Name: Does Gen AI deserve intellectual property rights?
Uploaded: 2024-01-15T14:32:47.721Z
Duration: 505.4

enrico again.mp4

Enrico Bonadio V3mp4.vtt

English

{{00:00:00}}
So it was a provocatively move on

{{00:00:05}}
argument that machines can be and
should be considered as inventors

{{00:00:11}}
and patents should be granted.

{{00:00:23}}
Intellectual property laws govern the
creations of the mind, inventions, so it's

{{00:00:30}}
a set of law which protects and rewards a
human ingenuity, human creativity, as well

{{00:00:37}}
as the use of trademarks in the market.

{{00:00:44}}
Well, AI, artificial intelligence,
machine learning poses a series of

{{00:00:48}}
challenges to intellectual property law.

{{00:00:50}}
In particular when it comes to
generative AI, these machine learning

{{00:00:55}}
systems, what they do is to be fed
with a huge amount of data, huge amount

{{00:01:03}}
of information to train the machine.

{{00:01:05}}
So one of the first challenges,
especially under copyright law, is this

{{00:01:10}}
massive use of data by the machine.

{{00:01:12}}
Does it amount to potentially
copyright infringement?

{{00:01:16}}
Because these machines  are fed,
are trained with content which

{{00:01:21}}
might be protected by copyright.

{{00:01:23}}
And the use of this content, of this
information, is an expressive use in the

{{00:01:29}}
sense that the machine uses such data
in order to process it and then create a

{{00:01:36}}
final output, a final output which might
be potentially in competition with the

{{00:01:42}}
data with which the machine has been fed.

{{00:01:50}}
What they can do is to
bring legal actions.

{{00:01:52}}
This is the legal tool provided, offered
by the system, by the law,  to start

{{00:01:59}}
a corporate action, a legal action.

{{00:02:00}}
Claiming that such use of data is
infringing their copyright, but

{{00:02:05}}
whether these actions are going to
be successful is not so certain.

{{00:02:11}}
What the developers of machine learning,
of these generative AI machines,

{{00:02:17}}
what they could claim back is that
the use of such content is fair use.

{{00:02:25}}
For example, under U.

{{00:02:26}}
S.

{{00:02:26}}
copyright law.

{{00:02:27}}
U.

{{00:02:27}}
S.

{{00:02:27}}
copyright law does have a fair use
doctrine, which is a broad doctrine,

{{00:02:33}}
exempting certain uses of copyrighted
content by third parties, especially

{{00:02:38}}
when those uses are transformative.

{{00:02:41}}
If the developer of these machines,
ChatGPT or similar AI generated tools,

{{00:02:48}}
if they are successful in convincing
judges and courts that the use of data for

{{00:02:55}}
training the machines is transformative
enough in the US, they might be exempted

{{00:03:00}}
from a corporate infringement finding.

{{00:03:07}}
Nowadays, these generative AI
tools can create content without a

{{00:03:12}}
role being played by human beings.

{{00:03:13}}
If they, the use of such
content is transformative.

{{00:03:18}}
So if, especially if the output created by
the machine is not so similar to the data

{{00:03:27}}
which have been used to trade the machine,
they might escape a finding of corporate

{{00:03:31}}
infringement, especially under U.S.

{{00:03:33}}
law.

{{00:03:33}}
In Europe and in the U.

{{00:03:35}}
K there is more, there are more risks
of infringing copyright of third parties

{{00:03:42}}
because in Europe and the U.K., we do
not have so broad fair use doctrine as U.

{{00:03:52}}
S.

{{00:03:52}}
law allows and permits.

{{00:03:54}}
And it is not a surprise, it does not
come as a surprise, that in the U.S.

{{00:04:00}}
tech giants, such as Google, Facebook,
Twitter, and many others, have   been

{{00:04:07}}
successful because they can rely
on a quite broad fair use doctrine.

{{00:04:19}}
What has been discussed in recent cases
before several patent offices, is whether

{{00:04:26}}
a machine can be considered an inventor.

{{00:04:29}}
There has been a case which has been,
which has attracted media attention.

{{00:04:33}}
I'm talking about the DABUS case.

