AI Is the Ultimate Philosopher: Here's Why

To answer the final question you pose, it all depends upon the topic under consideration, and the “source information” that the politician and the AI were operating from. 

\
For example, If a politician had specific training/learning in a fairly rigorous field of study, say medicine or engineering, then I would be more inclined to “listen” to that politician on matters pertaining to their training. 

\
On the other hand, if I was concerned with a question where I wanted a neutral distillation of common opinion, I would “trust” a language model referencing AP and Reuter’s articles, more than I would “trust” a politician. 

\
Any preference across the board for the sets of All Politicians, and All AIs, on All Subjects, would be ridiculously indiscriminate.

The second half of my argument is: 

* Any algorithmic distillation or reduction of human readable text (which I shall call a “book”) is necessarily a lower-resolution representation of the book from which it is derived.
* Any algorithmic distillation or reduction of multiple books is not a true synthesis, but rather necessarily a lower-resolution representation of the books from which it is derived.
* A language model is a formula, implemented in a neural net, that is derived from applying one or more algorithms (learning algorithms) to many books (the learning dataset), that can then be applied to new books.
* A chatbot like ChatGPT is the paring of a language model and a “reference” dataset.
* When one “asks” a chatbot a “question” what is actually happening at a technical level is you are requesting a partial distillation of the reference dataset by supplying some query parameters (your “question”).
* Therefore, what one receives from a chatbot in response to a “question” is a lower-resolution representation of the books from which it is derived. If one were to be able to read all the books from which the language model is referencing, one would have a much deeper and better understanding, or answer to one’s question.
* Therefore, the source “books” are the “ultimate” philosopher, not the language model that reduces them to lower-resolution information.

\


Does it matter if the interpolation is trivial or not? My counter example (going from hash algorithm to neural net) claims it does not. Your example establishes a phenomenological position that the source is irrelevant, only the receiver’s direct experience matters. My counter argument uses this same position to make it’s claim. I feel that discounting it for that reason would allow me to discount yours for exactly the same reason: the human neural net and algorithms generating the discourse are infinitely (or at least astronomically) more complex and sophisticated than the AI’s.

\
I feel you missed the point of my second argument. It is unconcerned with the question of awareness that you raise. There seem to be a length limit to replies so I will have to address that in a second reply.

Loris,

The conclusion you draw from your thought experiment would *ipso facto* require us to consider certain books as philosophers, and I have an intuition (I have not actually performed the logic) that this conclusion might well extend to the library the books are in because GPT is not “just code”. The neural net pathways are established by “reading” a vast collection of books (or other literature) and as such are actually a representation of an amalgamation of that literature in the same way that a hash is a representation of a file. 

* Do you agree with this conclusion, and if not, why?

Also, why do you consider that Anna is “making” philosophy. Does a hash function applied to a file containing the text of a book “make” literature, or does it simply make an abstract representation (in the form of a hexadecimal string) of the literature that is already made? And if the hash function is modified such that it produces human readable sentences instead of a hexadecimal sequence, would it then be “making” literature? Or would it simply be creating another, different, abstract representation of the original literature?

Questions are asked in good faith (not rhetorical). I enjoyed your article and would enjoy discussing it with you.

Read the inside story of the failure of enterprise IT

Read My Stories

On the Practicality of Regex for Email Address Processing

Creating a React Studio Plugin From an npm Library - a Walk in The Park!

IT Industrialization: How will it happen? With Adam Wasserman, CEO of Neonto

AI and The Consciousness Gap

Agile: the greatest struggle that never existed

Interpreted, compiled. what. ever.

Hidden Figurers

Is programming art…

Sic mundus creatus est

The over-eager programmer

Pyramids, salt mines, or sausage factories?

The first decade of programming: Speaking in tongues

The second decade of programming: All about Real™ Programmers

The third decade of programming: Just was *is* Productivity?

The fourth decade of programming: Bigger boxes, smaller talk

The Story of Me, a Real™ Programmer?

Framework or language? Get off my lawn!

The fifth decade of programming: J is also for JavaScript

The fifth decade of programming: J is for Java

The fourth decade of programming: How the Web was won

Adam Zachary Wasserman