our
14:25case,
14:25we
14:25call
14:25this
14:26technique
14:26direct
14:27memorization.
14:28But
14:29what
14:29are
14:29the
14:30problems
14:30we
14:31face
14:31in
14:31this
14:32kind
14:32of
14:32situation?
14:34The
14:35problem
14:35is
14:35legal
14:36and
14:37regulation
14:38problem
14:39both
14:39for
14:40artist
14:42player
14:43and
14:43also
14:44for
14:44AI
14:45company.
14:46for
14:46artistic
14:48company,
14:49for
14:49example,
14:50it
14:50is
14:51violation
14:51of
14:52their
14:53works,
14:55for
14:55example,
14:56and
14:56it
14:56is
14:58loose
14:59of
14:59their
15:01intellectual
15:01properties.
15:03And
15:03for
15:03a
15:04company
15:04like
15:04AI
15:05model
15:05creator,
15:06it
15:07is
15:07that
15:08they
15:09can
15:09face
15:09on
15:09a
15:10loud
15:10suite.
15:11It
15:11is
15:11why
15:12we
15:12must
15:13at
15:13beginning
15:13to
15:14create
15:15synergy
15:16with
15:16the
15:17two
15:17entities
15:17to
15:18work
15:18together
15:18in
15:19order
15:19to
15:19avoid
15:20problems.
15:21For
15:21that,
15:22we
15:22have
15:22some
15:22technique.
15:23That
15:23technique
15:25means
15:25to
15:27create
15:28the
15:29system
15:30that
15:31help
15:31AI
15:31builders
15:32to
15:34audit
15:34their
15:35datasets
15:35at
15:36the
15:37force.
15:38For
15:38example,
15:39they
15:39can
15:39audit
15:40the
15:40dataset
15:41before
15:41training
15:42to
15:42detect
15:43if
15:44in
15:44the
15:44data
15:44tests
15:45they
15:45don't
15:45have
15:46external
15:47AI
15:48templates
15:50or
15:50external
15:51key
15:52elements
15:53that
15:53belong
15:54to
15:54other
15:54artists
15:55and
15:56to
15:56use
15:56blocking
15:57technique
15:58to
15:58block
15:59this
16:00and
16:00help
16:01AI
16:01builder
16:01to
16:02avoid
16:03problems
16:04for
16:04example
16:05loud
16:05suite
16:05etc.
16:06In
16:07our
16:07case
16:07at
16:08Copernille
16:08Labs
16:09we
16:09work
16:09on
16:10this
16:10kind
16:10of
16:10models
16:11for
16:11example
16:11to
16:11detect
16:12illegal
16:12behavior
16:13in
16:13digital
16:13market
16:14for
16:14example
16:14to
16:15help
16:15creators
16:16to
16:17protect
16:17their
16:17assets
16:18digital
16:18assets
16:19for
16:20example
16:20creators
16:22like
16:23filmmakers
16:25or
16:27yes
16:28live
16:28stream
16:29makers
16:30for
16:30example
16:31Canal
16:31Place
16:31or
16:33others
16:33to
16:34detect
16:35similarity
16:35between
16:36their
16:38solution
16:38and
16:39illegal
16:42display
16:43on
16:43internet
16:43if
16:44you
16:45want
16:45to
16:46continue
16:47to
16:47exchange
16:48with
16:48us
16:48and
16:48know
16:49more
16:50about
16:50our
16:50work
16:50we
16:51are
16:51open
16:51to
16:52discuss
16:52with
16:52you
16:52and
16:53to
16:53present
16:54all
16:54our
16:55strategy
16:56at
16:56Copernille
16:57Labs
16:57we
16:58use
16:58to
16:59help
16:59a
16:59company
17:00like
17:00people
17:01like
17:01you
17:01thank
17:02you
17:02for
17:02attention
17:03thank
17:03you
17:03Mr
17:04such
17:04a
17:05level
17:05of
17:05similarity
17:06therefore
17:07suggest
17:07that
17:08AA
17:09memorizes
17:10my
17:11client's
17:11works
17:12we
17:13don't
17:13see
17:13to
17:15the
17:15point
17:15of
17:16being
17:16able
17:17to
17:18reconstruct
17:19it
17:19exactly
17:21the
17:22model
17:22apparently
17:23is
17:23artificial
17:24the
17:25expert
17:25forced
17:26the
17:26model
17:26to
17:26produce
17:27this
17:27image
17:27by
17:27tweaking
17:28parameters
17:28that
