Colorless green ideas sleep furiously | Hacker News Reader

Colorless green ideas sleep furiously

en.wikipedia.org

15 pointsBerislavLopac18 comments

I have that on vinyl.

Mentioned in today’s xkcd

Chomsky's Transformational Grammar[1] was applied to transcripts of gifted psychologists (Virginia Satir, Milton Erickson and Fritz Perls) to develop the "Meta-Model"[2] and eventually Neurolinguistic Programming, a very potent and rigorous school of applied psychology.

It's also the origin of the Chomsky hierarchy "a containment hierarchy of classes of formal grammars."[3]

Pretty neat, eh? Computer languages and a formal working psychological toolkit both have roots in Transformational Grammar.

[1] https://en.wikipedia.org/wiki/Transformational_grammar

[2] https://psychology.wikia.org/wiki/Meta_model_(NLP)

[3] https://en.wikipedia.org/wiki/Chomsky_hierarchy

I ran across the Chomsky hierarchy last week while researching expression languages (like jsonnet and Nix). Really blew my mind. Made me think how we could reimagine our legal language in a more structured way.

Legal language doesn't have any reason to be re-imagined, especially in the direction of making it more formal. If anything, it would probably benefit from stopping to pretend it is a kind of formal language. It doesn't really help to use a formal(-ish) language to describe informal semantics, and there's no way at this time to describe the world in formal semantics (for example, you couldn't formally, mathematically, define the concept 'contract' in a way that matches what the legal system generally understands by it).

I'm curious, why would it be impossible to define the concept of legal contract formally ?

Try it and it will quickly become apparent. Most of the concepts we work with are fuzzy and dependent on many implicit understandings that are very hard to formalize in satisfying manner. In particular, a contract must be willingly entered by the participants, which in itself is a very hard to formalize concept (e.g. if I am dying of hunger and believe that my only chance of eating is to enter this contract, am I willingly accepting it or am I being compelled? How is this different from someone putting a gun to my head and telling me to sign?)

Note that what I'm thinking of here is a formal definition that is rigorous enough to be able to apply formal mathematical methods to reason about it.

As legal codes become broader, more specific, and more voluminous, I believe we will need tools to automate enforcement of policy. Hard to do that without a formal language.

I'm not talking about the first amendment to the US constitution here. I am talking about legal codes used to standardize health and safety standards for goods manufactured in different jurisdictions.

How would you automatically enforce a policy that says that a good can't cause harm? What kind of machine could automatically verify this? Or am I misunderstanding what you mean by automatic enforcement of health and safety standards?

Pesticides and heavy metals are often measured in food, water, and other goods and have limits. Having these limits specified programmatically ensures there is a fast and efficient feedback loop between proposals to changes in limits and entities that would be affected by the policy changes.

Mandatory financial reporting can benefit from automation as well IMO.

The limits are specified explicitly in current laws as well. The problem is always things like deciding the limits, measuring the values in real products, establishing QA standards that give us a good chance that the levels in each unit match the levels that were measured when approved etc.

You'll also have all sorts of people lobbying for more lax standards claiming that stricter standards will have too much impact on businesses etc.

Financial reporting could almost certainly benefit from more automation. However, even there a lot of the problems happen around what revenue to recognize, how to prevent people from moving money around in fictitious transactions etc.

Another example, this time unrelated to health&safety or corporate governance:

A legislature should be able to express their intention to apply financial stimulus according to some expression that takes macroeconomic and/or quality of life metrics as input.

Macroeconomics is not a science, it is a mathematical model tangentially related to reality in a fuzzy way (if at all). Quality of life is not rigorously measurable.

Governments have and will continue to tie financial stimulus to financial indicators.

Governments tie their stimulus to political beliefs, and they usually choose various financial indicators that reenforce those beliefs. This government is targeting some unemployment number, that argument cares about debt, some other government is more concerned with the stock market etc. Some claim that this measure was good for workers because it lowered unemployment, other will claim that it was bad for workers because it increased underemployment. You choose which to believe, but you won't find a formal definition.

> Neurolinguistic Programming, a very potent and rigorous school of applied psychology

"Neurolinguistic Programming" is far from "a potent and rigorous" school of anything, it is gobbledygook that has nothing to do with Chomsky's insights into linguistics.

In particular, Chomsky's work focuses on understanding the basic patterns behind language itself as a general human trait, at a very basic level, because that's all that can be studied at the moment. The concepts from Transformational Grammar and Universal Grammar are far too simple to be used in analyzing actual human speech, it would be like trying to use Newton's laws of motion to analyze how athletes run and claiming you discovered something important: pseudo-science at its worst.

With respect, you are speaking from ignorance.

NLP arouses the ire of what I call "Capital-S Skeptics". That's why I didn't link to the Wikipedia article on it, as it's been glossed over with cries of "pseudo-science!" by ignorant skeptics so you can't really get any information about NLP from WP except that skeptics are skeptical of it.

I have personal experience of NLP and I can vouch that it is effective (you can change yourself quickly and easily) and rigorous (it develops repeatable algorithms that can be taught to others who can then replicate the results), e.g. the "Phobia Cure" pattern that cures phobias quickly and permanently typically in a single application.

I myself was cured of a terrible depression that was ruining my life in a one-off treatment that took less than ten minutes. Saved my life.

> I myself was cured of a terrible depression that was ruining my life in a one-off treatment that took less than ten minutes. Saved my life.

Having known many people who struggle with depression, please allow me to simply consider you a liar on this point at least. It is claims like this that make it obvious that NLP is either culty or snake-oil (or both).

Claiming that you can cure complex diseases with neurological and social and psychological causes in 10 minutes with repeatable algorithms inspired by applying linguistic analysis on the work of 3 therapists is so obviously bunk that I'm wondering if you'll next tell me about this one weird trick if I send you a one-off payment.

Reply on news.ycombinator.com