Went to a therapist as a teenager. They locked me away from my family threatened to have me removed from them permanently and I got to watch a child rip his eye out because they refused to give him his right drugs and changed them with no doctor approval.
I will literally never forget the sight
All because my therapist thought a depressed teenager was too much effort to deal with.
There are bad therapists. Lots of them. Just like any profession, unfortunately. There’s no way to just make them all be good, despite MSW training programs being fairly good overall, a lot of therapists can take a religious route in without taking past-grad level courses. These are overwhelmingly the “Let’s just talk about your mother for the next 15 years while I bill you” types. Properly trained therapists today are method-driven, using evidence-based modalities to solve specific problems. They can actually fix you. Badly trained therapists can wreak havoc.
We think of children and adolescents as the ones being unable to advocate for ourselves, but a huge proportion of adults can’t advocate for themselves in the mental health arena, meaning they aren’t going to get reliably good outcomes.
Actually this is an interesting question. Do local AIs not have or use agentic tools or access to the internet? I assume there are local AIs you can give access to the internet, in order to look up answers to your questions. In which case they absolutely could report you via any kind of online reporting tool that is available on wherever police website or hospital website. Whether the police or hospital accept that report and follow through is another matter. But if it can be done on the Internet, an AI can do it.
You can give local AIs access to the internet, but you don’t have to. You can restrict it to search-only, or you can give it access to tools that have the capability to write/send information. Even then, if you’re in control of the system prompt and the chat software, it won’t have some of the safeguards that an online chatbot does and your model itself would have to decide to snitch on you - on online chatbots I suspect that any such thing is implemented as a separate feature, not baked into the models themselves.
A local AI won’t 5150 you. Some people have trauma many therapists won’t touch.
Went to a therapist as a teenager. They locked me away from my family threatened to have me removed from them permanently and I got to watch a child rip his eye out because they refused to give him his right drugs and changed them with no doctor approval.
I will literally never forget the sight
All because my therapist thought a depressed teenager was too much effort to deal with.
I’ll have what he’s having
There are bad therapists. Lots of them. Just like any profession, unfortunately. There’s no way to just make them all be good, despite MSW training programs being fairly good overall, a lot of therapists can take a religious route in without taking past-grad level courses. These are overwhelmingly the “Let’s just talk about your mother for the next 15 years while I bill you” types. Properly trained therapists today are method-driven, using evidence-based modalities to solve specific problems. They can actually fix you. Badly trained therapists can wreak havoc.
We think of children and adolescents as the ones being unable to advocate for ourselves, but a huge proportion of adults can’t advocate for themselves in the mental health arena, meaning they aren’t going to get reliably good outcomes.
AI won’t improve on this problem.
Actually this is an interesting question. Do local AIs not have or use agentic tools or access to the internet? I assume there are local AIs you can give access to the internet, in order to look up answers to your questions. In which case they absolutely could report you via any kind of online reporting tool that is available on wherever police website or hospital website. Whether the police or hospital accept that report and follow through is another matter. But if it can be done on the Internet, an AI can do it.
You can give local AIs access to the internet, but you don’t have to. You can restrict it to search-only, or you can give it access to tools that have the capability to write/send information. Even then, if you’re in control of the system prompt and the chat software, it won’t have some of the safeguards that an online chatbot does and your model itself would have to decide to snitch on you - on online chatbots I suspect that any such thing is implemented as a separate feature, not baked into the models themselves.