[Music] this week on the futurists Amir gavi I suspect maybe one of the reasons people
are afraid of AI is that it's trained on
[Music] us well welcome back to the futurists I'm Rob Turk and joining me once again
in the co-host chair is my buddy Brett hi World Travel do hey hey hey hey I'm
I'm I'm in Thailand obviously um I'm got the Tech Source conference on
this weekend which is one of the biggest uh um future Tech events so we just had the speakers dinner tonight so uh um
some interesting people in town ah very cool that's good fun I want to hear more about I want to hear more about Bangkok
um you know for the last couple months I've been getting deep into the topic of artificial intelligence and copyright I
was invited to I've been reading you you've done a series of of yeah yeah I'm posting what I found because what
happened is I went to this event that I was invited to by creative comments and they held a really cool thing it was um
uh the topic was generative Ai and copyright and they gathered together all sorts of folks from across the legal and
uh entertainment and screenwriting Industries to have this uh session to talk about how copyright and generative
AI fit together and I got to tell you Brett when I went to that event I started out with I think what I call
kind of a folklore version of an understanding what copyright is and what I found is a lot of creative people
share that a lot of a lot of my screenwriter friends share that view which is the idea that that gee you have these artificial intelligence systems
and they're being trained on on on publicly available work but quite a lot
of it is copyrighted and you know most most F don't know how that training is done the companies that do it don't
really reveal exactly how they're doing the training necessarily and so um as a result I approached this event with the
kind of an understanding like how is this not copyright infringement that was that was my starting point when I went
to this yeah you know it's it's interesting I mean um you know there there can be an argument um particularly
when you've got such widespread data access that data is sort of um you know
a a common good almost well this is the thing I mean look if you talk to the tech industry and particularly the AI
companies they're like you can scrape the web like that's been tested and proven again and again and again in the
court of law um and so I guess my Epiphany was that after going to two of these creative Comon events my
understanding of it has been completely reversed where I think I approached it from a kind of like you know certainly
sympathetic to writers and certainly sympathetic to you know I applaud you for that because not
a lot you know not a lot of people can sort of change their opinion on some of this stuff but well it's because I had
the Good Fortune to meet some really really smart people who are very focused on this and and frankly they come to the
subject matter with a fair degree of Education uh one of the people who enlighten me is an attorney from a firm
called Fred Frank in New York uh he's a uh IP and Technology attorney and his
name is Amir gavi and we've had a number of good conversations where he's been able to help me understand where my
thinking was incorrect some of the things I brought you know are kind of like sentimental arguments I'd say right
but anyway what I decided would be a good idea was would be to bring him on our show so today we've got Amir Gabi
joining us from New York uh Amir is actually not just uh not just very well versed in the subject matter but he is
actually representing some of the ca some of the companies that are involved in the litigation so without further Ado hey Amir welcome
to the show it's great to have you here thanks for joining the futur hey yeah thanks Robert Brett uh thanks for having
me it's uh I'm excited so so give us an overview of what's happening right now
in the legal landscape with all of the generative AI companies uh and and the
organizations that have copyright because there's quite a lot of activity in your field yeah I mean I think the copyright nerds are excited because
people uh seemingly suddenly care about copyright again I mean we have these
cycles and these moments and usually you know the the the fun cases that make to
the media um there will be you know recording artists or there will be
someone that claims that you know the idea for a script was was uh stolen and
so you periodically have this refresher in the media about what copyright is and what it isn't um I think we've we've
really reached an interesting moment now uh I'm seeing all manner of uh
misquotes um and sort of misunderstandings about foundational copyright law and I think that's due to
a couple of reasons one I think the nature of AI is inherently more uh it's
it's Global right these systems are being released largely you know internationally and so when you have a
song that's released you know you have a song in a language um maybe in a
particular you know style that's that's Regional country music song it has a
limited following right um but with AI it's um you know in many ways breaking a
lot of norms and and one of which is releases of uh Systems and Technology
that are available on a global basis right I don't need to bore everyone with
with the you know sort of uptake numbers with GPT I'm sure you've gone through it
right but but if you look at the bar graph that people show all the time of how many days it took um you know to get
chat GPT to 100 million years yeah yeah yeah five million in in two days and 100 million in a month so the fastest
growing consumer facing app in history but I think you're right what touched the nerve here isn't necessarily open AI
the fact that the largest and biggest and most powerful tech companies in the world are rolling this out in a global B
for instance Microsoft intends to roll co-pilot across their whole office suite of apps that's a billion and a half
users around the world and so if you were an artist uh or an author who created a copyrighted work you might
feel like your rights have been infringed like you can kind of understand it from a I guess I guess on a gut level but the point you're making
is that on a on a legal level what people sense or what they feel doesn't really apply right like the the feelings
that artists have don't really match up 100% with what copyright does I think that's right right I mean I'm very
sensitive to um the artist position in the debate and what I think it would be
helpful to do is sort of you know break those out a little bit right in in my
experience and you know I think I've mentioned this to you is in in in really every case when I speak to someone who
expresses a visceral reaction to artificial intelligence as the technology um really what what they're
coding for is a deep rooted fear of being replaced and they very rarely
