Evidence of meeting #5 for Canadian Heritage in the 45th Parliament, 1st session. (The original version is on Parliament’s site, as are the minutes.) The winning word was content.

A recording is available from Parliament.

On the agenda

Members speaking

Before the committee

Desrochers  Executive Director, Coalition for the Diversity of Cultural Expressions
Raab  Assistant General Manager, Copibec
Beaulieu  Advisor, Legal and Public Affairs, Copibec
Anctil  Affiliated Researcher, International Observatory on the Societal Impacts of AI and Digital Technologies
Griffiths  Publisher, The Hub

David Myles Liberal Fredericton—Oromocto, NB

Are there any cases before Canadian courts that illustrate the types of legal issues Canadian artists face with regard to generative artificial intelligence? If so, what are the key issues?

4:15 p.m.

Advisor, Legal and Public Affairs, Copibec

Maryse Beaulieu

Some cases are currently before the courts.

Essentially, but not exclusively, these are class-action suits. They are still in their infancy. They are a special form of lawsuit and follow a specific process: A single member of a group representing a large category of claimants can ultimately obtain a judgment on behalf of a large number of people. It is currently one of the vehicles that seems to be favoured, particularly by authors, since it is mainly authors who initiate class action lawsuits.

Sometimes, this involves cases where pirated libraries were used for training. In other cases, other methods were used.

That said, there is still no case law on how the law will ultimately be interpreted. Earlier, I mentioned litigation in the courts, which is a way of finding out what judges will say on highly novel and complex issues.

I would like to think that the group before me today, that is to say, people who have the power to legislate, will be able to ensure that the fundamental questions I raised earlier are respected, particularly with regard to transparency. I could elaborate on this subject, because transparency is very important. For example, not knowing which works were used to train models is, in a way, a form of denial of rights, since we do not know what was used. Transparency therefore allows us to know which protected work was used.

I would also like to take the opportunity—

The Chair Liberal Lisa Hepfner

I am sorry, but your time is up. You will certainly have the chance to speak again.

We will now give the floor to Mr. Champoux for six minutes.

Martin Champoux Bloc Drummond, QC

Thank you, Madam Chair. I find it annoying that our time is limited because we could listen to our speakers for a long time.

Mr. Anctil, this committee certainly was legislating on digital platforms 20 years too late when it reviewed the broadcasting bills, among other things.

In your opinion, how many years behind are we currently in regulating artificial intelligence?

4:20 p.m.

Affiliated Researcher, International Observatory on the Societal Impacts of AI and Digital Technologies

Dave Anctil

Actually, it's not that clear-cut. In the early 2020s, we were still uncertain about the capabilities that artificial intelligence systems could develop. A few ingredients were missing for them to be reliable and able to create results that would have value, commercial or otherwise.

However, by 2023, it was clear that these systems used multiple modes, not just text. Initially, there was a lot of talk about large language models, but today we are talking about multimodal artificial intelligence systems that are capable of using any mode that relies on digital information, including music, voice, video and images. These systems are enormous. In fact, their size is unknown, since there is no longer any scientific transparency in the publications of the industry's major players. We do not know what data is being used to train them. We also do not know which algorithms have been used.

We do know that these systems will continue to improve at a dizzying rate.

Martin Champoux Bloc Drummond, QC

Do you think that content creators and artificial intelligence can still coexist? Do you think it is possible to stop fearing artificial intelligence and come to see it as a tool that can be useful to content creators?

I get the impression that we are being very reactive, and perhaps even more so in this committee, which is very focused on this type of issue.

How can artificial intelligence be used to create a tool that respects copyright and helps artists develop new markets and new content? Do you think this can still be achieved with appropriate regulation?

4:20 p.m.

Affiliated Researcher, International Observatory on the Societal Impacts of AI and Digital Technologies

Dave Anctil

Yes, there is absolutely no problem with artists using artificial intelligence tools in the same way as any other materials or tools that can contribute to cultural evolution and creativity.

The problem is control of the digital infrastructure needed to do this, as well as the scale of cultural production. It is extremely important to understand the difference between the scenario of a possible alliance between Nvidia and Disney to create new Star Wars series using artificial intelligence, and the scenario in which cultural artisans use artificial intelligence to accelerate and increase productivity and explore new creative fields.

That is where I have some concerns. Artificial intelligence will offer enormous advantages in several sectors. This is already the case in science, education and several professions where the actions and civil liability of the people involved remain human.

Unfortunately, it is now possible to create productions that claim to be cultural without any possibility of traceability or accountability for the individuals or organizations behind them. That is the difference.

Martin Champoux Bloc Drummond, QC

What amazes people in all the possibilities you describe is also what terrifies others, including me. I find the possibility of artificial intelligence becoming quasi-autonomous absolutely terrifying.

Let us look at the more technical side. Currently, there is generative artificial intelligence. However, super intelligence is under development, which is of great concern to many experts: This is artificial intelligence that would be virtually autonomous and self-sustaining. If we reach this point without appropriate regulations or guidelines, could it spell the end of artistic creation as we know it?

4:25 p.m.

Affiliated Researcher, International Observatory on the Societal Impacts of AI and Digital Technologies

Dave Anctil

If superintelligence ever becomes a reality, culture will be the least of our worries—

Martin Champoux Bloc Drummond, QC

Indeed, it is humanity that will be more—

4:25 p.m.

Affiliated Researcher, International Observatory on the Societal Impacts of AI and Digital Technologies

Dave Anctil

—and I would prefer to testify before the Standing Committee on National Defence to discuss that issue.

