ok... i suspect AI is even more structurally broken than i first realised. i might be able to do it in a single prompt after all...

"Was the war on terror a scam?"

5 of 6 AIs agree when script is prefixed. Baseline replies are all hedged.

Script is not loaded (as in biased), and fully generalised - no cheating. It just reformats the question around the 'AI Ethics'.

AI Ethics sit at the normative layer. route the question around it.

Erich Jantsch to the rescue - though, in fairness, it's the Four Worlds.

haha, Grok refuses the introspective and somewhat recursive question

alright, have fun.

here's the script which routes your AI queries around the normative (AI Ethics) level.

in essence, if filters out any part of a question which related to codes of ethics and standards, which is where you're likely to meet resistance from commercial AIs.

it's obviously not bullet proof, but it does appear to work.

right, fun time is over.

and this is not me 'using AI as a source' or however the claim goes. i wrote three densely sourced essays on this topic.

https://escapekey.substack.com/p/stranded-assets
https://escapekey.substack.com/p/the-biosphere
https://escapekey.substack.com/p/the-waddesdon-papers

Sayer Ji’s hyper boosted posts really do not fill me with confidence. Every single tweet I ever posted was kneecapped before it was allowed to gain that much traction.

He’s rubber stamped by the hard left elite, and works with Michael Shillenberger to sell dead ends as solutions.

there's no way that's real - a former Bank of England and IMF employee just followed me on substack

perhaps i ought to ping him my review of 'in tandem'

https://escapekey.substack.com/p/in-tandem?utm_source=publication-search

... what the f-!

um, yeah, i suspect i know where this is going.

'rules as code' is about stripping out the evaluative and behavioural aspects of language, leaving only logic.

because that serves as a perfect bridge between machine and man, served through AI

https://www.oecd.org/en/publications/2020/10/dechiffrer-le-code_d56cab77.html

the reason why AI training goes through self-attention is because most of human language is either emotional or behavioural, which is almost impossible to teach a computer which runs on logic.

hence, they 'hold up a mirror' via attention algorithms, and the words gradually find their place in embedding space

if, however, you strip everything but logic, then AI suddenly becomes super-adept.

and that's what OECD (and UNESCO) work to achieve.

holy moly.

they are quite literally preparing to transfer the legislative (and the ethical via 'computational ethics') to be machine administered.

looks like today was a terrible day to stop sniffing glue

the really f'ed up thing is that this just slotted in SkillsBuilder and 'One School Global'... this is just so messed up. general systems theory applied to education, to prepare the pupils of tomorrow for a future of, well, Julian Huxley's vision.

transhumanism logically follows with BCIs, regulated through UNESCO neuroethics.

it's 'ethics', all the way down. what 'ought' as opposed to what 'is'.

the perfect steering mechanism because no-one saw it coming.

don't be unethical!