6 Comments
Oct 20, 2023Liked by Étienne Fortier-Dubois

So, if I can ‘make’ a fire I ‘understand’ it?

Especially the ones who know the most about how this place functions admit having no clue. If we really lose the ability to see the magic of lightning because we think we’ve got it, then we are definitely lost.

If you grow up through dismantling your childhood bed and thus proving ‘there are no monsters’ you didn’t yet get the magic stories you were told.... and better prepare for some nasty surprises.

Picking up wands and using them without a long apprenticeship...is exactly what Goethe tried to point at. It is quite okay to not understand and live in awe and wonder. It is horrific to use the excuse of knowing to destroy the old tree....

This is a good discussion to have but it is not just about AI, a hammer is no less magical than AGI. Any bird or insect, any cell outcompetes our clumsy attempts at recreating parts of that. We are the most stupid of all beings if we do not start to ‘understand’ our limitations. The intelligence of reality is indifferent to our IQ.

Are you familiar with Iain McGilchrist’s theory? Very simply said he states our brains hold two opposing views of the world. One produces technology, uses maps, representations, calculations, plans, solidity, opaqueness. The other experiences the world through the senses, sees similarities, patterns, wholeness, uses metaphor, humour, poetic language, makes translucent, is capable of clarity without reduction. Only one of the two includes the other. And that’s the ‘one’ that should be king but no longer is in our society.

Sorry, this rant is not directed at you, I enjoyed the post. Good spotting of a trend. This is just very alive in me, and the outsiders view of magic feels a bit triggering.....dealing with the monsters under the hood of technology...with the unknown at the edge of the map....keep going....

Expand full comment
author

I do enjoy calling the things of nature "magic", but they're rarely called that in any seriousness, especially since we do have pretty good models of how most of them work. At least, mechanistically. There are large, complex structures that we don't quite understand, though, and I wouldn't balk at calling them magical (see https://etiennefd.substack.com/p/preamble-to-a-psychofauna-bestiary). In any case, this post was mostly about human-made capabilities.

Expand full comment
Oct 20, 2023Liked by Étienne Fortier-Dubois

Great post. I'll pick one nit: You say.

"The large AIs of today — large language models like GPT-4 — are the ultimate incomprehensible technology. They’re made of giant deep neural networks, containing millions of “neurons” and a similarly large number of connections between them"

From my research, this is incorrect. Large language models run via software on standard computers, not neural nets. In fact, I just posted about this yesterday: https://speclectic.substack.com/p/sentient-aisyes-no-when

I especially like your conclusions though. I think we may well be headed for a magical AI world.

Expand full comment
author

Large language models are complex pieces of software, but they absolutely do incorporate neural networks (specifically, transformers). Of course, all neural network programs run as regular programs on regular computers. Note that "neural network" in the machine learning sense has only a tenuous connection to brains and neuroscience.

Expand full comment
Oct 20, 2023Liked by Étienne Fortier-Dubois

... the EU has banned glitter.

Expand full comment
author

Now all they have will be virtual AI glitter. Software eating the world once again

Expand full comment