Ah. Now Eileen-II is starting to show weird symptoms. For instance, this strange blank horizontal line of dead pixels. Really starting to make the Steam Deck enticing, even though its price of nearly SGD1500 is really painful to look at. I really hope that the fix for Eileen-III will be successful, and that I can put away Eileen-II. What I don't want to do is to keep travelling back in time, from Windows 10 that is on Eileen-II now, to Windows 8 on Edythe-III---that would be even more terrifying.
------
A thought came to mind recently as I was just riding the bus between points. If I claim that the two biggest super powers of humanity over everything else is the amount of [self-]discipline that we can muster, and the fact that we can figure out how to transfer knowledge at a rate faster than one life-time's worth in one-life time, then by extension, anything that can emulate those two properties will therefore give humanity a run for their money at being the apex.
Guess what is currently being trained on the carefully (and not-so-carefully) curated knowledge created by people at a pace that will make the puritan blush and claim the need to respect the Sabbath.
That's right---all these so-called ``large language models''. We train them with copious amounts of written material, and the training proceeds without stop for months at a stretch using hardware that may fail but definitely does not sleep.
So now, I am at a cross roads of reasoning.
If the two biggest super powers I claim are all there is that makes humanity superior, then I think we may be doomed a little as we are naively building our replacement at best, and superior at worst.
If the two biggest super powers I claim are not all there is that makes humanity superior, then I need to start thinking about what other super powers are there that sets us apart from fellow creations of God.
Modern philosophers have come up with the notion that what counts as something with a consciousness is one that has a certain aspect of embodiment, which is the idea of an integration of a vast variety of sensory information. More importantly, the claim is that intelligence that arises from the integration of this vast variety of sensory information is the definition of consiousness.
Now, we have two different problems here, most literally. These philosophers are arguing that humanity is not the only example of a conscious being, and that any other creature that does integration of large amounts of sensory information is also conscious. In other words, spiders are also conscious beings.
And if one starts instrumenting a trained large language model with all kinds of sensors (CPU temperature sensors of all contributing chips, fan speed sensors... you get the idea), then the large language model itself is also conscious.
Oh and that J-space thing from Anthropic that seemed to be very related to the ``global workspace'' concept? Totally related then.
And that's the first problem---we effectively found some kind of equivalence relation between humanity, animals, and large language models residing in their natural ``body'' of a data centre in terms of consciousness.
The second problem then is that the next super power I claim that sets humanity apart from all therefore cannot be that of embodiment. Perhaps the true super power here is the depth of care and cooperation that many clearly not-the-same-individuals are willing to have together, to form a self-supporting society.
``But MT, what about the whole `agentic framework' use case of the large language models?''
If you think carefully, the ``agentic framework'' is less of a cooperation of not-the-same-individuals, and more of a self-consistent role-play attempt borne by the same large language model, but under different prescribed roles. In some fundamental sense, the same large language model shares the same foundation from which all inference is formed, but the same cannot be said of humanity given our current understanding.
The equivalent approach in the world of the large language model might be to have different large language models self-organising to work towards a common goal, instead of relying on a strictly hierarchical or even [externally] orchestrated approach.
That said, I've not really thought much about this whole gotong royong aspect of humanity as a major super power, primarily because it seems like something that is trite---many animals help each other, so what makes humanity different through possessing this power? At this point, it seems more like what the modern philosophers are trying to do---separate humanity from the so-called artificial intelligences.
But I don't want to just do that. I want an answer that separates humanity from both artificial intelligences, and the rest of the animals. ``Just because God created us in His image'' is great for a faith-based explanation, but that's a cop out in this case, because it is reality that we are the apex of the world right now.
If we knew why and how, maybe we can have a more rational response to the rise of the AI.
Till the next update.
No comments:
Post a Comment