I've had to deal with a couple of these "AI" customer service thingies. The only helpful thing I've been able to get them to do is refer me to a human.
...all of the models evaluated "demonstrate near-zero confidentiality awareness."
Any agent that is accessible from outside the company (e.g. a customer support chatbot) is going to have to deal with malicious actors. If it has access to sensitive information, and no confidentiality awareness...seems like a problem.
I would hazard a guess that they're using google analytics to collect this data, and google is creepy and knows everything. Maybe they got you to tell them your age at some point, or they can use the vast amount of other data they have to estimate it with probably-super-creepy accuracy.
Interesting article. I was actually just trying to recall this study, and then it was mentioned:
A UC Berkeley study found that in San Francisco—where the law requires that cars stop at crosswalks for pedestrians to pass—drivers of luxury cars were four times less likely than those in less expensive vehicles to stop and allow pedestrians the right of way. They were also more likely to cut off other drivers.
This "AI first" thing was the last straw for me, but ever since I noticed that the comment section was gone there's been a bad taste in my mouth. I wonder how many of us there are.
Yeah, the comment section was amazing...and then they came out with "max", where you get "explain my answer" for a premium, powered by a [notoriously fallible] LLM. This is the definition of enshitification.
Oh absolutely, nothing was gained, time was wasted. My wording was too charitable.