0_IAGenerative
Home / Dossiers / Science and technology / Generative AI: threat or opportunity?
π Science and technology π Digital

Generative AI: threat or opportunity?

3 episodes
  • 1
    ChatGPT, Midjourney: everything you need to know about generative AI
  • 2
    The ways AI will change the future of work
  • 3
    Demystifying generative AI: true, false, uncertain
Épisode 1/3
On November 21st, 2023
5 min reading time
Eric Moulines
Éric Moulines
Professor of Statistical Learning at École Polytechnique (IP Paris)
Hatim Bourfoune
Hatim Bourfoune
AI research engineer at IDRIS (CNRS)
Pierre Cornette
Pierre Cornette
AI support engineer at IDRIS (CNRS)

Key takeaways

  • Generative AI can create content from a database that has been ingested and according to the indications they are given.
  • These technologies, which remain new, are still being developed and there are still several areas for improvement: reliability, bias in the database, etc.
  • ChatGPT and Bloom are just two models of generative AI, but the concept can be extended to a multitude of applications.
  • These technologies raise a few questions, such as their ecological impact and the risk of using them for potentially malicious purposes.
Épisode 2/3
On January 10th, 2024
4 min reading time
Jeannine Berg
Janine Berg
Economist with the International Labour Organisation (UN)

Key takeaways

  • While generative AI worries workers, ILO economists have studied its impact on the global labour market.
  • The risk is not so much the massive replacement of jobs by bots, but rather the transformation of professions, which will affect 10-13% of occupations worldwide.
  • The professional category of low-skilled office jobs will be particularly affected by AI, since 82% of tasks could be entrusted to bots.
  • Women are particularly affected by automation, as they are twice as likely to be employed in these administrative positions.
  • AI will also increase inequalities, as low-income countries that do not have access to these technologies will have more jobs that could potentially be automated.
  • The current challenge is to support, organise and reflect on the deployment of AI to limit the social consequences.
Épisode 3/3
On February 7th, 2024
4 min reading time
Laure Soulier
Laure Soulier
Senior Lecturer at Sorbonne University in the "Machine Learning and Information Access" team

Key takeaways

  • Generative AI creates content, usually highly relevant and diverse (text, image, video), based on probabilities and deep language models.
  • Despite its performance, AI is neither comparable nor equivalent to human intelligence: rather than truth, it aims for believability.
  • The programme perpetuates the biases and errors of the dataset on which it has been trained.
  • This “work tool” is not expected to replace jobs on a massive scale but could even create new ones.
  • Its long-term development remains uncertain, but it will have to take account of environmental concerns and a move towards frugal AI.