> These chatbots are different.
Strong disagree, it's very easy to step back and say this is a program, input, output, the end.
All the people claiming this is some exhibition of personhood or whatever just don't want to spoil the illusion.
> These chatbots are different.
Strong disagree, it's very easy to step back and say this is a program, input, output, the end.
All the people claiming this is some exhibition of personhood or whatever just don't want to spoil the illusion.
It's not that these are magical machines, and TFA shouldn't have gone that direction, it's that "what if we are also just a repeated, recursive, story that endlessly drolls in our own minds"
> Seeing and being seen is apparently just neurotic streams of interleaved text flowing across a screen.
... Sounds to me a clunky analogy of how our own minds work.
Without the language to express "thoughts" and not-self in that level of abstraction, one cannot form the thought at all that it might not be those thoughts.
That should be all you need to know to see that a human being doesn't solely consist of language. If you do a deeper dive into this you'll also find other cases such as Genie[0] where people with intact brains didn't initially acquire language and yet were still able to develop cognitively. We are fully capable of functioning without having a language, which makes sense if you think about it. The ability to understand language must necessarily precede the invention of a language in the same way that the ability to grasp a tool must precede the invention of the tool.
[0] https://en.wikipedia.org/wiki/Linguistic_development_of_Geni...
If so, then the author is absolutely, terribly wrong. The human personality, how it develops, why it develops, the mechanisms by which it operates- those are some of the greatest unsolved mysteries, some of the most interesting questions in science. The nature of the world, perhaps, comes first, in the list of scientific questions; then, the nature of ourselves, of humanity.
If the authors is saying that there is nothing special about being a person, that's because they, like the rest of us, all, do not understand what it means to be a person.
You know, despite being one, and everything.
> In fact, it is hard to argue in 2023, knowing what we know of online life, that online text-personas are somehow more impoverished than in-person presence of persons
It is in fact very easy to argue. No one on the Internet knows you're a dog, there is no stable identity anywhere, anonymization clearly creates a Ring of Gyges scenario, trolling, catfishing, brigading, attention economy, and above all, the constant chase for influence (and ultimately revenue) - what passes for "persona" online is a thin gruel compared to in-person personas.
When you bump into a stranger at the DMV, you aren't instantly suspicious of their motives, what they're trying to sell you, are they a Russian influence farmer, etc.
Night and day. Extremely impoverished.
I don't know whether to attribute that to a defense mechanism that marketing has forced me to construct, or if indeed it is due to 9/10 they are actually trying to sell me something.
this is a jumping off point into considering your own mind as an illusion. your own self with its sense of personhood: i.e. yourself as the it-element in a I-it interaction.
But if we leave it at that, it's essentially a very nihilistic (deterministically reduced), so either turn back, or keep going:
the fact that your own personhood is itself very much an illusion is OK. such illusion, however illusory, has real and potentially useful effects
when you interact with your computer, do you do it terms of the logical gates you know are there? of course not, we use higher level constructs (essentially "illusory" conceptual constructions) like processes and things provided by the operating system; we use languages, functions, classes: farther and farther away from the 'real' hardware-made logic gates with more and more mathematical-grade illusions in between.
so the illusions have real effects, in MOST contexts, it's better to deal with the illusions than with the underlying implementations. dunno, what if we tried to think of a HTTP search request into some API in terms of the voltage levels in the ethernet wires so that we truly 'spoil the illusion'??
Just because people don't actively think all the time in terms of low level contexts doesn't mean that only simulating the high level contexts is a sufficient substitute for the whole process.
I think this whole concept is conflating "illusion" (i.e. allowing oneself to be fooled) and "delusion" (being involuntarily fooled, or unwilling to admit to being fooled.)
I personally don't enjoy magic shows, but people do, and it's not because they think there's real magic there.
See also Aristole's description of a 'soul' (Lat. anima/Gk. ψυχή), which is embodied above all, unlike the abstract description of the soul that the West would go on to inherit from Neo-Platonism via Christianity.
Even though today we know full well we are indissolubly embodied entities, the tendency to frame identity around an abstraction of that persists, but it seems thinking around this hasn't completely succumb to this historical artifact, see 'Descartes' Error: Emotion, Reason, and the Human'
That argument relies on presumptions of what a program can and cannot be.
It’s very easy for me to step back and say my brain is a (self-modifying) program with input and output, the end.
Some basic text based APIs that would quickly improve LLM utility:
Calculators
Database storage and retrieval
Web access (already kind of done by bing)
Shell scripting
Thinking further into the future of multimodal models, it’s not hard to imagine this sort of thing could be extended to include image based APIs. Imagine a LLM looking at your gui and clicking on things. The sky’s the limit at that point.
Checkout toolformer, they’ve got this mostly working with a much smaller model than gpt3.5.