Human Resources Committee on Nov. 22nd, 2023

[Yana Lukasheh]

I'll answer that question in terms of what is probably relevant for SAP and what can be given as a response for SAP.

Larger companies, when it comes to data and AI and machine learning, or large language models, have been quite deliberate in the way they design many of those algorithms, so that they are reliable, safe and responsible. We have many strategies and ethics that go behind it and that are in place. I will start with that.

[Michael Coteau, Liberal, Don Valley East, ON]

I would like to know about the ethics piece. How do you make sure the ethics piece is kept if it's really behind the company? The company needs to preserve some of its—

[Yana Lukasheh]

Absolutely.

There are compliances we have to address and abide by, as do other industry members in the different sectors as well. The data that goes behind....SAP does not own that data. It is the customer's data. We provide the technology, the tools, and the customer maintains that data. It's hard for me to answer that question from that perspective, but there's a lot that goes into these tools and how they're used.

The data, depending on the sources it's coming from, yes, has to be validated. It has to be verified to make sure that it does not cause bias and unintended consequences. The developers in our industry are consistently looking at how to improve that technology and how to improve leveraging of the good or clean data, I should say.

[Michael Coteau, Liberal, Don Valley East, ON]

Thank you, I appreciate that.

Thank you, Chair.

[The Chair, Liberal, Bobby Morrissey]

Thank you, Mr. Coteau.

Go ahead, Ms. Chabot. You have six minutes.

[Louise Chabot, Bloc, Thérèse-De Blainville, QC]

Thank you, Mr. Chair.

Thank you to the witnesses for being here.

This is our last day hearing from witnesses on the implications of artificial intelligence for the Canadian labour force, and I'm not sure we've gone as far as we need to. We are actually still missing quite a bit of the information we need to measure the impact.

Mr. Soucy and Ms. Blais, thank you for your input. Some of your fellow union representatives told the committee that it is detrimental to workers when they aren't told ahead of time about the implementation of new technologies like AI or the purpose of those technologies.

You said that AI could even cause upheavals in the working world—hence the importance of regulating AI.

What do you mean by regulating? My Liberal colleague pointed out that not all workers are unionized. How do we regulate AI in a practical and effective way?

[Danick Soucy]

It's important to have clear laws that define the responsibilities around the use of AI. That will ensure that even non-unionized workers are protected.

[Louise Chabot, Bloc, Thérèse-De Blainville, QC]

I gather, then, that more effectively regulating AI could also mean amending labour laws.

Is that correct?

[Danick Soucy]

Precisely. Labour laws and the Labour Code have to be adapted to address new technologies.

[Louise Chabot, Bloc, Thérèse-De Blainville, QC]

A common refrain is that no one is against new technology because it helps society move forward, but it has to work for humans, not the other way around.

Do you have any specific recommendations to support workers as far as privacy, data and workplace health and safety are concerned?

[Danick Soucy]

It's important to see the technologies as tools, not a way to replace humans and the work they do. The technologies absolutely mustn't put the health and safety of workers at risk. Everything has to be laid out clearly, including straightforward and publicly available remedies in case the system fails.

[Nathalie Blais]

It's really important to educate the public and companies. The government is really pushing AI. In the fall, Montreal hosted an AI event called All In 2023. Minister Champagne and Prime Minister Trudeau were there. There's a big appetite in the government for AI. There's a push to move in that direction, but it's also important to keep in mind that public literacy may not be at the level it should. When interacting with AI or when faced with the collection of certain data, people may not understand they need to be mindful and take appropriate precautions.

The first step is education. Next is making sure that not only labour laws, but also specific AI legislation is responsive to this reality. The proposed artificial intelligence and data act, currently being studied by the Standing Committee on Industry and Technology, could include guidelines to ensure that the use of AI does not infringe on workers' fundamental rights or jeopardize their health and safety.

[Louise Chabot, Bloc, Thérèse-De Blainville, QC]

ChatGPT has revolutionized the online world. Some are calling for a moratorium on AI technology before things go any further, precisely to educate the public and employers. I heard the same thing in the wake of that big conference in Montreal.

Is it absolutely necessary, in your view, to do the work on the front end before continuing down this path?

[Danick Soucy]

Yes, we have to make sure the law is clearly defined before we go full steam ahead with AI. Otherwise, workers' rights could be violated until the legislation comes into force.

