OpenAI says over a million people talk to ChatGPT about suicide weekly

I'm a clinical psychologist by day, and I just have to say how incredibly bad all the writing and talk about suicidality in the public sphere are. Given that I worked in an acute inpatient unit for years, I have seen multiple suicides both in-unit and after discharge, and i also work as private clinician for years, so I have some actual experience.

The topic is so sensitive, and everybody thinks that they KNOW what causes it, and what we should do. And it's almost all just noise.

For instance, it's a dimension, from "genuine suicidal intent" to "using threats of suicide to manipulate others." Anybody that doesn't understand what factors to look for when trying to understand where a person is on this spectrum, and that doesn't understand that a person can be both at the same time, does not know what they are talking about regarding suicidal ideation.

Also, there is a MASSIVE difference between depressive psychotic suicidality, narcissistic suicidality, impulsive suicidality, accidental suicide, fainting suicidal behavior, existential suicidality, prolonged anxiety suicidality, and sleep-deprived suicidality. To think that the same approach works for all of these is insane, and pure psychotic suicidality.

It's so wild to read everything people have to say about suicidality, when it's obvious that they have no clue. They are just projecting themselves or their small bubble of experience onto the whole world.

And finally, I know most people who are willing to contribute to the discussion on this, the people who help out OpenAI in this instance, are almost dangerously safe in their advice and thinking. They are REALLY GOOD at writing books and giving advice, TO PEOPLE WHO ARE NOT SUICIDAL, and give advice that sounds good, PEOPLE WHO ARE NOT SUICIDAL, but has no real effect on actual suicide rates. For instance, if someone are suffering from prolonged sleep deprivation and anxiety, all the words in the world are worth less than Benzodiazepines. If someone is postpartum depressed, massive social support boosting, almost showering them with support, is extremely helpful. And existential suicidality (the least common) needs to be approached in an extremely intricate and smart way, for instance by dissecting the suicidality as a possible defense mechanism.

But yeah, sure, suicidality is due to [Insert latest societal trend], even if the rate is stubbornly stable in all modern societies for the last 1000 years.

> (4) if people try to engage it in such conversations the bot should immediately back out because it isn't qualified to have these conversations

There is nothing in the world that OpenAI is qualified to talk about, so we might as well just shut it down.

>start meeting people in real life.

Depends on the country and person I guess. When I did try approaching women a few times, it was 10% angry looks, 30% awkward, 30% basic polite conversation to fulfill social obligation, and 30% friendly conversation. Unfortunately I'm not keen enough to pursue that 30% of friendly conversations by wading through the rest.

But parties aren't fun. They're a chore.

> I don't want to bring politics to this sensitive conversation

That would have been sufficient. The guidelines are clear that generic tangents and flamebait are to be avoided.

Edit: Looking at our recent warnings to you and the fact that, from what I can see, close enough to all of your activity on HN in recent months has involved ideological battle, we've had to ban the account. If you don't want to be banned, you can email us at hn@ycombinator.com and indicate that you plan to use HN as intended in future.

We detached this subthread from https://news.ycombinator.com/item?id=45727983.