The subject of technological biases created by artificial intelligence is key for any digital business. To prevent inequalities from widening, Numeum is launching Ethical AI. The initiative brings together companies that want to make AI more virtuous. Explanations with Céline Bayle, Director of Enterprise Market Product Marketing France & Export of Sage and member of the Ethical AI steering committee.
All companies that develop artificial intelligence tools face the risk of emerging biases created by technology. AI can indeed bring out results that worsen certain inequalities or set aside certain profiles of people. To avoid these pitfalls, the initiative Ethical AI publish a practical methodological guide intended to bring out proposals and recommendations for each professional.
–
In-depth work carried out by Numeum including Céline Bayle, Director of Enterprise Market Product Marketing France & Export at Sage, and member of the Ethical AI steering committee, explains the details.
How to qualify an ethical AI?
An AI system can qualify as ethical if it is able to preserve fundamental human rights throughout its lifecycle. We are talking here about the rights to dignity, mental and physical integrity, liberty, autonomy, fair treatment, privacy and privacy. It is important to keep in mind the fact that Man remains solely responsible for the systems he creates.
Artificial intelligence must also be respectful of personal data. I would add that it must be fair. This means that the system must be careful not to create or reproduce discrimination. In this sense, it will promote inclusion. AI must be transparent in its operation. This implies that its conclusions can be explained and that it is auditable. From this perspective, the system is loyal in its relationship with human beings and in control (under human control). Finally, AI must be secure and robust against cyber attacks.
To set up such an ethical AI, we recommend within Ethical AI, to set up, at the highest level, a specific governance aimed at defining and enforcing the company’s AI policy. It is therefore a question of raising awareness for each project while establishing a point of contact for the company on these questions. And that, as well vis-à-vis the interior as the exterior.
Why should companies join Ethical AI?
Ethical AI makes it possible to give a positive DNA, a virtue. It is about asking the right questions around projects in order to apply the sieve of ethical AI. The goal is therefore to break your own habits to exceed your expectations. It is virtuous to “put down your pencil” and to think about the impacts that your tool can have. The process can seem complex insofar as the objects very often have an aim oriented mainly towards the business. But adding this new filter gives rise to different thoughts. The very nature of reflections changes in favor of more ethics.
It is therefore a question of passing each of the methods through these ethical sieves. This allows everyone to explore the gains that this brings to their business and to their own customers. To achieve this result, it is still necessary to evangelize the market. Everyone should understand that theartificial intelligence does only what is asked of it. Ethics are therefore taken into account as soon as the code is implemented. The Ethical AI initiative therefore aims to prevent vicious circles from setting up in the economy. It is up to humans to lead the processes towards virtue. Each actor must therefore be made aware so that he can then have a significant impact.
How to make people understand the interest of developing ethical AI?
One of the ways to spread awareness of the importance of developing ethical AI is to organize hackatons. We have done several that have sparked important reflections on the subject of artificial intelligence lifecycles. The feedback was generally positive and made it possible to answer new questions. In this sense, the Ethical AI initiative helps everyone understand that they can have positive impacts on society.
This gives digital a part of societal DNA. Each engineer and developer is therefore justified in asking how to no longer reproduce socially unacceptable biases. The technology must, for example, make it possible to combine profitability and environmental sobriety. Until recently these new parameters were not taken into account. But mentalities are changing in the right direction.
–
How to “put people” in AI?
Artificial intelligence puts the notion of impact back at the center of discussions. In recent years, robotization or digital technology have been deployed without, however, measuring the impact on society. We believe AI is measurable. That it is not a black box. It is therefore up to those responsible to impose limits on this technology. This type of process is relatively new insofar as it incorporates parameters such as transparency or even accessibility.
The development of this technology follows a logic similar to that of the GDPR. This text sets out to regulate how we use data, and for what purposes. This has generated a real awareness on the subject. It is in this sense that our manifesto intends to bring these subjects back to the fore. Putting people in should not be seen as a constraint but as an opportunity to come out on top vis-à-vis these competitors. It also allows you to have a detailed reading of your own services and to ask yourself the right questions.
How to implement ethical AI? what is your advice to companies within Ethical AI?
I advise everyone to take real cases of what their own company is already doing in terms of artificial intelligence. The manager can then carry out workshops with internal referents to pass the tools through the sieve of ethical values. It is thus a question of asking the question of the existence of possible biases. I recommend tackling smaller projects first in order to reach firm conclusions quickly. The objective is then to determine whether or not it is necessary to apply a different filter on the subject.
This way of anchoring in the concrete is resolutely positive for all parties concerned. All companies have an interest in focusing on these questions. Within Ethical AI, we do not oblige partners in any way in terms of constraints. We try to infuse a healthy word that carries values. The objective is, in fine, that everyone recognizes these ethical values. It is our dearest wish for the world of tomorrow.
More information on the initiative site Ethical AI.
–
