By clicking “Accept”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information.
March 24, 2020

Newsletter - March 2020 - AI for Humanity: for an artificial intelligence at the service of humanity

March 24, 2020

Artificial intelligence (AI) is more than ever at the heart of the game when it comes to innovation. And the phenomenon is growing: according to the IDC, global spending on AI systems is expected to reach $97.9 billion in 2023 compared to $37.5 billion in 2019, with an annual growth rate of nearly 30%. This is a fundamental trend, a silent revolution that affects all sectors. If transport and telecommunications are doing well, the mapping carried out by France digitale of the 432 AI start-ups in France shows the vast diversity of applications: robotics, marketing, HR, adtech, fintech, agriculture, health, big data, commerce... Even fields that were thought to be exclusively reserved for human creativity such as writing, painting or music are no longer out of reach for AI.

Schizophrenia

This apparent omnipotence elicits contrasting reactions from most of us: the hopes raised by artificial intelligence are matched only by the concerns it engenders about our place, role and future as humans. On the one hand the promise of early detection of disease, the end of road deaths, optimised services, increased innovation; on the other the fear of becoming obsolete, of losing control, of a distant if not improbable singularity. All this, fuelled by science fiction and a few dubious experiments that find an echo in the press, like those American researchers who have developed a system for detecting sexual orientation from simple photos of people. It is time to dispassionate the debate and remove the prejudices.


Human after all

For artificial intelligence is not and never will be anything other than human technology. It is not so much the technology that is at issue as the way it is designed and used. We have seen this many times: AI tends to reproduce human prejudices, hidden in the data it is given to practice. For example:

  • as a member of the jury of a beauty pageant, the AI eliminated black candidates first,
  • In the context of predictive justice, the COMPAS software that assesses the risk of recidivism has assigned a risk rate more than twice as high to African-Americans,
  • An algorithm that was supposed to link words has reproduced stereotypes by more readily associating "woman" and "home", or men with scientific professions.

Most often, this is not due to the malicious intent of the creators of these AIs, but to a poor quality, incomplete or badly calibrated dataset. The Institut Montaigne has just released a report on this subject, "Algorithm: Bias Control Please".

Mirror, my beautiful mirror

The AI therefore reflects an image of ourselves and those who designed it. But who are the workers behind AI today? There are of course the young geeks in Silicon Valley, but also and above all there are millions of little hands that annotate images to allow the algorithm to learn. With each click, they leave a bit of their identity, their culture, and sometimes their prejudices. Let's be clear: tomorrow's artificial intelligence cannot be built by a group of white male engineers. Diversity, here more than anywhere else, is the key. At isahit, we are developing a community of women around the world to help meet these challenges. It's not just a question of ethics, but also of precision and efficiency. For example:

  • in the case of facial recognition, a person of a certain ethnicity will find it easier to recognize another person of the same ethnicity,
  • in the food sector, an African will be able to identify more finely the dishes or ingredients specific to African cuisine,
  • The same applies to waste analysis.

This may seem marginal, but they are fundamental keys to the development of a successful AI. Through the diversity of our community, we can guarantee our clients accurate and valid, specific and contextualized data.

Emancipatory power

Further on, having recourse to these populations from developing countries is a powerful lever of economic and social emancipation for the latter. The first to have felt this was the late Leila Janah, founder of Samasource, who left too early last January at the age of 37. Since 2008, she has been connecting these poor populations with large American groups to carry out the basic but no less essential tasks of data entry or image annotation for the development of their AIs. As a result, more than 50,000 people, mainly in Uganda and Kenya, have been able to benefit from the digital revolution while gaining access to a decent income. A much more important lever for development than humanitarian aid. It is this legacy that isahit is trying to continue today with its dynamic, diverse and dynamic community of women entrepreneurs. With our own identity: operating as a platform, horizontal management and greater diversity (26 countries, 3 continents). And with a requirement: that they have a concrete personal or professional project, so that isahit is not an end but a springboard to other activities and so that they can multiply the positive impact of their experience within their communities.

Ethics IA

As you can see, ethics is a central issue for AI. This may be an opportunity for France and Europe to catch up in this sector. But there is no more time to lose. The AI for Humanity report, submitted by Cédric Villani to Emmanuel Macron in 2018, paves the way: "If we wish to develop AI technologies that comply with our values and social standards, we must act now by mobilizing the scientific community, public authorities, industrialists, entrepreneurs and civil society organizations. "And to propose several lines of work:

  • Opening the black boxes, i.e. deciphering how an AI arrives at a result, something that researchers still sometimes find difficult to understand and which can be a source of discrimination and bias.
  • To train researchers, engineers and entrepreneurs in ethical issues while implementing impact studies on the risks of discrimination during the development of AIs, based on the model of the RGPD on data processing.
  • Create an AI ethics committee to shed light on major issues and issue recommendations, even standards.
  • Acting for mix and diversity so that AI does not become "a new machine to exclude".

To action!

The European Union, too, with its new Commission, seems at last to have realised what is at stake and the strategic importance of developing a European AI more in line with our interests and values. Let's get down to business! All over the continent, startups are moving forward in this field and innovating. Isahit, with the strength of his community, is determined to take its full part in this exciting and exciting adventure, at their side.

Isabelle Mashola, CEO of isahit.

Read all about it on Linkedin

Read the forum in your web browser

Read the gallery in PDF

You might also like
this new related posts

Want to scale up your data labeling projects
and do it ethically? 

We have a wide range of solutions and tools that will help you train your algorithms. Click below to learn more!