macrolocal

I’m enjoying “Uproar!” by Alice Loxton.

https://www.the-tls.co.uk/articles/james-gillray-tim-clayton...

1.4T according to Michael Douglas [1].

[1] https://youtu.be/B3Az2EONCHE

Nb. Altman wants lenient regulations for companies that might leverage OpenAI's foundational models.

These days English majors do tend to know about entropy, thanks to Stoppard.

G.H. Hardy’s sense of humor was famously self-deprecating.

One complication is that mainstream tech spends billions precisely to prevent them from doing that.

I wonder why there isn’t more support for replacing possibly the entire industry with generative AI. Pornography is such low entropy content.

“As weak as” means “as strong as.” But in your scenario, weakness includes failing to adequately limit political or moral liability internally.

Although, silencing whistleblowers is as weak a strategy as security through obscurity.

As long as you don’t bump into Gal(ℝ/ℚ) somewhere along the way.

I say pictures of cattle mean “oxox” or “oxen” like pictures of Tom Hanks mean “thanks.”

Yosida is a great reference for this functional analysis.

For a much broader generalization, albeit with expensive concepts, cf. Tannaka-Krein duality.

Yep, there’s probably a relationship between lack of creativity and overfitting.

I don’t buy this gender dimorphism. For starters, immature and manipulative women are about as common as immature and manipulative men.

“I can bring whole cities to ruin

And still have time to get a soft-shoe in”

Maybe for oversight and liability.

Nb. There are plenty of specialized architectures for convolutions: Nervana, Habana, Groq, etc.

Thar be dragons here, but here’s a nice overview by Chalmers:

https://youtu.be/-BcuCmf00_Y

Personally, I’m inclined to believe that consciousness requires an integrated attention-self-world model, but maybe not much more that that.

A 6D being even! We live from aglet to aglet.

The point of quantum computing is that you get much of that matrix math for free.

You couldn’t tie your shoelaces in 4D because the laces would just slip around each other.

Sure, easily. On example is vision: you'd just have to adjust input preprocessing in existing models.

No matter what your ontological commitments happen to be, imagining what an LLM's umwelt might be like, and how it might approximate ours, is a useful (and fun!) exercise.

I sometimes like to imagine the umwelt of an LLM. Tokens are their sensory atoms, and the statistical relationships between them form their entire reality.

Of course, one could easily build a model whose tokens line up with qualia we experience.

Haskell is useful to stop other people from breaking your scripts.

Probably the Universe is full of things even more amazing than life as we know it.

I've more often seen cruelty paired with a willingness to shrug off reason the moment it becomes inconvenient.

Without enlightenment values, disagreements almost always degenerate into power struggles.

Science is more like a willingness to admit you're wrong when some agreed upon standard is met.

How about a Rubik’s cube with Spin(3) rotations?

For example, cf. ZionEX and Pathways.

AI accelerators are important.