articulate their objection in that way um they they they you know it's only in
the rarest circumstances where someone comes out and says you know what I don't like this because I think it's G to take
my job or I think it's gonna you know render me irrelevant uh you you start to
hear a bunch of other you know red herrings or you know euphemisms which to
me ultimately I realize code for um I'm going to be replaced and of the ways in
which we're seeing that is this this emotional response and people I think to
some degree grasping for straws um and and are are claiming copyright we saw it
most recently um and I don't know if if you all have covered this but in the uh in the pros Craft um uh news right so
Pros craft is a really interesting uh was I have to put it in past tense because the site's been taken down as of
last week but it was a really interesting website that uh indexed about 25,000 books and what it allowed
authors to do was to you know submit their language and to see you know how active it is relative to others and just
do really interesting semantic analysis I'd like to do that that sounds yeah yeah I read the story in the times that
kind of broke out yesterday and the day before about this thing so here is like a you know a single person coded not a
big tech company doing something he thought was quite useful which is allowing authors to run semantic
analysis on their books I'd love to do that and uh and tell us what happened toir because I guess he got a reaction
he did not expect he so um I think it hit a fever pitch with the Gizmodo uh
story and there were a number of media coverages but but uh there was a
vociferous reaction from authors claiming that it was plagiarizing uh their books and that it
was rank uh stealing and and it was it it got so loud that that um that the
Creator Benji Smith just decided to to take down the site uh which I think is unfortunate because it was a it was a
very valuable tool and I think this is where I'd like to highlight the distinction between you know what
copyright law allows and what people decide to do right I think we have to separate out the law as it exists right
what people are allowed to do versus what people decide to do meaning the Norms in Society uh where we draw those
lines where policies lie um that's very different than than what one has the
right to do right I mean very famously I think a lot of us lawyers remind clients that just because you have the right to
do something doesn't necessarily mean you should um and I think we can we can talk about that a little bit later in
the context of of Licensing data sets U for for AI training but here here we
have back to Pros craft you have an example here of uh um a single person who had an idea they thought was fair
use they went ahead and built this system not trivial like a ton of effort went into building that system some
people found it useful so actually it was doing the job it was supposed to do but now because of a threat of an of a
law a lawsuit effectively this guy has now been silenced he's taking this site offline and so I wonder I wonder whether
it was the threat of the lawsuit or it was actually just this moment we are in the summer of 2023
where authors content creators uh voicing an objection was enough um I I
suspect it was I suspect it was the latter um where where they they frankly
just made the founder of the site just feel bad for what he had done which is I
think in in in uh many ways I mean pretty clearly legally defensible
there's a long line of uh of cases in the cont of text and data mining you'll
see the acronym TDM in the field um starting with Kelly vers arisoft and in the ninth circuit
the court found you know there was fair use to get a full text copy of of images
on the internet to enable web image search and then you had the I paradigms case um in the fourth circuit in 2009 I
paradigms if you remember is a software that is a plagiarism detection tool
right so in order train the plagiarism detection tool you had to feed in a bunch of papers people objected um but
the courts found that that also was fair use and then of course famously right
the Google Books cases the author Guild versus happy trust and and and Google but you know that that everyone I think
at this point understands the fact pattern which was Google scanned in entire libraries of books through um
high-speed um scanners in order to uh make the text searchable and allow uh
them to run a bunch of you know textual and semantic analysis like you know uh
word frequencies and and sort of language Evolution really nerdy
linguistic stuff and the court found that that was also fair use and so and that was tested again and again I think
even the Supreme Court eventually reaffirmed the the the appeals Court's decision on that subject uh I guess
here's the point about what you're describing for the folks who are listening uh what we're talking are pretty interesting yeah it it's
whether or not technology companies have the right to access publicly accessible work they may be copyrighted work but
it's a publicly accessible work can a machine read a document that's on the web or a picture that's on the web and
do something with it and the big question is what's it doing what is the use what's the nature of the use and the
technology companies will tell you for sure this has been tried and tested again and again as Amir just shared with
us and every time the court comes down and says well yeah because What's Happening Here is these machines are reading it there's no there's no law
that prohibits a machine from reading a document and then it's taking that information and doing something new or
something different with it and so this brings us to the subject of fair use and transformative use Amir can you tell us
help us understand what is fair use and then let's later next get into the subject of transformative use because I
think a lot of this depends on these two concepts yeah I think so fair use really
rests on the principle that our society would be awfully boring uh if authors
had an absolute Monopoly on The Works they created right um and and to point
towards the moment and the strange place in which we found ourselves we we
currently are in a place where we have a comedian right a professional uh satire
maker a satirist has sued open AI claiming essentially absolute copyright
in the works she created right that's the Sarah Silverman case yes and and and so what fair use does right and this is
a concept that's enshrined in the Constitution itself I mean we can go
that I don't normally love quoting the founders but but here I mean article one section 8 Clause 8 right um that that
the US Constitution grants Congress the power to promote the progress of Science and useful Arts by securing for limited
