macrolocal

6 ms·

31.
by macrolocal 3y ago
Nb. Altman wants lenient regulations for companies that might leverage OpenAI's foundational models.
32.
by macrolocal 3y ago
These days English majors do tend to know about entropy, thanks to Stoppard.
33.
by macrolocal 3y ago
G.H. Hardy’s sense of humor was famously self-deprecating.
34.
by macrolocal 3y ago
One complication is that mainstream tech spends billions precisely to prevent them from doing that.
35.
by macrolocal 3y ago
I wonder why there isn’t more support for replacing possibly the entire industry with generative AI. Pornography is such low entropy content.
36.
by macrolocal 3y ago
“As weak as” means “as strong as.” But in your scenario, weakness includes failing to adequately limit political or moral liability internally.
37.
by macrolocal 3y ago
Although, silencing whistleblowers is as weak a strategy as security through obscurity.
38.
by macrolocal 3y ago
As long as you don’t bump into Gal(ℝ/ℚ) somewhere along the way.
39.
by macrolocal 3y ago
I say pictures of cattle mean “oxox” or “oxen” like pictures of Tom Hanks mean “thanks.”
40.
by macrolocal 3y ago
Yosida is a great reference for this functional analysis. For a much broader generalization, albeit with expensive concepts, cf. Tannaka-Krein duality.
41.
by macrolocal 3y ago
Yep, there’s probably a relationship between lack of creativity and overfitting.
42.
by macrolocal 3y ago
I don’t buy this gender dimorphism. For starters, immature and manipulative women are about as common as immature and manipulative men.
43.
by macrolocal 3y ago
“I can bring whole cities to ruin And still have time to get a soft-shoe in”
44.
by macrolocal 3y ago
Maybe for oversight and liability.
45.
by macrolocal 3y ago
Nb. There are plenty of specialized architectures for convolutions: Nervana, Habana, Groq, etc.
46.
by macrolocal 3y ago
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 th
47.
by macrolocal 3y ago
A 6D being even! We live from aglet to aglet.
48.
by macrolocal 3y ago
The point of quantum computing is that you get much of that matrix math for free .
49.
by macrolocal 3y ago
You couldn’t tie your shoelaces in 4D because the laces would just slip around each other.
50.
by macrolocal 4y ago
Sure, easily. On example is vision: you'd just have to adjust input preprocessing in existing models.
51.
by macrolocal 4y ago
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.
52.
by macrolocal 4y ago
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 experi
53.
by macrolocal 4y ago
Haskell is useful to stop other people from breaking your scripts.
54.
by macrolocal 4y ago
Probably the Universe is full of things even more amazing than life as we know it.
55.
by macrolocal 4y ago
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.
56.
by macrolocal 4y ago
Science is more like a willingness to admit you're wrong when some agreed upon standard is met.
57.
Deep Learning in 2020, 2030, and 2040(gilkalai.wordpress.com)
3 points by macrolocal 4y ago|0 comments
58.
by macrolocal 4y ago
How about a Rubik’s cube with Spin(3) rotations?
59.
by macrolocal 4y ago
For example, cf. ZionEX and Pathways. AI accelerators are important.
60.
by macrolocal 4y ago
Thank you!

More ›