Canadian Heritage Committee on April 20th, 2023

[Kent Walker]

Let me look into that. The only constraint I can see is under SEC, the United States Security Exchange Commission rules with regard to disclosure. We traditionally don't do a lot of individual breakouts of sources of revenue, but it may well be that the information is public through the Canadian equivalent of Freedom of Information Act requests and the like, so we will look into it.

[Martin Shields, Conservative, Bow River, AB]

Thank you.

The U.K., EU, Taiwan, U.S. and Canada have all been involved in the discussions and are dealing with Google on this All those names have been brought up today.

Are there any other countries you're dealing with on this type of issue?

[Richard Gingras]

I couldn't give you the precise number. It's many more than we've mentioned today. It's very much a part of my role.

We have worked with the ecosystems and are working with policy-makers in countries throughout the world, in every region of the world, and we'll continue to do so.

[Martin Shields, Conservative, Bow River, AB]

Could you give us the names of the countries that you're dealing with around this policy? If we have policy, we'd like to know what the policies are of the other countries you're dealing with as well.

Is that possible? Great. Thank you.

[Marilyn Gladu, Conservative, Sarnia—Lambton, ON]

All right. I just want to help my colleague out.

I had a conversation with a local media journalist who was at a convention, who called me up and said that the federal government spent $1.7 million last year on advertising with Google and only $140,000 with local media. They were making the point that if the government wanted to help local media, it should advertise more there.

I have a question about a statement that you made. You said that when you did the test, news content was less than 2% of the total volume of information shared on Google.

Does that mean that 98% of the content that Canadians were viewing was not impacted by this test?

[Richard Gingras]

To be precise, I was referring to 2% of search queries. However, to your point, yes, I think it would generally affirm that 98% of the queries on the part of Canadians—as I mentioned, from academia to government information, restaurant information and so on—were certainly available to them, and we saw no changes—

[Marilyn Gladu, Conservative, Sarnia—Lambton, ON]

Mr. Walker, I want to ask about the transparency report. Can you describe to me what the purpose is of the different categories that you're tracking there?

[Kent Walker]

We are trying to provide visibility into the nature of government requests for personal information, as well as for removal of content. We think it's an important part of the democratic process that people be aware of requests for removals, to the extent permitted by law.

There may be legitimate criminal or national security reasons for some of that information not to be public, but whenever it is legitimately public, we want to do our part in making that more accessible.

[Marilyn Gladu, Conservative, Sarnia—Lambton, ON]

How does Canada measure up against the rest of the world in terms of the frequency of government requests to take down content?

[Kent Walker]

I honestly haven't looked into that, but the report has an area where you can break that down in a spreadsheet and do a fairly easy comparison.

[Marilyn Gladu, Conservative, Sarnia—Lambton, ON]

Okay. Very good.

Rachael, do you have any final comment?

[The Vice-Chair, Conservative, Kevin Waugh]

Okay. You're all good on this side.

We'll extend the Liberal side a minute or two.

Ms. Hepfner, the floor is yours for five-plus minutes.

[Lisa Hepfner, Liberal, Hamilton Mountain, ON]

Thank you, Chair. I will happily take those five-plus minutes.

I want to go back to the comment that you made earlier about your difficulty in accessing a particular news article. I just want to make the comment that when Google uses tactics like blocking Canadians' access to news sources and information, it fails to be a reliable service for consumers. I guess it's not surprising that we hear reports that companies like Samsung are considering dropping Google as the default search engine on their phones.

I wanted to make that comment and then turn back to my questions about artificial intelligence, its ability to and the likelihood that it will upend linking entirely. We've seen this with some of your competitors. They include artificial intelligence in their search engines. In some cases, the language model tools scrape data and content from behind paywalls.

We heard this from your CEO. He seemed to hint at it when the Wall Street Journal asked him if he sees link-based search as the dominant way people access information on the Internet in the future. His answer was, “I think the experience will evolve substantively over the next decade.”

Mr. Walker, please tell us whether you will be following in the footsteps of some of your competitors and getting rid of links altogether?

[Kent Walker]

I don't believe we have plans to get rid of links altogether. As I said, as we were talking earlier, we believe that links directly to publisher content are an important part of the ecosystem. They're a complementary part because, again, these generative models are very creative, very interesting and very useful in a lot of different ways, but they're not necessarily completely reliable in every case....

It's an unsolved research problem—

[Lisa Hepfner, Liberal, Hamilton Mountain, ON]

That's a good point, thank you.

We know that links are just one way that news is accessed. If Google starts pivoting to AI, describe how users would access news.

[Kent Walker]

All of this is evolving quickly, so it's hard to have a clear, long-term perspective. For the relevant near term, my guess for the immediate future is that we will continue to look at a blend of traditional experiences with links going directly to publishers and advertisers, and it will be complemented by some of the new AI tools you're seeing. We will be thinking about new ways to deliver traffic to various original sources and the like.

For example, if you did a search on fun things to do in Ottawa during the day, you might get a generative AI list of topics, but each of them might have a link to a place to buy tickets, to get a tour, to get tickets to a museum or whatever else that might be.

You can see the model evolving over time. I think we and the entire industry are still working out what that might look like.

[Lisa Hepfner, Liberal, Hamilton Mountain, ON]

That's an interesting point. When you're talking about how Bard develops and trains, is it using news content in its training?

[Kent Walker]

Most large language models largely trained on the open web. That's true of Canadian researchers at the University of Toronto, commercial competitors or other academic models out there. It's difficult to know exactly what was in or what was out, but the training is very broad.

[Lisa Hepfner, Liberal, Hamilton Mountain, ON]

I imagine the Bard model can use news content with no attribution, and it's capable of writing convincing articles, but we don't know how accurate it is.

[Kent Walker]

As I said, we are continuing to research the science and engineering of how to increase accuracy and reliability. In a funny way, it's not directly using the content to return that content to the user. It's using it to determine the weights of billions, in some cases trillions, of parameters that allow it to make more accurate estimates as to what the next words in a response should look like.

[Lisa Hepfner, Liberal, Hamilton Mountain, ON]

I'm not sure if I understand that entirely.

Tell me whether you think Bard could replace journalists, for example.

[Richard Gingras]

I'd be pleased to—

[Lisa Hepfner, Liberal, Hamilton Mountain, ON]

No, please, I'd like to hear from Mr. Walker. Please continue.

[Kent Walker]

It's with a caveat that Richard might give you a better answer.

No, we think of this as a tool that will enable journalists to do a wider variety of things. It may make it simpler to do the rote tasks of journalism in the same way it makes it simpler to do the rote tasks of computer programming, of being a lawyer or of being a doctor. I think we all need to be thinking about how we could have a labour force evolution so that people are more comfortable using these tools to make themselves more productive and to provide more value. It could be extraordinarily powerful.

[Lisa Hepfner, Liberal, Hamilton Mountain, ON]

A labour force evolution...that's interesting. Thank you.

I've heard from you that the news sector needs to innovate, and it needs to do a better job of monetization. I can tell you, in 20-plus years in journalism, it has never made money. It's never been about making money. It's a public service. Here in Canada, we believe it's essential to democracy, and that's why we support it with government regulations.

Why would news organizations go to the trouble and the expense of digitizing everything and putting their content, which costs a lot of money and takes a lot of talent to produce, on a search engine that could just arbitrarily, by a company decision, whip away their ability to share their content?