Neural Networks Simulate Schizophrenia with Hyperlearning Theory | Hacker News Reader

Neural Networks Simulate Schizophrenia with Hyperlearning Theory

livescience.com

27 pointsjgv12 comments

Without example in- and output or at least more detailed descriptions of the actual experiments, it's hard to judge the value of the research. It's (unfortunately very) easy to write a system that spews out grammatically correct but otherwise incoherent sentences, a large part of NLP is about this precise problem. I'd like to know more about how the structure of the sentences by DISCERN match that of schizophrenic utterances.

"You can't really prove anything with a computer model like this,"

Indeed, the behavior the article describes sounds like a normal failure mode for a natural language processing computer program, "spewing out a jumble of grammatically correct yet disassociated sentences".

Interestingly enough, this seems to also explain why schizophrenics tend to self-medicate with cigarettes and why smoking tends to increase focus in non-schizophrenics (nicotine increases dopamine levels in the brain).

Edit: I misunderstood, schizophrenia seems to be caused by an elevated dopamine level, not a deficiency.

Could you elaborate on that? If the idea is that dopamine causes schizophrenia (to oversimplify things), I don't see why increasing dopamine with a cigarette would work as self-medication for a schizophrenic.

"A brain chemical called dopamine may help encode what counts as relevant information in the human brain, researchers say."

I think the theory is that a lack of dopamine is what causes schizophrenia, by making it harder for the brain to determine what information is relevant or not.

Aha. But I think it's the other way around, schizophrenia is treated by dopamine suppressors, so too much dopamine means the brain takes all information to be relevant.

Schizophrenia is a complex disease, but most hypotheses have centered on hyperactive rather than hypoactive dopaminergic activity:

http://en.wikipedia.org/wiki/Dopamine_hypothesis_of_schizoph...

There is indeed greater nicotine use by those diagnosed with schizophrenia, but the reasons are likely complex. Self-treatment of some symptoms may be involved, but they may be more associated with nicotine's anxiolytic activity or relief of some of schizophrenia's negative symptoms (e.g., apathy), rather than positive symptoms like frank psychosis.

Ah, ok I misunderstood what I read.

A lack of dopamine (especially in the extrapyramidal system) is what causes Parkinson's. An excess of dopamine (especially in the mesolimbic pathway) is what causes schizophrenia.

Nicotine does increase dopamine in the brain, but only in the same way that my car plays music. It does that, but it's not really relevant to the way nicotine works -- nicotine acts to stimulate an ion channel called the nicotinic acetylcholine receptor. This ion channel stimulation causes several downstream effects, most notably the release of endorphins (biomolecules which mimic the function of morphine), and also the release of adrenaline (which increases focus and concentration).

A lot of people say schizophrenia is basically a lung disease for this reason. The lungs are also where DMT and possibly 5-MeO-DMT are made, so it's possible that smoking moderates this also. Obviously that's extremely speculative though.

If their theory is correct then the researchers have basically proved the inevitability of SKYNET. When the Internet becomes self aware (able parse and understand itself) its inability to forget will cause it to become schizophrenic and start issuing terrorist threats against humans.

you mean GLaDOS

Reply on news.ycombinator.com