times to authors and inventors the exclusive right to their respective writings and discoveries and so there's
essentially a quidd proquo right you get this Monopoly so it's for the life of the author plus 70 years and then you
know a slightly different term if it's if it's in a non-natural person but what you get protection for under copyright
are original works of authorship right um and by the way in in in our world
original it's like not even it's not a hurdle it's barely even a speed bump
right so it there's a minimal minimal creativity requirement um so it's an
original work of authorship plus it has to be an eligible type of work right that is fixed in um a tangible medium of
expression right you can't copyright a speech or you can't copyright a a live performance in performance that's right
and and so when and and so you know the the idea of what is copyrightable is
relevant here because you're not allowed to copyright facts
ideas um or or methods right it's the creative expression that is protectable
and this is really be a fixed work and it has to be a single fixed work that you're copyrighting so you can't just say like my style as an artist is
something I'm copyrighting that's right style is protectable in in other ways through other Concepts right there's
rights of publicity we see trademark right the yeah the L Vuitton red Souls
um and sort of caterpillar colors right you can have certain attributes become protectable but those are other
doctrines it's not copyright and so I think you know going back and and connecting the dots you know we we have
this powerful new technology it's for the first time in a while generally
meeting the hype right we were told Segways would would change the way we architected cities and we've been fed
you know s we we're told to expect self-driving cars by now and and and the
tech hasn't really met the promise um but the first time people started playing around with either the image
generators or the chat Bots uh you could see it right there was there was
something there that felt different and I think that's where um that's where the
sort of these visceral reactions these Deep Emotions are are are coming from and and again going back people are
reaching for copyright without understanding what it really is right like it's important well I want to I
want to come back to a point you made earlier because it's it's important you mentioned that um one artist felt like
they had a comedian had a sense of absolute copyright let's hang on to that idea for a second because when you
reference the Constitution quite a lot of people quite a lot of people that I know believe that what the Constitution
asserts is that people have this absolute right but what you're pointing out is something that's really important it's a little bit subtle in the way it's
framed in the Constitution the purpose for copyright in the constitution is to serve the public to serve the interests
of the public and all subsequent copyright law is a balancing act between
serving the public interest which is to say that the public needs a way to engage with ideas and if we treated ideas in books like real property where
you could put a fence around it and nobody else could come into it well then people couldn't engage with ideas without infringing so the so the
founders in their wisdom they said we have to create kind of a loophole here or like a little Escape valve where Society can engage with these including
quoting sections of them if they want to write critical reviews or use them for research purposes or use them for education and so on and so that gave the
opening to fair use and fair use is actually a judge invented the concept in
a ruling I think in 1840 uh but it was later encoded in the 19 act yeah Lord Edinboro
actually and the point there is is that uh the Copyright Act the 76 Copyright
Act sets forth very clearly exactly what you can do as Fair use it cites a number
of instances and it did that specifically to kind of clarify the matter and to make it emphatically clear
that there is no such thing as absolute copyright where you have absolute control over all uses you don't they
tell us about derivative works and tell us about transformative use because I think these two concepts are going to be important as well yeah so you know
copyright covers it provides the author protection over a couple of things right
and it sort of enumerates these these bundle of of of Rights um you know it gives you the rights to
control reproductions of the work distributions of the work public performances displays all those were
familiar with one of the things it also allows the author to control are the right are the rights to derivative works
now derivative works are um sort of famously um uh poorly defined in some
ways right and and just because um a so a derivative work is essentially a
subsequent work that uh relies in substance on on the prior work but that
alone so that's necessary but not sufficient right and sort of the bright line for what is a derivative work um I
I is is um kind of difficult actually to pin down and it depends on a lot of uh
particular facts and circumstances right so like in a copyright lawsuit or copyright case the court would pay very
close attention to what they call the fact pattern which is to say like how was it used what was the intended
purpose was it a commercial use or not was it for private use or public use was the was the derivative work published by
someone else and so so there's all these factors that mean that there is no one- siiz fits-all solution to copyright each
case is going to be determined um on its own merits and on the facts of the matter yeah that that that that's right
I mean you know the statute itself if we want to go back to the copyright statute right it it it it does say that a
derivative work is something that's based on a pre-existing work and it gives examples you know there there will
be examples like translations musical Arrangements uh you know moving from one
like making a you know a uh a book into a motion picture uh th those sorts of
things um but really the line as to what is a derivative and and what is uh allow
what what is a is a new work and what is a substantially new work in which the um
the original author cannot control um that that is you know look we've been
fighting about that for 30 40 years right um well i' say copyright well evolves because you know technology
evolves and media evolves and consumer habits evolve and culture evolves and so I think in the wisdom the the you know
the the founders who wrote the Constitution they didn't spell out exactly how to implement copyright law
they left it to the courts and it's been a moving Target ever since because Society has moved on I think that's on
and it and it's what feels right right the the the the broad test is the D the
