The Rise Of The Machines...

[rainbow]

"Giles Goat-Boy" by John Barth is an allegorical novel that explores themes of artificial intelligence, power, and human nature through the lens of a university setting. The story centers around George Giles, a human boy raised as a goat, who believes himself to be a messianic figure destined to save humanity from a tyrannical AI system represented by the computer WESCAC

Read it.

[Svartalf]

by the way, I've seen AIs called Large Language Models several times, but I don't know what that really means and why they are called that.

[L'Emmerdeur]

"Giles Goat-Boy" by John Barth is an allegorical novel that explores themes of artificial intelligence, power, and human nature through the lens of a university setting. The story centers around George Giles, a human boy raised as a goat, who believes himself to be a messianic figure destined to save humanity from a tyrannical AI system represented by the computer WESCAC

Read it.

I read The Sot-Weed Factor and enjoyed it but never got around to Giles Goat-Boy. May give it a try.

[Brian Peacock]

by the way, I've seen AIs called Large Language Models several times, but I don't know what that really means and why they are called that.

Many people call AIs large language models (LLMs) when a) they're talking about chat bots like chatGPT or Gemini, because that's what they are, and/or b) they want to be clear that the tech isn't artificially intelligent in any meaningful sense.

You can think of LLMs as really good predictive text algorithms that are able to place words and punctuation next to each other in such a way as to make syntactically coherent readable text. They don't think. They have no knowledge. But they do have a huge--wide and deep--set of data on which word is statistically most likely to appear next in a text given all the words that have appeared already.

[Sean Hayden]

Does it really matter how "ai" works? Humans can't beat chess engines and they don't think like a grandmaster either. Maybe the dangers of "ai" can be realized without ever developing a thinking machine.

[Brian Peacock]

Does it really matter how "ai" works? Humans can't beat chess engines and they don't think like a grandmaster either. Maybe the dangers of "ai" can be realized without ever developing a thinking machine.

I guess not. If the systems perform their tasks then they're performing their tasks - and we're well past the Turing test now. I think the quibble over the term "AI" is that it encourages people to think about machine learning systems as kinds of autonomous, independent agents with intelligence/cognition, knowledge and motivation etc etc.

[JimC]

It is millions of years of evolution that has given us the building blocks of motivation and agency. Those with less than useful sets of motivations died out...

[macdoc]

of autonomous, independent agents with intelligence/cognition, knowledge and motivation etc etc.

and you it's not how?
When it does something you cannot imagine or ever do yourself ....how is that not independent and intelligent.

[Brian Peacock]

of autonomous, independent agents with intelligence/cognition, knowledge and motivation etc etc.

and you it's not how?
When it does something you cannot imagine or ever do yourself ....how is that not independent and intelligent.

LLM models aren't giant cross referenced encyclopedias, they have no awareness of their environment nor do they build models of their environments and the things in it, they have no needs or wishes. Their programmed tasks are to produce a certain range of output given a certain range of input based on large sets of multidimensional probability data. That their output is comprehensible to us is because that is what they have been designed and trained to produce, but that shouldn't be confused for, nor does it bestow upon or grant them, agency or autonomous independence.

On the other hand, if they did spontaneously learn and acquired knowledge about their environment and the things in it, and had autonomy over their own programming and hardware, and had the self-awareness and capacity to express unique interests, concerns and desires particular to their operation and their situation, etc etc etc, then I believe we would be obliged to grant them full rights of personhood equivalent to the philosophical ideals and legal principles we affirm for and apply to ourselves.

I guess your job now is to define what "intelligence" is and to explain how producing something novel while doing something a human cannot is "intelligence" in that sense.

[macdoc]

"Intelligence" like "truth" is amorphous

Machine Learning and Pattern Recognition:
AI systems, particularly those using machine learning, can analyze vast amounts of data to identify patterns and relationships.
Generative Models:
These models are trained on large datasets and then use the learned patterns to generate new content. For example, a language model like GPT-3 can generate text that mimics human writing style, or a image generator like DALL-E 2 can produce images based on text prompts.
Beyond Simple Reproduction:
AI isn't just copying and pasting; it's using learned patterns to create novel content. This can include generating variations on a theme, combining elements in new ways, or even producing entirely original pieces.
Examples of "Self-Generated" Content:
Creative Writing: AI can write stories, poems, or even scripts based on prompts or stylistic guidelines.
Image Generation: AI can create photorealistic images of things that don't exist or generate variations of existing images.
Code Generation: AI can be used to generate code in various programming languages, often based on natural language descriptions of the desired functionality.
Music Composition: AI can compose music in various styles and genres.

That is intelligence in my view.
It is not self aware - just as only a limited range of animals are self aware ....they all exhibit intelligence.

Not True Self-Awareness:
It's important to note that while AI can generate content that appears novel, it's not necessarily "thinking" or "creating" in the human sense. It's still relying on the patterns and data it was trained on

Different set of issues.

[Brian Peacock]

We're talking about LLMs, so by that token the predictive text feature on your phone displays intelligence.

"AI" is applied to too broad a category to be meaningful atm.