"Anthropic reported that Project Deal involved 186 deals made by AI agents, totaling more than $4,000 in value, showcasing the effectiveness of AI in a marketplace setting."
"The experiment revealed that when users were represented by more advanced models, they achieved objectively better outcomes, yet they were unaware of the quality differences."
"Anthropic noted that the initial instructions given to agents did not significantly impact the likelihood of sales or the prices negotiated during the experiment."
Anthropic conducted Project Deal, a pilot experiment where AI agents acted as buyers and sellers in a marketplace. A total of 69 employees participated, each with a $100 budget. The experiment resulted in 186 transactions valued at over $4,000. Four different marketplace models were tested, revealing that advanced AI models led to better outcomes for users. However, participants did not recognize disparities in agent quality, suggesting potential gaps in awareness regarding their performance. Initial instructions did not significantly influence transaction outcomes or prices.
Read at TechCrunch
Unable to calculate read time
Collection
[
|
...
]