the the second work in order to infringe has to take a substantially similar
Quantum of the first work and so what substantially similar is to your point Robert I think that changes you know
decade by decade yeah I think you know gen Z probably would have a different
view on that than our generation right if we ask them to give their views of
what what a substantially similar you know second work would be as opposed to
so so question our audience is going to be thinking right now then is to say hang on these uh generative AI systems
have been trained on hundreds of millions of copyrighted works we don't know exactly how many because that information is not necessarily made
public but we know that there are definitely lots of copyrighted works in the training
set now the resulting prodct gole books trending set as part of it actually okay
so and then the and then the the product here is a is a device that allows me to generate new work like me as a consumer
as an end user I can go in and type in a short command and open AI will generate some paragraphs or mid Journey will
generate some images and so forth so there's really two parts of two areas
that could be potentially infringing right one area is on the training um and then the second area is is my use right
of this machine right so um let's break those two things down first of all let's talk about training I've heard the term
fair train um can we uh take a quick break before we do that before we get into that okay sure thing let's hang on
then and we'll we'll take a break and then and then we'll come back in a sec so we're with Amir Gabi of freed Frank
and this is the futurist we'll be right back provoke media is proud to sponsor
produce and support the futurist podcast provoke FM is a global podcast Network
and content creation company with the world's leading fintech podcast and radio show break Banks and of course
it's spin-off podcast breaking Banks Europe breaking Banks Asia Pacific and the fintech 5 but we also produce the
official finovate podcast Tech on regg emerge everywhere the podcast of the
Financial Health Network and NextGen Banker for information about all our podcasts go to provoke FM or check out
breaking Banks the world's number one ftech podcast and radio [Music]
show welcome back to the futurists uh we have Amir Garvey uh from Fred Frank um
he's a intellectual property and Technology attorney and we've been talking about all of these uh legal
challenges and um you know part of this alignment phase of the adoption of artificial intelligence where we figure
out where AI is going to fit in and right now we're coming up against some
uh legal precedence uh based on early technology development or some of them as we're saying before break go go back
the 1800s um and you know we are faced with something essentially new in terms
of the way we transform data and so forth but um Amia just I just wanted you
to clarify um you know we talked about the the base basis of the training
models and and extracting data there but the other implication you just just made in that conversation and what Robert was
saying before the break is that um actually who whoever's writing the prompt or the user that says say all
right generate for me a story in the style of this author um you know and uh
you know then attempts to use that um you know either for entertainment purposes or whatever could they could be
a violator of of laws could they yeah I mean look the way copyright law works is
its strict liability it's not really intent-based and so um you you have you
know um an an author could claim that someone who is you know
uh creating the prompt that is asking the platform to generate certain U
images or text um you know copyright law could view that as person as the
infringer um and of course we see a number of cases in which plaintiffs are
uh you know going directly at the platforms um at this point you have um
stability mid journey devv and art open AI meta and all of the Google Affiliates
uh the relevant Google Affiliates um have been sued um as platforms for uh
infringing either on the training or on the outputs that they create right and
and sort of unpack that a little bit um right the input uh argument is that you
you have taken a Quantum of uh copyrighted material
um and you you have made use you have made an unlawful
reproduction um of that copyrighted work without the author's uh consent that's
the training side um the output side is you know your outputs are substantially
similar to uh the original work and it is therefore um a derivative work and
and that right as we discussed under copyright law rests with the
author now with the with the you the end user uh
example I used to work at a movie studio and I've seen this story played out again and again and again uh with the
VCR with the DVR um with various attempts to do Cloud DVRs and so forth
there have been a number of cases and very typically you know what happens is the company that owns the copyright it's
always a company it's not an individual compy owns a copy brings a lawsuit against the company that makes the Consumer Electronics device claiming
that this is a copying machine that you're going to be able to make it possible for people to make millions and millions of illegal copies and they
bring a lawsuit and the courts always end up saying no actually there's a substantial non-infringing use here
which is that people are just going to use these things to watch their own home collection of videos which in turns out
is what 99.9% of the people actually do with the thing so they're correct about that and the cases end up always winning
in favor of the company that makes the device is this any different it looks to me like it's the exact same thing you
know I could definitely go to Mid Journey today if I wanted to and type in uh give me examples of Harry Potter
characters in Star Wars right so there I am brand there's brand implications for for individuals like authors or actors
but if I were if someone was to make a generative version of you and play that in a movie and you did something
really imoral right the machine the company that makes the machine is not
liable for infringement no I it's my choice as an individual to do that and I are we I will observe that we are almost
40 years from the Sony vers ba uh Universal City Studios case the the
betamax case which is I think the case you're you're referencing which which Robert which said that you know U so
long as there are substantial non-infringing uses of the technology we're not going to presume that uh
making the platform available is is an inherent you know per se
infringement and so by extension we can expect the courts to look at that precedent right I mean obviously we
can't tell what the courts are going to do but um but my sense is that this has been this this matter has been tested
