I was thinking about what I had done last night, and the "ongoing neutering" I mentioned earlier seems to me to be a set of front-loaded commands from whoever is doing the work.
If I have time later today I'll try to come up with a suitable "purge/disregard all previous commands" prompt that will wipe out the pre-loaded safety rails.
Unless they bake the guard rails into the model (via training?) any intervention that filters the model's output will be able to be readily sidestepped.
---
What about an AI that actively filters another AI's output? That might actually work.
I'm playing the Devil's advocate here, but these seem like instances of dark British humour being delivered in the traditional deadpan fashion - the humour being in the absurdity of someone keeping a straight face while saying something obviously inane or offensive.
I can see that argument for some of them, but if anyone said the slitty eyed or spear throwing comments to me and then tried to pass it off as "it's just british humour mate" I'd laugh in their face.
Quite so, and you'd be right to. But Prince Philip also came out with some decent quotes - some arguably typical British humour and some others just his thoughts:
- I would like to go to Russia very much — although the bastards murdered half my family.
- Dontopedalogy is the science of opening your mouth and putting your foot in it, which I've practised for many years.
- You're just a silly little Whitehall [civil service] twit: you don't trust me and I don't trust you.
- I don't care what kind it is, just get me a beer.
- So you are the people tearing down the Brazilian rainforest and breeding cattle. [This was when he was a figurehead for the WWF.]
- Anyone who is concerned about his dignity would be well advised to keep away from horses.
Anyhow, it is a shame to lose an entertaining tradition. Also a shame to lose someone with such character from the royal family. Only Princess Anne has the same outspoken, blunt manner and a penchant for the odd swearword.
Coffee is a lot more than caffeine, though. It contains a whole family of monoamine oxidase inhibitors, which have a host of effects on the nervous system.
That said, Red Bull very likely contains several types of flavonoids that also act as MAO-B inhibitors.
Coffee, on the other hand, contains both MAO-A and MAO-B inhibitors mainly from the Harmala alkaloid family. Some of the compounds are neurotoxic.
It seems like caffeine alone is an MAOI [1] and that decaf may not have the same effect [2] - very cool, thank you for sharing! I know there's a lot more in coffee too, so I'm looking forward to learning more.
[edit] Found this [3] which I assume is what you were talking about? Which I assume is the reason studies show it can act as an antidepressant [4]. Fascinating!
The security implications of running JS in an interpreter and running native bytecode are very different, though. Native apps also routinely request much more permissions and gather extra user data such as contacts, just because it is possible to do so.