To be honest I don’t really understand what a token is. Like is it something that’s preset or not

yes i agree, removed into a psych ward where they have mandatory treatment, and stay there for the length that a prison sentence would be, for their benefit and to mitigate the danger they pose to the community. its irrelevant whether it was beause a dirnk was spiked, they were up 4 days on meth, they were grieving. the incident that most likely triggered it, shoudlnt impede them getting the medical care they need, or dictate sending someone who is dangerous to society either in a prison with other people, or in the regular world.

can this thing yank my crank

If I take acid and I see a happy tree talking to me that I kiss. Or I take acid and I see a devil that I stab. What’s the difference

Small snippet of characters, basically the language AI "thinks" in. All you really need to know is more tokens per second = faster

Not yet, but I'm working every day to make this a reality.

But I mean is a token dependent on the AI or is a token an intrinsic measurement

“ I can't engage in conversations that are harmful or derogatory towards any individual or group. Can I help you with something else?”

Annnnd trashed! Dont link woke slop

u tried to f*** it

Humans will be f***ing robots within 100 years guaranteed

Don’t fight it

well ill tell you subjectively, since ive had psychosis and also tripped hundreds of times since, they are completely different feelings and processes. when youre tripping you can tell, even if there are moments where youre disoriented. In psychosis I literally cannot tell there was somehting wrong with my thoughts. it wasnt an even remotely similar experience.

Acid doesnt impair regular people to the point that they stab a devil that wasnt there.

that person has a psychotic break from the acid, or were a murderer coming up with an excuse.

a happy tree that you kiss, is a temporary and harmless passing thought while you were on a psychedelic, that in 99% of people, would integrate in their life in a healthy way.

you can kill somebody on d**** and not be in psychosis. you can take a d*** that triggers a psychosis and kill somebody.

these things are pretty clearly distinct in medical literature. even if psychedelics are techincally a form of temporary psychosis according to some theories.

but now we are getting pedantic about what psychosis is and youre, i guess trying to use the idea that blurry line, to justify that theyre always interchangeable situations. which is foolish

they already do and i did this loads in fallout new vegas

100? more like 5

a happy tree that you kiss

If someone with schizophrenia wakes up and believes they are an angel sent from god who’s sole mission is to feed homeless people, are they psychotic?

i also dont really see the logic in trying to be pedantic about this. If somebody is in psychosis and killed somebody, they should be a ward for a couple decades because theyre a danger to other poepl and themsleves. the cause doesnt really matter

adding pihole to the homelab cus we hate ads!

It’s well accepted that certain d**** induce TEMPORARY psychosis. I feel like you are maybe jumping through hoops to try to prove that d**** aren’t bad lol

It's not a universal measurement. The methodology is largely consistent between models afaik, but the token length and rules used to create the tokens are dependent on the specific AI model and the tokenizer it uses.

I'm far from an expert on this stuff, but that's my general understanding of it.

Basically

Human sends model a message -> tokenizer splits that message up into hundreds or thousands of tokens according to model-specific rules -> the model receives the tokens and runs them through its transformer to predict the "correct" response -> outputted response is translated back from tokens to english and returned to the user

There's more to it than that obviously, but thats the bare bones ideas. Modern models have a "thinking" process where they repeatedly feed potential responses back into themselves for filtering and comparison, until a response hits the criteria to be sent back to the user

The next horizon, and a potentially very dangerous one, is letting the models reason in their own language as opposed to forcing them to use tokens. Think of it like this: Current AI models are something like a person with short term memory loss, they have to write their current thoughts down immediately in what is more or less plain english, and keep a constant paper trail of these to reference so it can orient itself. The main benefit of this is that we can a***yze their reasoning and find out when and why a model is being malicious. But if you don't force the model to do that, and allow it to perform all of it's calculations internally, it gets much smarter, faster, efficient and context aware. All at the cost of knowing what the model is "thinking". OpenAI is already dabbling in this with Astra, but just a little bit. It's already a controversial topic and somewhat of an AI taboo.

they are definetly having pretty significant delusions, which most likely mean theyre having a psychotic break.

In that case they need to get to a hospital, a psychiatrist to fix their medication and probably watch them until the episode ends. the difference this they werent violent.

which the vast majority of people in psychosis are not. theyre a much greater danger to themsleves than others. if theyre not violent, the focus is on them being able to function in society and be healthy, not to isolate them because theyre a danger to others

Heard a pop song by someone called Bella Kay

I liked it!

If it's because of them having psychosis why don't they just stop

We been on that