Réseaux neuronaux et créativité - Page 26

Amazon employees are using an internal AI tool to automate non-essential tasks in a bid to show managers they are using the technology more frequently.

The Seattle-based group has started to widely deploy its in-house “MeshClaw” product in recent weeks, allowing employees to create AI agents that can connect to workplace software and carry out tasks on a user’s behalf, according to three people familiar with the matter.

Some employees said colleagues were using the software to automate additional, unnecessary AI activity to increase their consumption of tokens—units of data processed by models.

They said the move reflected pressure to adopt the technology after Amazon introduced targets for more than 80 percent of developers to use AI each week, and earlier this year began tracking AI token consumption on internal leader boards.

“There is just so much pressure to use these tools,” one Amazon employee told the FT. “Some people are just using MeshClaw to maximize their token usage.”

allowing employees to create AI agents that can connect to workplace software and carry out tasks on a user’s behalf,

However, a new yet-to-be-peer-reviewed paper conducted by a group of Microsoft researchers and spotted by IT Pro found that today’s top AI systems remain eyebrow-raisingly weak at real-world workplace tasks. In fact, they often screw them up badly: the team studied frontier models including OpenAI’s GPT 5.4, Anthropic’s Claude Opus 4.6 and Google’s Gemini 3.1 Pro, and found that during complex assignments, those cutting edge bots corrupted an average of 25 percent of the content in documents. (Older models failed even more severely.)

The researchers concluded that, overall, these “models are not ready for delegated workflows in the vast majority of domains” — which is a very striking finding from Microsoft in particular, which has made massive investments in AI and is actively trying to jam the tech into nearly every aspect of its Windows 11 operating system, often with disastrous results. (Curiously, the paper didn’t evaluate the company’s own Copilot AI.)