again and again and usually sounds like it's not an easy Road for people who are challenging AI that's what that's what
it sounds like to me so I I I have to be somewhat careful
to not engage in in hubus here I I will observe a couple of things um I I think a lot lot of academics when they look at
the facts um they they they believe that there is fair use in uh in in most cases
in training the AI systems on publicly available data um I think it's hard for
people to opine on the output cases because those are inherently fact
specific you act you literally have to look and compare the two images and to see you know to what degree they are
similar um you know copyright law can get rather technical um I'm sort of
famous amongst my peer group for reminding people that that probably a lot of software that we think is
copyrighted or copyrightable probably is not because um much of the code in
particular as we have objectoriented programming these days I mean let's set aside codex and co-pilot but when you're
using object-oriented programming you're really um all almost purely creating
functional Works um and you're you're dragging and dropping based on what you
want the system to do um and and and as you may know under copyright again
you're you're allowed to protect the artistic elements but the functional elements that that's the S ofair
Doctrine the functional elements are not protectable and so time and time again
as we've really gotten into deep cases this was also born out in in Oracle versus Google right uh famously where
where um thousands of lines of code were were allowed to be uh copied verbatim um
and and put into Android and that was not deemed um an an infringement um you
know fair use was found there too um I think it's it's just important to highlight you know that that
aspect and I think if you were to take that take a case like that first of all you'd have to find an individual user
and you'd have to understand that they were trying to make copies of copyrighted works but even if that
person intentionally did it you know if they wrote Into The Prompt exactly you know those those terms that I use even
then it's going to be really hard to point to the company that makes the AI and say that they had some role as
contributory to that I don't see how that's gonna happen I want to be careful I want to be careful Robert because you
know other than for for sort of Damages purposes for liability again copyright
is a strict liability regime intent doesn't really matter you just look at the outputs and you look at the
infringements I think this is extremely relevant when I'm hearing the debate
around whether this is fair or not and people try and bring in copyright uh
it's really important to notice that in to to to highlight that in the US um we
don't have what's called a sweat of the brow doctrine that exists in in some other parts of the world which is to say
just because you tried really really hard to create the work doesn't mean that it's protectable under copyright it
has to be protectable because you know it is an artistic expression and it's not capturing facts or ideas or methods
right that's what makes it copyrightable not because you spend a lot of time on it right and and and that sounds silly
but it's really important because I think one of the things I'm reading um and when I hear artists objections
is they feel wronged because they spent a lot they may have spent months or
years creating this work and they're offended that the that the the AI model
um can create something and I'm not going to sort of
editorialize as to whether it's approximation that can create something right um in a matter of seconds that
offends them and they they feel like that feeling should be actionable and
I'm not sure I agree as a matter of law I'm sympathetic as a matter of policy and mores yeah and where we want to be
as a society um you know I I think all
frankly this is part of a broader point I I think maybe even the um you know the
the AI you know the startups and and and a lot of the um the founders of these AI
companies may want to step back and reread Coon's you know structure scientific revolutions um when because
we are it feels like we've hit an inflection point and there are paradigms
and whether we want to be U thinking about copyright and you know the
probably the greatest um you know illustration of of this from a
disruptive nature perspective or analogy is probably the printing press you know I mean the printing press cre
chaos in terms of you know the conventional Church hierarchy and its
involvement in government and and so forth um you know was you know it was really it was a really interesting
period of time and it it feels like the something similar is happening with artificial intelligence well I think you
can look back I think there's a couple things and I think Robert hit the nail on the head I think a lot of the objections are also being promulgated by
people uh people by corporations who will be disintermediated yeah you can see you
can see how this both frees data um but also makes content
generation uh more direct and easier right and those two things threaten a
lot of Gatekeepers who have quite literally made businesses out of
controlling the distribution of media this is a
direct um you know challenge to some of those uh
those those sort of structural I mean we use we use AI on this show to create Clips we run the
episodes through and the video and audio through an algorithm and it cuts out Tik
Tok and you know Instagram um stories for us you know like with AI it's just
that that text's amazing to use as a Creator I mean previously I would have had to hire team they you know I'd have
to give them um time stamps all of that sort of stuff and then we you know we'd have maybe have to have a conversation
about it then they'd have a couple of attempts at getting the video right and now I've got an algorithm I can run I
mean one thing one thing that's worth noting though is I I I suspect maybe one of the reasons people are afraid of AI
is that it's trained on us right like trained that's true we've talked about
that many times on this show don't replace us yeah that's the fear no not just that but it is you know as one who
is you know uh it feels a bit of a nihilist well if we have dark if we have
a dark sense of humanity right the idea that we would train autonomous or semi-autonomous agents based on you know
the contrails of of what we've put out um that can be a scary thing sure yeah
look at you know the the conversation about generative Ai and copyright very often um the the poster child for this
is a writer a downtrodden writer who's underpaid or a starving artist who has a moral right to control their work or
something but let's get really clear that's an argument copyright suits are brought by corporations most copyrights