17:29a
17:29regular
17:29user
17:30would
17:30never
17:30use
17:31under
17:32normal
17:32use
17:33the
17:33AI
17:33would
17:33never
17:34have
17:34produced
17:34a
17:34carbon
17:35copy
17:35of
17:35the
17:35work
17:36this
17:36is
17:37a
17:37textbook
17:37case
17:38created
17:38to
17:38instill
17:39fear
17:41it's
17:42you
17:44no
17:44it's
17:45me
17:45no
17:46I'm
17:46sorry
17:46on
17:47contrary
17:47contrary
17:49the
17:49expert test
17:50reverse
17:50a
17:51fundamental
17:51flow
17:52the
17:53training
17:54data
17:54set
17:55was
17:55not
17:55properly
17:56controlled
17:57and the
17:58model
17:58returns
17:59direct
18:00trust
18:01of
18:02the
18:02works
18:02even
18:03if
18:04only
18:04an
18:05expert
18:05can
18:07prove
18:07this
18:08is
18:08in
18:08laboratory
18:08seating
18:09the
18:11fact
18:11man
18:12remains
18:13the
18:14model
18:14scraped
18:15my
18:16client's
18:17works
18:17and
18:18he's
18:19capable
18:19for
18:19robotics
18:20seeing
18:22them
18:22this
18:23is
18:23a
18:24very
18:24incense
18:25of
18:26the
18:26counter
18:26for
18:27ring
18:29how
18:30does
18:30the
18:31defense
18:31respond
18:32to
18:32this
18:32if
18:34the
18:34model
18:34can
18:34substantially
18:35reproduce
18:36a
18:37protected
18:37world
18:38even
18:39marginally
18:40are we
18:41not
18:41dealing
18:42with
18:42unauthorized
18:43reproduction
18:46your
18:47honor
18:47the
18:48model
18:48was not
18:48designed
18:49to
18:49copy
18:49the
18:50test
18:50result
18:51is
18:51rare
18:51and
18:52unintentional
18:53we
18:53take
18:54it
18:54seriously
18:54and
18:55we'll
18:55refine
18:55our
18:55filters
18:56remember
18:57the
18:57AI
18:58trained
18:58on
18:58millions
18:58of
18:59images
18:59yet
19:00only
19:00one
19:00close
19:01match
19:01was
19:01found
19:01under
19:02lab
19:02conditions
19:03this
19:04does
19:04not
19:04prove
19:05systematic
19:05copying
19:06eliminating
19:07all
19:07risk
19:08is a
19:08technical
19:08challenge
19:09but
19:10we're
19:10working
19:10on
19:10it
19:13this
19:14case
19:14this
19:15case
19:15is
19:16about
19:16you
19:16understand
19:17production
19:18creators
19:19wise
19:19in
19:20the
19:20face
19:21of
19:21powerful
19:22AA
19:23system
19:24copy
19:25right
19:25law
19:26is
19:26meant
19:26to
19:27give
19:27artists
19:28control
19:28of
19:29other
19:30their
19:30work
19:30and
19:31ensure
19:32they
19:33can
19:34pay
19:35that
19:36the
19:36two
19:36questions
19:37authorization
19:37and
19:38payment
19:39but
19:39today
19:40individual
19:41artists
19:42can
19:43track
19:43or
19:44control
19:45how
19:45their
19:46work
19:47is
19:47used
19:48by
19:48AA
19:49this
19:50system
19:50like
19:51AA
19:52corps
19:52use
19:53much
19:54amounts
19:55of
19:55data
19:56in
19:57secret
19:58without
19:59transparency
20:01so we are
20:01now
20:02in the
20:02life
20:03this creates
20:04a serious
20:05imbalance
20:05on
20:07one
20:07artist
20:08alone
20:08can
20:08stop
20:09can
20:10stop
20:10up
20:11to
20:12a tech
20:13giant
20:15that's
20:16not
20:17just
20:18unfair
20:18it's
20:19under
20:19term
20:20in
20:20the
20:20very
20:20purpose
20:21of
20:21copyright
20:21which
20:23is
20:24to
20:24reward
20:25creativity
20:26we
20:28support
20:28innovation
20:29they are
20:30not
20:30doubt
20:31about
20:31this
20:31but
20:32innovation
20:32must
20:33respect
20:33the
20:34right
20:34of
20:35the
20:35artist
20:35simple
20:37solution
20:37exists
20:38ask
20:39permission
20:41paid