{{00:04:35}}
DABUS is a machine which has been
developed by several, by a team

{{00:04:38}}
of researchers and scientists, and
such machine is capable of producing

{{00:04:42}}
and generating new inventions.

{{00:04:44}}
So what this team has done is to
file several patent applications,

{{00:04:50}}
covering two inventions in particular.

{{00:04:53}}
Designating in the paperwork,
designating as inventor the machine.

{{00:04:58}}
So it was a provocatively move on the
part of this team, pushing the argument

{{00:05:04}}
that machines can be and should be
considered as inventors and patents

{{00:05:10}}
should be granted in these specific cases.

{{00:05:12}}
Well, unfortunately for the team,
most of the patent offices where these

{{00:05:16}}
applications have been filed have
refused to grant the patent saying,

{{00:05:21}}
look, inventors should be human beings.

{{00:05:25}}
At least under current patent laws,
which were devised many decades ago

{{00:05:33}}
when artificial intelligence was not
so popular, they could not grant the

{{00:05:38}}
patent, but in several of these cases,
judges have also said, okay, I'm not

{{00:05:44}}
going to give you the patent, but we
are aware of the debate around AI being

{{00:05:49}}
capable of producing new inventions,
and therefore now the task is for the

{{00:05:55}}
policymakers, the parliaments, to verify
whether it is time to change current

{{00:06:01}}
patent laws to accommodate AI innovation.

{{00:06:10}}
Governments have already started
consultations with several stakeholders:

{{00:06:15}}
AI people, AI organizations.

{{00:06:18}}
They are basically in the
process of collecting opinions.

{{00:06:22}}
ideas, proposals, suggestions,
recommendations, which will

{{00:06:27}}
then be evaluated, assessed by,
by the lawmaker itself, right?

{{00:06:31}}
In order to verify whether it
is time really to change current

{{00:06:36}}
patent laws, at the moment it is
difficult to foresee if and when

{{00:06:42}}
such policy changes will materialise.

{{00:06:46}}
From how fast this AI revolution is
going, judging from how, how fast

{{00:06:52}}
this, this technology is developing
in terms of generating new content,

{{00:06:57}}
new content which resembles more and
more the content produced by humans.

{{00:07:01}}
Sooner or later, policymakers,
they will need to make a move

{{00:07:06}}
and to modify these, these laws.

{{00:07:10}}
One thing that must be taken into
account, I think, is the need to

{{00:07:15}}
guarantee a balance, a balance
between two different needs.

{{00:07:19}}
On the one hand, the need of AI people,
AI organisation to be incentivised, right?

{{00:07:25}}
Because they might not like the idea
of the final output generated by this

{{00:07:31}}
machine not being protected, right?

{{00:07:34}}
And I understand their concerns.

{{00:07:36}}
They invest money, time, resources
to come up with these machines,

{{00:07:42}}
with these,  technologies
capable of producing new output.

{{00:07:46}}
What they ask is protection for the final
output, either with a patent or with

{{00:07:50}}
copyright and understand their contents.

{{00:07:53}}
But on the other hand, we also need
to take into account the role of human

{{00:07:57}}
creators, human artists, musicians,
journalists, novelists, and inventors.

{{00:08:04}}
If we give too much protection to
output created by the machines, we might

{{00:08:13}}
indirectly disincentivise and discourage
human creativity and human ingenuity.

argument that machines can be and
should be considered as inventors

Intellectual property laws govern the
creations of the mind, inventions, so it's

a set of law which protects and rewards a
human ingenuity, human creativity, as well

Well, AI, artificial intelligence,
machine learning poses a series of

In particular when it comes to
generative AI, these machine learning

systems, what they do is to be fed
with a huge amount of data, huge amount

So one of the first challenges,
especially under copyright law, is this

Does it amount to potentially
copyright infringement?

Because these machines  are fed,
are trained with content which

And the use of this content, of this
information, is an expressive use in the

sense that the machine uses such data
in order to process it and then create a

final output, a final output which might
be potentially in competition with the

data with which the machine has been fed.

What they can do is to
bring legal actions.