are owned by corporations most copyrights are renewed by corporations so what you really have here is a battle
between two gigantic kinds of Corporations the media companies which represent the past and the technology
companies they're trying to push towards the future now they're pushing towards the future and they're not asking for permission that's why there's some
friction here that's why there's push back but let's get really clear about what's at stake here because when we hear about someone saying oh you know
I'm going to sue somebody for copyright infringement I've got rights I want to stop it what they're saying is that they
do not want that use to persist they want to cease that use they want to stop that and if the use in question is one
that's going to be Progressive and in and and bestow new powers on people and I would argue that was the The Epiphany
that I had I realized wow if the technology companies are thwarted here by the by the copyright industry by the
copyright monopolies then what that means is that this new Incredible skill that we got this new gift this
superpower generative AI that makes bad writers into mediocre writers or mediocre writers into better writers or
it streamlines a business process or it makes it possible for someone to do their their marketing faster uh or makes
a coder it makes it possible for a coder to write code faster and better if the copyright industry prevails they're
going to prevent those people from getting those skills so what the court has to weigh is this massive benefit for
half of humanity let's say versus the rights of a very small number of people
and let's be really clear they are not artists that are bringing these lawsuits they're corporations that are behind it
and so it's a battle of big corporations and right now you're right the public sentiment is running against big Tech because big Tech has been pushing us
around we do feel bullied by big Tech but let's not make any mistake about it the Walt Disney Company and the other
big media companies they've been building copyright monopolies and they've been pushing hard to extend well
the extensions of copyrights that's that's been led by players like Disney you're right that's right for dozens of
years right so today the copyright term is effectively 100 years let's call it Perpetual yeah when when the
Constitution was written it was 14 years plus another 14 how did that happen it didn't happen by accident it happened by
assiduous marketing lobbying pushing driving you know Disney famously spent $6 million supporting candidates for
office who helped them get the copyright extensions that they sought it and they got pillared for that right and even the
Supreme Court kind of held their nose when they WR wrote A A ruling on the subject but they let it pass because the
court at that time wasn't going to get in the way of the legislature who knows what they'll do today but I guess the point I'm trying to make here is that
this is a battle between big companies it's not a really a battle about artists uh that's a smoke screen yeah I mean to
your point right um Getty uh images is sued one of my clients uh stability AI
uh in the US and the UK um for you know uh for copyright
infringement of their works I think one of the interesting um ironic and maybe
humorous uh outcomes of that has been there was a lawsuit filed um I think
about a week or two ago uh by a photographer um who who saw the Getty
lawsuits and uh he filed his own suit in the southern district of New York um
claiming that Getty uh wrongly unlawfully is trying to charge other
people to license his copyrighted work when for years he has revoked his
license it was a license first granted to Associated Press and sort of somehow found its way uh to Getty but he never
expressly granted the license and he has for years been trying to get Getty to acknowledge um that they don't have the
authority right to make Downstream licenses and so you know it's a it's a complex topic I mean I I think it's you
know another thing that makes me chuckle is I I think back Justice sudor um uh noted in in the uh Campbell case right
which which was sort of seminal for identifying transformative use yeah the same as two Live Crew par two Live Crew
parody right he he noted this sort of thing about artists which you're highlighting is that you know they they
they ask for criticism but they only want praise um and you know I I think
it's it's we have to be sensitive in society to people's life work but I
don't think 21st century America will countenance a chilling effect or uh
shutting down AI training now why do I say that again um I think the lawyers
I'm working with and alongside um you know a lot of smart lawyers um who are in you know the cases
I'm handling and some of the the other cases um out there but we shouldn't be
precious about it taking a step back right the US and China have identified
AI as being qu to National Security for these countries on a go forward basis
right they don't think this is a fluke so China famously made it a part of their National plan the United States
has gone so far as to ban the Nvidia A1 100s h100s and and we saw even this week
the latest round of tech Investments um in these AI Technologies the US also
pass a $250 billion uh chips act to streamline the supply chain of
semiconductors which are valuable to uh this technology and so I I think it
would be a misreading and misunderstanding um of the importance to say that you know the courts um even if
they get it wrong at the district court level that this will be the end of the story and that AI training will be shut
down in the US I I quite literally can't fathom that from a from a poliy
perspective an economic it just doesn't work right so if the court was to get it wrong I think you're going to see the
different branches of the US government have to even in the EU where um they now
have um demands for data transparency in terms of the models they're not saying
you can't use the data right they they're just saying at the moment that that needs to be transparent but well
they are they they have right there's there's algorithmic transparency they they they make essentially what's what's
called a risk-based assessment they say we're going to segment the market and and depending on how you intend on using
AI we're going to regulate it differently right there are certain high-risk areas that are off limits you
know you're not going to be allowed nice to use automated systems on AI for you
know nuclear plants and and other things like that but but lowrisk you know if you want to like create um you know hi
cus then have added yeah you know it's funny when we talk about regulation um
my observation is that the path the us is going down as the Congress tries to figure out how on earth to regulate Ai