20:42artist
20:42use
20:43listen
20:43see
20:44if
20:45i
20:45a
20:46corpse
20:48is
20:48truly
20:49innovative
20:50it was
20:52a debt
20:53to
20:54the
20:54artist
20:55whose
20:55works
20:56made
20:57it
20:57possible
21:00your
21:00honor
21:01we
21:01understand
21:02these
21:02concerns
21:03icorp
21:04supports
21:04artists
21:05and
21:05seeks
21:05a
21:06fair
21:06balance
21:06between
21:07innovation
21:07and
21:08respect
21:08for
21:08rights
21:09we
21:10back
21:10the
21:10european
21:11code
21:11of
21:11good
21:11practice
21:12for
21:12transparency
21:13if
21:14laws
21:14change
21:15we'll
21:15comply
21:15but
21:16today
21:17icorp
21:18followed
21:18the
21:18law
21:18strictly
21:19even
21:20if
21:20not
21:20always
21:21in
21:21spirit
21:21it's
21:22not
21:22fair
21:23to
21:23punish
21:23us
21:23for
21:23obligations
21:24that
21:24don't
21:25yet
21:25exist
21:25requiring
21:27permission
21:27for
21:27every
21:28use
21:28would
21:28stifle
21:29innovation
21:29this
21:30AI
21:30might
21:31never
21:31have
21:31existed
21:32we
21:33must
21:33balance
21:33AI
21:33freedom
21:34with
21:34artist
21:34protection
21:35a
21:36corp
21:36is
21:36committed
21:37to
21:37finding
21:37that
21:37balance
21:41now
21:42the
21:43court
21:43has
21:44heard
21:44both
21:45sides
21:46this
21:47mock
21:47trial
21:48reveals
21:48legal
21:50and
21:50ethical
21:51gaps
21:51that
21:52current
21:52law
21:53can't
21:54fully
21:54address
21:55as
21:57judge
21:57of
21:57this
21:57experimental
21:58trial
21:59I
22:00will
22:00not
22:01issue
22:01a
22:01final
22:02verdict
22:02instead
22:03I
22:04suspend
22:05judgment
22:05and
22:06refer
22:07these
22:07key
22:08questions
22:09to
22:10a
22:10broader
22:10ethics
22:11forum
22:11first
22:13should
22:14a
22:14licensing
22:15system
22:16be
22:16created
22:17for
22:18AI
22:19training
22:19on
22:20protected
22:21works
22:21to
22:22ensure
22:22artists
22:23concern
22:24consent
22:25and
22:25compensation
22:26second
22:27is
22:28global
22:29regulation
22:30needed
22:30needed
22:31to
22:32manage
22:32AI
22:33training
22:34and
22:34copyright
22:35protection
22:36internationally
22:37this
22:39question
22:40demand
22:41collective
22:42answers
22:43so
22:44court
22:44now
22:45awaits
22:46your
22:46influence
22:48okay
22:50thanks
22:50sorry
22:51the first
22:52time
22:52okay
22:52now
22:54it's
22:55to
22:55turn
22:56to
22:56take
22:56action
22:57you
22:57okay
22:58what
22:59three
22:59questions
22:59three
23:00legal
23:00questions
23:01I
23:02need
23:02a
23:03voluntary
23:03to
23:04take
23:04this
23:05dress
23:06and
23:06to
23:07to
23:07receive
23:08the
23:09answer
23:09about
23:10this
23:10legal
23:11questions
23:11any
23:12persons
23:13with
23:13interests
23:14we have
23:15five
23:15minutes
23:15no
23:17we are
23:19candidate
23:19there
23:21no
23:22okay
23:23we
23:25we
23:25asked
23:26about
23:26sorry
23:27about
23:28one
23:29questions
23:50you
23:52remember
23:53about
23:53the
23:54text
23:54we
23:55speak
23:56we
23:56spoke
23:57about
23:572000
23:58directive
23:59european
24:01in
24:01copyright
24:02okay
24:02when we
24:04use
24:05non
24:05c'est pas
24:06on a plus
24:07l'image
24:07ah
24:08okay
24:08c'est pas
24:18ça marche pas
24:21ok
24:26sorry
24:27is the
24:27direct
24:30it's good
24:30tu peux me faire un copier
24:32connu
24:34non
24:37ça marche pas
24:38évidemment
24:42okay
24:43i just
24:44want
24:44to
24:44see
24:45to
24:46to
24:46to
24:46show
24:46one
