We can tell that other people are intelligent and conscious because they're like us in important ways and we are intelligent and conscious. I haven't ruled out intelligent machines in general. My point was about the inadequacy of the Turing Test.

You can be reasonably sure about your friends, as they're sufficiently like you to leave little doubt. I disagree but can understand you thinking that LLMs are intelligent, but unless by conscious you mean something like being self-aware, I don't understand how you could conclude LLMs are conscious. For example, although LLMs can tell you what colours things are, how the visual system distinguishes colours in humans and other animals, and what wavelength mixtures different colours are, they don't know what it's like to see colours because they cannot see colours. All the input of LLMs is text, and you can't describe, in words, what it's like to see colours, or hear sounds, or feel pain, in a way that someone incapable of these sensations would understand. And to remove any remaining doubt, there was an expert in colour vision (Knut Nordby) who, as he could not see colours because his retinas lacked cones, could not imagine what it's like to see colours, despite all his knowledge.

Arguably, ELIZA passed the Turing test ca. 1967: https://archive.computerhistory.org/resources/access/text/20... and not that long ago, ELIZA was still more convincing: https://arstechnica.com/information-technology/2023/12/real-...

The Turing Test can be gamed. This was done in some Loebner Prize entries: see https://www.behavior.org/resources/325.pdf and https://en.wikipedia.org/wiki/Eugene_Goostman

To determine whether a machine is intelligent, I think it's necessary to understand how it works internally and not just take a black-box approach. This was already understood when Alan Turing wrote his 1950 paper. Elmer and Elsie, William Grey Walter's tortoise robots built during the preceding two years, appeared to behave as if they had some degree of self-awareness. But their "brains" consisted of only two thermionic valves, so self-awareness can be ruled out.

A propos of breathing marks, iota subscripts, and the three different accent marks of Classical Greek, when I learned it at school we had to remember the breathing marks and iota subscripts, and would lose marks if we omitted them, but we didn't need to learn the accents. Modern Greek now has only (acute) accents, which you need to know to stress the correct vowels, exactly where the accents were in the equivalent Classical Greek words.

But it appears to be valid in PDP-1 Lisp (though not in either Lisp 1.5 or modern Lisps). From https://s3data.computerhistory.org/pdp-1/DEC.pdp_1.1964.1026...

"Doing a CDR of an atom is permissible and will get the atom's property list. Doing a CAR of an atom may very easily wreck the system."

Comments URL: https://news.ycombinator.com/item?id=48575074

Points: 7

# Comments: 1

Even if they become more reliable, I like to understand and work things out for myself, rather than just be given the answer.

I assume, though, you mean LLMs. I haven't used them first hand, but I have fairly recently implemented a multi-layer artificial neural network in C, mostly as a learning exercise, but as I had previously built a speech spectrogram in Lisp, I thought I'd try to use it to recognize phonemes, with one hidden layer. The Lisp communicated with the ANN via a Unix pipe. It worked reasonably well for just vowels, but when I added other sounds (e.g. l, r, s, z), its performance deteriorated. I think the C is bug free, but I don't know an easy way to train the ANN. I've tried adding to the training set, adding an extra layer, changing the number of neurons in the hidden layer. The usual debugging skills don't seem to help there.

There are others who just want a problem solved, and it doesn't matter to them how the program which solves it gets written, as long as the program (appears to) work and it's done as quickly as possible, so they outsource the development to AI. That is not hacking.

And here's the clip with aspect ratio corrected: https://www.stretch.site/?videoUrl=https%3A%2F%2Fwww.youtube...

The test here is for the inverse square law of gravity. The rival theory in this case isn't GR, but MOND: https://en.wikipedia.org/wiki/Modified_Newtonian_dynamics

We agree on Algol, Lisp, Forth, APL, and Prolog. For ground-breaking functional language, I have SASL (St Andrews Static Language), which (just) predates ML, and for object oriented language, I have Smalltalk (which predates Self).

I also include Fortran, COBOL, SNOBOL (string processing), and Prograph (visual dataflow), which were similarly ground-breaking in different ways.