Despite the electrifying promises of generative artificial intelligence (AI) since the arrival of ChatGPT two years ago, its adoption by the economic sector is nuanced, with companies often favoring caution as the year 2025 approaches.
Getlink, Eurotunnel operator, would like, for example, to create for its teams “a maintenance companion based on generative AI,” Denis Coutrot, the group’s director of artificial intelligence, explains to AFP.
But first, the company that runs the 38-kilometer tunnel between France and the United Kingdom, with 400 trains a day, is carefully experimenting.
It started small, implementing a conversational agent (or “chatbot”) that consults its internal regulatory documentation.
“We were very afraid of the issue of “hallucinations” (phenomenon that leads to absurd, incoherent or false answers, NLDR)”, he adds, referring to the “very strict procedures” with 800 m long railway shuttles, the equivalent of four TGVs. .
A methodical and thoughtful, if not revolutionary, approach that many companies have.
“No more coders in 100 years”
“ChatGPT is obviously amazing, but it’s quite difficult to use it really effectively in our daily work,” James Sutton, founder of London-based legal services firm Avantia Law, told AFP.
While generative AI excels at simple tasks like searching legal databases and summarizing documents, complex work requires human supervision, he continues.
The lack of reliability of this tool is an obstacle in the legal sector.
For the annotation of a contract, “AI can do it perfectly and, in another contract, it will only be 40% correct. Because of this uncertainty, lawyers always have to check everything,” emphasizes James Sutton.
It is not surprising that “tech” companies are less careful in adopting this technology.
Nearly a quarter of the new computer code produced at Google is generated by AI, Sundar Pichai, head of parent company Alphabet, said in late October.
Next year, 75 to 80% of the tasks assigned to developers will be able to be performed using artificial intelligence, according to Kirill Skrygan, CEO of software publisher JetBrains.
Imaging tools, such as DALL-E, Midjourney and Stable Diffusion, have a strong impact on creative industries such as fashion, transforming work habits and collection marketing times.
lack of curiosity
On the other hand, in the field of health, doctors are hesitant to take the full step, despite the fact that a recent American study demonstrates its effectiveness when making diagnoses.
Costs, innovation, surveillance are the elements that companies must deal with.
“The market will need time to evaluate all the costs and benefits, especially in a context in which companies are already hesitant about technological investments,” Seth Robinson, of the American association CompTIA, observes to AFP.
According to Anant Bhardwaj, director of San Francisco-based AI platform Instabase, AI is strong at analyzing pre-existing data and patterns, but lacks the human curiosity needed to push boundaries.
For him, within ten years, most sectors will incorporate artificial intelligence into their activities, but with humans watching over the grain, and the fantasy of autonomous AI is still far away.
Source: BFM TV