24:47article
24:47three
24:48five
24:49of
24:50AA
24:50act
24:51do you
24:51know
24:52what
24:52is
24:52AA
24:53act
24:53whole
24:55whole
24:56professional
24:58in
24:58AA
24:59have
25:00to
25:01know
25:01this
25:01text
25:02this
25:02text
25:02is
25:03in
25:03reality
25:03a
25:04text
25:05to
25:06impose
25:07for
25:07the
25:07professional
25:08to
25:09to
25:10to
25:10to
25:11to
25:11write
25:12to
25:12that
25:13if
25:13they
25:14use
25:14a
25:15protected
25:15creation
25:17for
25:17training
25:18a model
25:18you have
25:19to
25:19to
25:20learn
25:20it
25:20this
25:22is
25:22the
25:23one
25:23text
25:24and we
25:25spoke
25:26about
25:26along
25:27the
25:29directive
25:312002
25:31sorry
25:34it's really
25:35hard to
25:36to make this
25:38in quick time
25:38but I have
25:39three minutes
25:42look
25:43remember
25:44about this
25:45directive
25:472002
25:48I
25:50use
25:51I
25:52I
25:52oh
26:06okay look
26:08look
26:10how
26:12we can use
26:13what can
26:15a lawyer
26:16lose
26:16IA
26:18graph
26:19directly
26:19demonstration
26:21we see
26:22that
26:22in reality
26:23data mining
26:24is possible
26:25but we have
26:26exception
26:27first exception
26:30condition
26:31sorry
26:31the artist
26:32must to
26:33not reserve
26:35his right
26:36he don't use
26:37opt out
26:38okay
26:38and
26:40just for
26:41data mining
26:42is possible
26:43just for
26:43academic
26:45or scientific
26:45use
26:46we cannot
26:47use
26:48IA
26:49models
26:49without
26:51a
26:52consent
26:53or payment
26:54the artist
26:55what
26:56want I
26:57say
26:58I
26:58got
26:59outside
26:59with this
27:00mock trial
27:01for me
27:02it's a good
27:03method
27:03sorry
27:03it's a good
27:04method
27:04to
27:05to build
27:06in
27:08same time
27:09in same
27:10time
27:10business
27:11and
27:12innovation
27:13because
27:15we know
27:15innovation
27:16in Europe
27:17is really
27:18heavy
27:19for many
27:20person
27:20but
27:21start
27:22by
27:23the
27:23because
27:24trial
27:24is the
27:26situation
27:27we can
27:29manage
27:29you
27:34that
27:37is my
27:38don't
27:40have
27:40we don't
27:42have
27:42the
27:43pinnacle
27:47I'm sorry
27:49did you
27:50understand
27:51what
27:51demonstrated
27:53it's the
27:55first time
27:55we
27:56used
27:56did you
27:57have a
27:57question
27:58well
27:5830
28:00seconds
28:02no
28:03you
28:04you
28:04all
28:04understand
28:05yeah
28:05madame
28:10was that
28:11scripted
28:12but did you
28:13know what the
28:13AI was going
28:14to say
28:17it's my
28:18circuit
28:20it's the first
28:21time in fact
28:21we prepare
28:22the answer
28:23and the
28:24question
28:24before
28:25but
28:26next here
28:27with a real
28:28agent
28:28we can
28:29imagine
28:30the lawyer
28:30because
28:31it's
28:31easy
28:32now
28:32okay
28:33and
28:34think about
28:34this
28:35imagine
28:35a mock
28:36trial
28:36on your
28:37company
28:38you can
28:38this
28:38because
28:39it's
28:40really
28:40interesting
28:40to trust
28:41your
28:42investor
28:42and
28:43the
28:45other
28:45hedge funds
28:46because
28:47you
28:48manage
28:48the
28:48risk
28:49with
28:49this
28:49thanks
28:49a lot
28:50okay
28:51thank you
28:51very much
28:52to our
28:53legal team
28:54for this
28:55mock
28:55trial
28:56as we can
28:57see it's
28:57very much
28:57a work
28:58in progress
28:58and
28:59we'll see
29:00them
29:00maybe next
29:01year
29:01with
29:02the real
29:03thing
29:03all right
29:03thank you
29:04very much
Commentaires