AI agents get office tasks wrong around 70% of time, and many aren't AI at all
AI agents get office tasks wrong around 70% of time, and many aren't AI at all
www.theregister.com
AI agents wrong ~70% of time: Carnegie Mellon study
7 comments
-
IT consultancy Gartner predicts that more than 40 percent of agentic AI projects will be cancelled by the end of 2027 due to rising costs, unclear business value, or insufficient risk controls.
That implies something like 60 percent of agentic AI projects would be retained, which is actually remarkable given that the rate of successful task completion for AI agents, as measured by researchers at Carnegie Mellon University (CMU) and at Salesforce, is only about 30 to 35 percent for multi-step tasks.
To further muddy the math, Gartner contends that most of the purported agentic AI vendors offer products or services that don't actually qualify as agentic AI.
This is the way a consulting company breaks the awkward news to US clients that the US economy is about to absolutely tank as AI swallows up the money we desperately needed for real problems like climate change.
-
It’s kind of hilarious that “AI” is taking us down without even being actual AI. Maybe hilarious isn’t the right word.
-
And yet I laugh. Maybe this is finally my Joker moment?
-
-
-
Good thing we're about to pass a bill preventing us from regulating it in any meaningful way.
-
Correction: NONE of them are AI at all
-
I like when I Google how to change some setting on my phone and I always get a hallucinated set of instructions, without fail.
-
Copilot is ass for day to day tasks. Okay for generalized research like if you have a big task and zero idea where to start.
I think these tools are much more useful when given a specific task. They're also super powerful in the dev space for faster coding, automated testing, and ETL
7 comments