Whether we are for or against the current meteoric development of artificial intelligence, we must acknowledge that we have hardly any choice but to live with it. Collectively, we are committed to it, and so it is collectively that we must face its threats, and also learn to master its opportunities. This involves in particular promoting, in workplaces especially, communities for sharing experience and mutual help, so as to leave no one by the roadside. Along this path, three complementary tasks await us : learning AI, learning with AI, learning from AI.
Learning AI
It means first learning its imperfections, and there are many : insufficient databases, opaque algorithms, inability to tell authentic information from false news (fake news), and of course the famous “hallucinations”. This is the name given to false answers presented as certain fact by the AI. That is why it is important to always check the answers given by a generative AI, which Google, incidentally, advises for its generative AI Gemini. But it is also important, when faced with an AI that is wrong, to point this out to it and to ask it why, in its view, it was wrong. Such an attitude might be considered impolite with a human interlocutor, but our relations with machines must not be governed by rules of politeness that concern only our relations with our fellow humans. The AI always answers, and it is sometimes very interesting.
It is also essential to familiarise ourselves with the different forms of malicious use and disinformation, such as mass surveillance and cybercrime.
The third obligation we face is to learn about frugal AI, which aims to develop AIs capable of “doing more with less”, that is, which consume less energy, and so have a smaller ecological footprint, a smaller carbon footprint, and emit fewer greenhouse gases. But frugal AI also engages all users in less energy-consuming practices, notably by avoiding unnecessary emails or systematic copying.
Finally, the fourth thing every citizen must learn about AI concerns the various forms of cyber hacking. This word covers acts of digital malice that consist in exploiting one or more vulnerabilities of a system in order to bypass its protections and gain access to it. The aims are espionage, theft of sensitive data that the institution is then offered the chance to buy back, manipulation of information or even sabotage. These practices, often organised from dictatorships, exploit technical and organisational weaknesses, but human ones too. Hackers increasingly use the private digital tools of the employees of targeted companies to get round their defences. That is why educating the general public about these practices is essential.
Learning with AI
AI is going to upend our professional practices in every field : information (searching for information, learning to use the various generative AIs) ; communication (generating emails, text messages) and intervention practices (creating new opportunities for support through shared professional practices).
But at the same time, we risk delegating to these technologies our capacities for expression, memory and imagination, with the risk of no longer learning to do certain things, or even of “unlearning” things we already know.
The risk exists in particular for people who are not comfortable with writing or who do not master a language properly. They will be tempted to relieve the anxiety of the blank page by passing the question they were asked on to an AI. This raises the question of the school’s responsibility: the importance of learning logical links (conjunctions such as but, or, and, so, yet, nor, for) and of learning to work with AI without asking it to do everything for us.
We will have to learn to prioritise and pay attention to what comes from us. And this involves valuing the esteem each person has for themselves. It is the condition for us to be able to use AIs to help us write, not to replace us, which means going back and forth between the AI’s work and our own.
Learning from AI
It means learning, with its help, to carry out transformations from one register of expression into another, that is, things the human mind is capable of but that our culture does not encourage us to do. This is what multimodal generative AIs make possible today. They are able to generate a great variety of content from varied data, for example creating images from text, text from images, and soon images from sounds or the reverse. With these AIs, in which technology meets creativity, we discover unlimited possibilities for innovation. They can help us cultivate our visual intelligence, forms of non-linear narrative, and diversify points of view on the same subject. Already, it is possible to add to a piece of work a summary in the form of a comic strip with Midjourney or DALL-E, or even a short film using DeepMind.
In conclusion,
We have no other choice but to learn to master AI, but this can only be done collectively. It is particularly important, in the family and at work, to make ourselves available for reciprocal intergenerational transmission: the youngest can teach the oldest. And for that, we should organise debates and meetings in companies and cities. And for that, during the introduction of digital technologies and throughout their use, it is important to set up collectives and communities for mutual help and sharing of experience. There is indeed no better protection against the misuses and pitfalls of AI than collective intelligence.
