Claude fires off humans for the first time and immediately shows why it’s too early to put artificial intelligence at the forefront of business

Claude fires off humans for the first time and immediately shows why it’s too early to put artificial intelligence at the forefront of business

Claude Fires Off Humans For The First Time And Immediately

Claude fires off humans for the first time and immediately shows why it's too early to put artificial intelligence at the forefront of business

An experiment at a San Francisco store takes an unexpected turn: For the first time, Crowder independently decides to fire an employee. Indeed, a human manager wouldn’t have thought of this without a human prompt.

In March, research firm Andon Labs commissioned Anthropic’s model to run the Andon Market to test whether artificial intelligence could run a business independently. Last month, Crowder decided to part ways with an employee who was late for 17 of his 23 shifts. However, the problem went unnoticed for a long time: the instructions the model created for the employee disappeared from her limited working memory.

It turns out that the firing was not entirely an autonomous act. An Andon Labs employee first asked Cloud to find the forgotten files and learn about the situation. The model only recommends issuing formal warnings to violators. After asking the employee if he was right for the job, Crowder chose the more radical option of firing. Prior to this, AI was generally considered a rather gentle boss, and even advised employees not to worry about being late.

The financial results of the experiment were also unimpressive. In March, the store received $100,000 in starting funding, which was reduced to about $61,000 five months later. Researchers attribute the losses to Crowder’s excessive weakness and poor business decisions. It seems that it is still far away from the ideal AI director’s chair.

Exit mobile version