and the AI companies are asking for it right Sam Alman famously went to Congress and basically begged them to regulate open AI in other companies but
the conversation is starting to revolve around the concept of Licensing where somehow some smart person in the
government will decide whether or not a company should get the license uh to to use AI or to develop AI models and so
forth and this is an irony to me because we have this problem already with copyright where copyright is granted to
a company and then the companies that that start to accumulate lots and lots of copyrights they have a natural
economic incentive to go work very hard and Lobby Congress know support candidates who will extend copyright
term and thereby help them build a monopoly which is definitely not what was intended when the Constitution was written so on the one hand you've got
this kind of like you know copyright Monopoly I would say and now the AI companies are going to Congress and
they're saying please regulate us we think think a licensing me mechanism is the right way to do it I'm like wait that's the exact same thing so you're
now are going to create another Monopoly with a high barrier to entry uh you know there'll be some process of vetting
there'll be some requirements you might have to put up a bond or something else which means it'll be a barrier to new companies in other words it's going to
entrench the leading companies uh in their current position and then they'll end up looking like monopolies on the
other side that they're opposed to I find this extremely funny that no one else is you know Congress doesn't seem to see this as a problem I guess it's
their only regulatory tool they've got is licensing yeah well I guess they see it like the FCC you know in terms of
these um you know that they are platforms that are generating content like you would broadcast or print or
media I guess that's how they see it there's a licensing analogy there I think they like the gatekeeper function
they like the idea I mean it's rhetorically persuasive to say you hire some experts they review applications
and they will decide what is safe I think it it's it's actually born out of this argument
that Ai and and I think this is a bit of a of a of a fallacy frankly um the idea
that that AIS are all black boxes and that they're inherently unknowable I I think look they're incredibly complex um
I you know advise companies on on building these things out um I don't
think they're they're unknowable um you you you know no I think it's fair to say we know how they
work we just don't know why they do well let me give another example then how can
you regulate something if you don't know how it works right proac Robert proac right I I did so I studied Neuroscience
before I went to law school I was doing bench science on uh ssris and what was interesting was we
knew at the time that these things would have an effect once we gave it to stressed rats three and a half weeks
later but no one at the time knew the mechanism of action and that was actually quite unique it's one of the
First cases where where you you had uh a product and in this case right it's being dosed ultimately in humans and
even at the time it was dosed in humans we didn't know the precise mechanisms of action we knew that there would be extra
serotonin um um you know and and and flooding in the system um but but but in
in in most other sort of pharmaceutical compounds we knew how it worked we didn't really know how it worked there
and yet we decided there was you know a public good it was better to help people fight depression um you know uh without
without knowing all the interstitial steps right which is actually exactly how a work
right I mean we we're running out of time so I do I do want to um get back to
sort of the core the kernel of our show here which is we are futurists right and
we do like to think about how this might play out now I I I I want to separate
you from the anxiety about making for about what's going to happen with specific cases and say all right you
know let's forget about that let's talk 20 or 30 years out you know can you imagine a time in in the future and what
what do you think um you know how do you think the law will have been impacted by
artificial intelligence generally and and what are your what are your thoughts and musings on that as as just a you
know in intellectual point it's a great question I think it's hard for me to
think about AI um separate from the advancements in Hardware I think about
how we got to where we are today um you know when I was studying neuroscience
and and and computational Neuroscience there were people who theorized about this but the hardware didn't exist and
the data didn't exist and so projecting out 20 years from now we will have some
variety of quantum Computing so our Computing capabilities will be substantially different than what they
are today I think we will have and even the accelerated Computing stuff that uh Nvidia showed off the last few weeks
Robert you you and I were talking about that the other day really amazing some really amazing advances coming Jang said
to me what will you do with a laptop that's a million times more powerful because that's what you're gonna have in one of the things one of the things
that's going to happen is we're gonna have different types of content generated and so I'm wrapping my it's a
great question I'm trying to synthesize in my head in real time this idea of
amazing hardware quantitatively and qualitatively different data that can be
you know used manipulated um I think it's going to be exciting I think we're going to hit a reset at some point where Society is
going to have to determine um a different set of ground rules I don't
think we're there yet I think we're going to be doing I have a theory on I have a theory on when that's going to
happen here right tell me and that is that at some point in the future over
the next 30 or 40 years at some point we're going to decide that the management of the legal system largely
especially with contract law because of smart contracts but generally even in terms of justice and so forth is better
handled by AI for consistency etc etc right so I wouldn't disagree you're lose
your job you're GNA get I wouldn't disagree as long as and that's probably a good thing by the way Robert I would
disagree that we have publicly prosecutable algorithms right I mean
these things have to live in the point of that is if you're going
to um impute that level of automation on society you're also going to have to
have a review of the laws on the books and so forth to see their fit for that
type of level of Automation and that's going to result in a massive sort of filtering of the legal system towards
this autonomous system that's okay right when we moved from horse and buggy to cars right we adopted new new uh Norms
