GPT-4, orgasms and Stochastic Parrots

This is a text automatically translated from Italian. If you appreciate our work and if you like reading it in your language, consider a donation to allow us to continue doing it and improving it.

The articles of Cassandra Crossing I'm under license CC BY-SA 4.0 | Cassandra Crossing is a column created by Marco Calamari with the "nom de plume" of Cassandra, born in 2005.

Cassandra come back to have his say on AI and on ChatGPT.

This article was written on March 17, 2023 from Cassandra

Cassandra Crossing 534/ GPT-4, orgasms and Stochastic Parrots

Distilling important things from news events and long technical documents is not easy, but forgetting them is very easy; with the release of GPT-4 it is worth doing everything from scratch.

Agreed, the orgasms of "influencers" and those who are already hanging out as registered users on GPT-4 are humanly justified.

The influencers, even the few competent in the matter obviously have to say things that raise the ratings, and therefore spend their time publishing wonders and horrors produced by the false "Artificial Intelligences" of OpenAI/Microsoft, and they try to arouse orgasms and fears in their childrenollower.

Occasional users, almost always completely unaware of the problems of ICT technologies in general, and those of artificial intelligence in particular, are victims of a very strong version of what is known as ”Eliza effect”, and fall prey to sensations they have never felt before, sometimes bordering on orgasm.

Since readers are usually not informed about the basic characteristics of fake “Artificial Intelligences,” they actually have every reason to be amazed, fascinated and excited.

They are victims of Clarke's Third Law: “Any sufficiently advanced technology is indistinguishable from magic”.

Let us therefore state the two incredibly and enormously dangerous aspects of this problem.

Primarily: GPT-4/ChatGPT is part of one of the largest and most dangerous marketing campaigns the computing world has ever seen. It is a campaign aimed at encouraging the adoption, as extensive and as less controlled as possible, of a new product, false "Artificial Intelligence" technologies based on "Deep Learning".

It's a campaign aimed at getting GPT-4 and its ilk accepted as an all-purpose "algorithm" to make our lives better, and save us the trouble of choosing, making decisions, being free. In reality these are new products that only the GAFAM, equipped with immense IT resources, can produce and sell under cartel conditions, opening up new markets for new profits. With the absolute contempt for ethics and the good of humanity that is typical of non-human entities such as multinational companies.

Secondly: the non-intelligence of false “Artificial Intelligences” based on Deep Learning technologies. This fact, totally established and shared but always deliberately hidden, the knowledge of which is vital, can and must be repeated tirelessly, in a simple and completely comprehensible form.

An excellent formulation recently appeared on the “Nexa” mailig list by Professor Enrico Nardelli, whom I thank here. In summary it is:

“GPT4/ChatGPT exhibits human-like proficiency on the syntactic level but is light years away from our semantic proficiency. She has no real understanding of the meaning of what she is doing. Unfortunately, this is a problem of great importance for us on a social level, since what it does it does well, expressing itself in a beautiful and sophisticated form that takes on meaning for us. However, this only happens to the extent that we "project" onto it a meaning that is within ourselves.”

Syntax and Semantics. Language and meaning. The false "Artificial Intelligences" perfectly understand the rules of the language, be it natural, programming or even visual, and can generate excellent-looking documents.

But these documents are devoid of any semantic content, because the false "Artificial Intelligences" do not have the slightest understanding of what they receive as a question, and what they provide as an answer.

These documents have no meaning. They are not "real".

They are, to use an expression known in the artificial intelligence sector, products produced by “Stochastic Parrots”, who repeat what they have heard without understanding absolutely anything about its possible meaning. And they are therefore able to produce the creaking of a door or a poem equally well. Anyone wishing to undertake a challenging read on this can read "On the dangers of Stochastic Parrots: Speech patterns can become too large?, a paper by Emily M. Bender et al.

These two basic problems, we repeat again, that is, the current false "Artificial Intelligences" are:

1. a misleading product, not “Intelligence” but “Parrots”;
2. a huge and dangerous commercial operation with certain negative impacts on very important aspects of society;

have already been tirelessly repeated in various ways this series of 7 utterances by Cassandra.

The eighth statement tries to be concise but complete, to try to overcome the cacophony of completely incorrect and misleading messages and discussions that the media, both popular and technical, they are deafening the vast majority of listeners; the current false “Artificial Intelligences like those of OpenAI/Microsoft:

1. they are not intelligences but dangerous software;
2. they are a huge commercial operation carried out with total disregard for human rights.

And finally, give your favorite prophetess a little personal vent: what a monotonous and thankless job divination is!

Marco Calamari

Join communities