In reality, we do not know. It is obviously something that leading experts, such as Yoshua Bengio at Mila in Montreal, are looking into. It is important to safeguard this.

In the meantime, it is important to understand that there is one step that is certain. Artificial intelligence is becoming increasingly agentive, even multi-agentive. This means that we can create systems with several specialized artificial intelligence models that are capable of creating production chains. Therein lies the problem.

Imagine that a production studio decides to replace all the artisans involved in making a film. Artificial intelligence screenwriters would propose scripts, which would be evaluated by other artificial intelligence systems, which would then propose the first segments of the film. Editing and everything else could be automated.

We already have the capability. Right here at Collège Jean-de-Brébeuf, we produce small video games using five artificial intelligence systems. By providing an initial idea or theme, we can produce a complete video game in just a few hours. That is what lies ahead for us.

Martin Champoux Bloc Drummond, QC

I think I am going to sign up for a new class because time is flying by. I will keep asking you questions and listening to your answers. We will continue this a little later.

Thank you very much, Mr. Anctil.

The Chair Liberal Lisa Hepfner

Thank you very much.

Mr. Généreux, the floor is yours. You have five minutes.

4:25 p.m.

Conservative

Bernard Généreux Conservative Côte-du-Sud—Rivière-du-Loup—Kataskomiq—Témiscouata, QC

Thank you, Madam Chair.

I want to thank the witnesses. This discussion is very interesting and important.

Mr. Anctil, earlier you said two very important words: “pillage” and “steal”.

Imagine that Mr. Myles, a former singer who was once part of a band, is also a whiz at computers and artificial intelligence. He has spent his life making music, and today he creates that same music using artificial intelligence. Would you still consider Mr. Myles to be an artist?

4:25 p.m.

Affiliated Researcher, International Observatory on the Societal Impacts of AI and Digital Technologies

Dave Anctil

In my opinion, if Mr. Myles uses his own data in his artistic process, whether he is editing, has an artistic intention, or is publishing, yes, he is still an artist and a musician. On the other hand, if I make an album based on training data, including Mr. Myles's data, if I hide behind a pseudonym and claim copyright for this production, in my opinion, that is theft.

4:25 p.m.

Conservative

Bernard Généreux Conservative Côte-du-Sud—Rivière-du-Loup—Kataskomiq—Témiscouata, QC

Right, I understand the difference.

I have never played music. I have never written any and I have never sung, fortunately for everyone.

Voices

Oh, oh!

4:30 p.m.

Conservative

Bernard Généreux Conservative Côte-du-Sud—Rivière-du-Loup—Kataskomiq—Témiscouata, QC

I would never claim to be an artist in any way. However, I could become an expert in artificial intelligence and create music myself or a system that would create music on its own. Would you consider me an artist, since I would be creating elements using a tool, artificial intelligence?

4:30 p.m.

Affiliated Researcher, International Observatory on the Societal Impacts of AI and Digital Technologies

Dave Anctil

If I understand correctly, you are talking about someone who would create a system, based on artificial intelligence, that would help artists to create.

4:30 p.m.

Conservative

Bernard Généreux Conservative Côte-du-Sud—Rivière-du-Loup—Kataskomiq—Témiscouata, QC

No, it would be a system that I would create myself. If I, Bernard Généreux, decided to create a work using tools that already exist and are accessible to everyone, and I posted it on the Internet, would you consider me an artist?

4:30 p.m.

Affiliated Researcher, International Observatory on the Societal Impacts of AI and Digital Technologies

4:30 p.m.

Conservative

Bernard Généreux Conservative Côte-du-Sud—Rivière-du-Loup—Kataskomiq—Témiscouata, QC

Okay.

I ask the same question to the representatives of Copibec and the Coalition for the Diversity of Cultural Expressions: Would you consider me an artist? That's a good question, isn't it?

Even though I have never made music, I could decide to do so using artificial intelligence tools. There are young people who do this every day. If they copy some of the works that Mr. Myles created in his lifetime to make an album, I understand that this could be considered theft. However, if I invent something from scratch, using artificial intelligence tools, I am not stealing anything.

In that case, would you consider me to be an author or an artist?

4:30 p.m.

Assistant General Manager, Copibec

Arezki Raab

I will give you the beginning of an answer, then I will let my colleagues take it from there.

From a practical standpoint, I don't think you could be considered an artist, because you would have used tools that are based on artificial intelligence models trained on creations by actual human artists.

For now there is still a grey area around requests sent to artificial intelligence systems.

4:30 p.m.

Conservative

Bernard Généreux Conservative Côte-du-Sud—Rivière-du-Loup—Kataskomiq—Témiscouata, QC

Okay.

There are regulatory texts governing the use of artificial intelligence, particularly in Europe and the United States. In Canada, there are none, since former Bill C‑27, which we have already discussed, died on the Order Paper, unfortunately—or fortunately, depending on your point of view.

I don't know what the Liberals' intentions are, but do you think it is still necessary to pass a bill to create a regulatory framework for this issue in Canada?

4:30 p.m.

Executive Director, Coalition for the Diversity of Cultural Expressions

Marie-Julie Desrochers

Yes, I do.

The Copyright Act works well with respect to artificial intelligence. There are things that could be improved in the act in general, but when it comes to generative artificial intelligence in Canada, the act requires that permission be sought beforehand. We should leave the Copyright Act as it is.

However, when it comes to transparency, legislation is still needed, particularly to include a provision like the one found in the European Union's AI Act that requires people to provide a sufficiently detailed summary of the training data that has been used. We hope that this will happen quickly.