This is the legal tool provided, offered
by the system, by the law,  to start

Claiming that such use of data is
infringing their copyright, but

whether these actions are going to
be successful is not so certain.

What the developers of machine learning,
of these generative AI machines,

what they could claim back is that
the use of such content is fair use.

copyright law does have a fair use
doctrine, which is a broad doctrine,

exempting certain uses of copyrighted
content by third parties, especially

If the developer of these machines,
ChatGPT or similar AI generated tools,

if they are successful in convincing
judges and courts that the use of data for

training the machines is transformative
enough in the US, they might be exempted

Nowadays, these generative AI
tools can create content without a

If they, the use of such
content is transformative.

So if, especially if the output created by
the machine is not so similar to the data

which have been used to trade the machine,
they might escape a finding of corporate

K there is more, there are more risks
of infringing copyright of third parties

because in Europe and the U.K., we do
not have so broad fair use doctrine as U.

And it is not a surprise, it does not
come as a surprise, that in the U.S.

tech giants, such as Google, Facebook,
Twitter, and many others, have   been

successful because they can rely
on a quite broad fair use doctrine.

What has been discussed in recent cases
before several patent offices, is whether

There has been a case which has been,
which has attracted media attention.

DABUS is a machine which has been
developed by several, by a team

of researchers and scientists, and
such machine is capable of producing

So what this team has done is to
file several patent applications,

Designating in the paperwork,
designating as inventor the machine.

So it was a provocatively move on the
part of this team, pushing the argument

that machines can be and should be
considered as inventors and patents

should be granted in these specific cases.

Well, unfortunately for the team,
most of the patent offices where these

applications have been filed have
refused to grant the patent saying,

At least under current patent laws,
which were devised many decades ago

when artificial intelligence was not
so popular, they could not grant the

patent, but in several of these cases,
judges have also said, okay, I'm not

going to give you the patent, but we
are aware of the debate around AI being

capable of producing new inventions,
and therefore now the task is for the

policymakers, the parliaments, to verify
whether it is time to change current

patent laws to accommodate AI innovation.

Governments have already started
consultations with several stakeholders:

They are basically in the
process of collecting opinions.

ideas, proposals, suggestions,
recommendations, which will

then be evaluated, assessed by,
by the lawmaker itself, right?

In order to verify whether it
is time really to change current

patent laws, at the moment it is
difficult to foresee if and when

From how fast this AI revolution is
going, judging from how, how fast

this, this technology is developing
in terms of generating new content,

new content which resembles more and
more the content produced by humans.

Sooner or later, policymakers,
they will need to make a move

One thing that must be taken into
account, I think, is the need to

guarantee a balance, a balance
between two different needs.

On the one hand, the need of AI people,
AI organisation to be incentivised, right?

Because they might not like the idea
of the final output generated by this

They invest money, time, resources
to come up with these machines,

with these,  technologies
capable of producing new output.

What they ask is protection for the final
output, either with a patent or with

Expert Insight

Intellectual property rights and copyright laws were first conceptualised during the 1700s as a way of protecting invention and inventors. Fast forward to 2024 and generative AI is everywhere - both consuming content and producing new content. So, is it time to change the law to recognise this shift?
Enrico Bonadio, Reader in Intellectual Property Law at University of London,  explores the seesaw debate on whether the law needs to accommodate for AI, or if invention laws should be reserved for human ingenuity.

Artificial Intelligence

Patents

IP & Competition Law

AI Standards

AI-ML

Machine Learning and AI

AI automation

thumbnail_ENRICO BONADIO.jpg

xsmall_ENRICO BONADIO.jpg

ENRICO BONADIO.jpg

Enrico Bonadio

Reader

Before diving into the deep end of the debate, first we must understand the challenges surrounding AI with intellectual property. 


Primary challenges of making intellectual property with AI

Different countries have different laws that define the parameters of copyright infringement. Bonadio talks us through how the USA's fair use doctrine might be more lenient than European laws, and shines light on the DABUS AI copyright case.

What counts as infringement?

Judging by the speed that AI is developing, some believe lawmakers need to make changes towards current copyright and intellectual property right laws now. Bonadio raises the argument that there must be a balance between the protection of human creativity and the encouragement of AI inventions.