[Nathalie Blais]

Certainly, it's tough to take action on the front end, before AI is widely implemented. I take your point, Ms. Chabot, but the train has already left the station. Something has to be done to put things on hold if the idea is really to do the work beforehand.

Some companies still haven't adopted AI systems, so in their case, it would be possible to take action on the front end. I don't see how we can stop a train that's already coming down the track.

[Louise Chabot, Bloc, Thérèse-De Blainville, QC]

I know your union has a lot workers in the telecommunications sector. Does implementing AI technologies pose any specific risks in that sector or other sectors? Can you give us any actual examples?

[Nathalie Blais]

In the communications sector, more broadly, a closed captioning company comes to mind. We found out purely by chance from an employee in our union that the company was introducing AI.

What happened is that the employee was asked to revise some captioning that had been done. She thought she was revising a contractor's work. She was asked twice more to revise texts, and each text was better written and better overall than the time before. She eventually realized that, unbeknownst to her, she was training an AI system.

That's a good example of the problems associated with AI—companies are not being transparent. We are also seeing a lot of jobs being moved.

[The Chair, Liberal, Bobby Morrissey]

Thank you, Ms. Blais and Ms. Chabot.

We have Madame Zarrillo for six minutes, please.

[Bonita Zarrillo, NDP, Port Moody—Coquitlam, BC]

Thank you so much.

It's so interesting to hear the testimony today, and there's just so much that we still need to learn and know.

I wanted to go to Mr. Kiron. You spoke about the potential for dehumanizing workers, and I'm interested in exploring this a little bit. I wonder if you could share what risk factors there are that would contribute to dehumanization in regard to AI.

[Dr. David Kiron]

Sure.

One of the big threats around dehumanization comes with surveillance in the workplace. There are AI technologies that enable business owners and business managers to track very specifically what is happening with workers. It's super-constrained. It's not only keystrokes, but it's whether or not you're being attentive, whether you're focused on a screen or what your biosignals are. You have that level of intrusion, and the job could be the only kind of job you could get, or you might even feel lucky. But the actual performance of the job is under so many different tabs that it's like you're a machine, being played by the manager elsewhere. That's one example.

[Bonita Zarrillo, NDP, Port Moody—Coquitlam, BC]

We're thinking about that, because we're federal regulators here and we want to make sure that we're protecting workers and the human factors of people in our community. Can you share with the committee your thoughts about what we need to do as regulators to protect from that dehumanizing risk?

[Dr. David Kiron]

When I go to my doctor and he asks what my problem is, I say, “Well, my shoulder hurts.” He says, “When does it hurt?”, and I say, “When I do this.” Then he says, “Well, don't do that.” Regulations can be very targeted in saying that there are certain types of control in the workplace that are just unacceptable. They constitute a dehumanizing effect on workers [Technical difficulty—Editor].

[Bonita Zarrillo, NDP, Port Moody—Coquitlam, BC]

It looks like we lost that witness.

I'm going to move over to Madam Lukasheh on the information around disability. This committee also considers persons with disabilities and their inclusion. I just wonder if you wouldn't mind expanding a little bit on opportunities to have AI make the workplace more equitable and inclusive, and also any risks that you see in that space.

[Yana Lukasheh]

Absolutely. We certainly see the opportunity for software applications that leverage AI tools to be made available to members of the disability community across the spectrum. The applications that we use are accessible and compliant, and they will allow different members to play a role in the workforce, depending again on the user needs and user experience.

Whether it is an individual who is visually impaired...there are ways that AI can be worked into a software application to allow them to still participate in that workload. I'm happy to provide more details on this particular business council, but also about how SAP more broadly thinks about disability inclusion, which is factored into the design phase of our applications.

Stemming from our global CEO's office, we have a full expertise in how to make sure that our applications are disability and accessibility friendly. That goes across the board for all of our applications that are used right now.

[Bonita Zarrillo, NDP, Port Moody—Coquitlam, BC]

Thank you for offering additional information. To have the makeup of that council and how it works would be great.

In testimony at the committee for this study, it's been suggested that there should be an advisory council for the rolling out of regulation as they relate to workers. I wonder whether you could share what you think would be important representation around any kind of an advisory council, federally, that would look at the implementation of AI in the labour market.