Accessible since November 2022, ChatGPT has become the symbol of a new form of Artificial Intelligence, generative AI (GAI). A new universe of possibilities opens up to everyone. It is no longer enough, like on a search engine, to ask a question to get answers. With ChatGPT, we converse and content is created. From an elaborate question – a “prompt” – we obtain images, videos, music, code, slides. Welcome to the world of multi-modality, personalized assistants and automation available to everyone.
Huge progress… but not without risk
For individuals, businesses and governments, the opportunities are immense: saving time, discovering a new form of creativity, helping with data analysis, making more rational decisions, optimizing information in real time, faster innovation, testing of new ideas, but also the ability to create drugs, apply personalized medicine, learn differently with specific exercises for each person, and automatically create communication campaigns.
At the same time, the risks are serious and numerous. First of all, false information (deepfakes) can be created much more easily and it will become increasingly complicated to distinguish true from false. This risk of mind manipulation already existed. With AGI, it is increased tenfold. Let us then point out the ecological risk.
For Luc Julia, the creator of Siri, if we use AGI tomorrow as much as search engines today, the consumption of energy and water used for the creation of language models and during queries will generate a unsustainable situation. Let us not forget, finally, the danger posed by the opacity of algorithms, through data – in recruitment, for example – or the risk of disappearance of many professions, the surveillance of citizens.
European AI Act
States are aware of the risks. On November 1 and 2, the first international summit on the regulation of AI took place in Bletchley (United Kingdom). A few days earlier, on October 30, the most advanced economies (Germany, Canada, United States, France, Italy, Japan, United Kingdom and the European Union) adopted a code of good conduct on AI, in addition to the European AI Act (December 2023).
And you ? If you decide to test chatGPT and AGI, be careful not to make any mistakes! The first would be to believe that ChatGPT is easy to use. It is not enough to ask him a question, but then not at all.
To obtain a qualitative response, you must learn to configure the “personalized instructions” and follow the steps of a good “prompt”. Don’t underestimate how long it can take. The other fatal mistake would be to trust AI too much. ChatGPT lies, hallucinates and cites false references. Always check what he says. In short, to explore the IAG galaxy, travel in a group and keep a critical mind!
This article is originally published on pourleco.com