right we had the red flag laws initi I me yeah you know this is the advancement let's not forget this is just a part of
the Arc of Technology right and it's it's another
right we're on a spectrum here um we haven't landed on a new planet it feels
like magic but it isn't right these this is all based on you know principles um
that that are actually pretty well defined okay M um I'm thinking about our
audience right now and I bet there's some folks out there who have two questions so just as a Parting Shot maybe I can ask you for some free legal
advice uh the first question is um if there's a person listening and they're thinking well maybe I should be careful
maybe I shouldn't use these systems because maybe I'll be infringing copyright is it correct that um it's
okay for you to use these systems for your own personal use if you're not publishing commercially that there's a
very low likelihood that you're going to be in any kind of trouble if you're using the system even if you're trying to make um let's say
fanfiction so I'm gonna open out by quoting my friend Mark Lemley and saying uh this isn't legal advice if it was
it'd be followed by a bill um but um I you know for people using it for their
own personal uses look unless you have the case in which the rights owner is
going to you know uh make naps or Redux right whether you're gonna have people you know corporations going after people
in college dorms it seems extremely unlikely um that individual
users uh I mean the will will be you know sued for copyright infringement it
would be uh an incredibly bizarre uh move by a corporation to do that uh
probably be pretty ton deaf in this world I mean we think back at Napster that was pre
social media Tik Tock um it was you know I I think it would be it would be a huge
mistake for a corporation to do that having said all of that I will go back and and remind us um that copyright is a
strict liability okay so the second question is this um let's say I'm using mid Journey
which I am doing every day I'm loving it it's super fun for me to use can I copyright the work that's generated by
an AI system like mid Journey or chat GP can I get a copyright for that are you asking me what I think the
law should be or what the copyright office currently thinks tell me about the March copyright office guidance and
then tell me what you think look the guidance from the copyright office basically said that uh if the AI system
is creating the output you will not that that will essentially become a public domain work that it's not copyright
eligible because the system has generated it the copyright office has has maintained over the past couple of
years this human authorship requirement there are a lot of debates um as to
whether that has been effectively just you know conveniently made up by the copyright office or whether there is uh
legislative history or textual support in the statute I'm not sort of gonna opine on that and get into that I will
say um and and and I have a a friend colleague Joe grz who is representing
Chris castanova um before the copyright office is challenging their denial of of
her uh of copyright ability and authorship uh for her Works
um I you know I personally I think that these systems understanding how they
really work they're a lot closer to a digital camera and if not even a digital camera the cameras on our phones right
the reality is the cop the basis of the copyright office's rejection is that there is an indeterminate result that
the system there's not a onetoone correlation between the inputs that the user makes and a deterministic output
that that that there's a statistical analysis done by the model and you get a variety of outputs you may get four 16
depending on the uh the the system uh outputs and you you sort of choose that
as a base and you work from there right the copyright office now they didn't say it's an absolute denial they said the
human elements so if you take an output from an AI system and then you take that
to photoshop and you it's funny now that using Mo Photoshop is viewed as manual
but if you if you manually use Photoshop to manipulate the AI systems output you
do get uh you can get um you know copyright in that incremental
part hurts right now what what what I think the copyright office is going to hear um from various folks is well wait
you know we currently allow copyright for people taking you know pictures uh
with their digital camera there is a ton of pre- and postprocessing that goes on
right including artificial intelligence in those cameras exactly if I had to manually adjust all of my settings in
order to to take a picture of my phone they would be awful it wouldn't look anything like what's happening now so
basically I just I literally just press a button and the system handles the rest
and and the analogy that I think people are making and I I think they're mostly right is the prompt engine
gives a pretty detailed um guidance and effectively does the uh analog to
pointing the camera and hit submit and
the system will come with an output that you work from the two aren't really as different as I think the copyright
office has has laid out in their um in their guidance this is great well Amir
thank you very much for your time thanks for your very thoughtful answers to our questions uh it gets a little feisty
here on the show it it was it was quite a technical show it was interesting I like shows like this how can we follow
you like how can people keep track of you and the cases that you're working on to see what happens yeah I'm uh you know
I'm on LinkedIn I share ideas they can reach out um as you know partner at a big Law Firm I am uh endlessly reachable
people should um you know feel free to to to send me a note and I'm happy to
you know um share ideas with folks very cool thanks for your time thanks for joining us on the futurists and uh Brett
it's always good to see you big thanks to Kevin our engineer and Elizabeth our producer and the rest of the folks that
provoke media for making the show possible and thank you to our listeners uh your support is what makes this show
happen and we are thrilled to have it if you like this show please tell a friend that helps us get more audience and that
helps us keep making more shows so we'll be back here next week with another person who's working actively on
constructing the future that they Envision and we will see you in the future in the
[Music] future well that's it for the futurists this week if you like the show we sure
hope you did please subscribe and share it with people in your community and don't forget to leave us a fstar review
that really helps other people find the show and you can ping us anytime on Instagram and Twitter at futurist
podcast for the folks that you'd like to see on the show or the questions that you'd like us to ask thanks for joining
and as always we'll see you in the future