mlsu

30 ms·

481.
by mlsu 2y ago
I wonder if there is some kind of mesh that you could put on the bottom of the pool to absorb the interior waves. Sort of like soundproofing in a recording studio.
482.
by mlsu 2y ago
It’s a slog to do the “which one of these is least a ripoff” calculation, at the grocery store or online. I try to do the calculation nonetheless. But every time I just need to buy some ordinary staple, I’m up against the Relentless Spirit
483.
by mlsu 2y ago
What else can be done? If you are sitting on 1 billions $ of GPU capex, what's $50 million in energy/training cost for another incremental run that may beat the leaderboard? Over the last few years the market has placed its bets t
484.
by mlsu 2y ago
But that's a semantics game. Sub probability for entropy. Why do we live in a world where low probability states were in the past and high probability ones are in the future? What intrinsic property of the universe causes this asymmetr
485.
by mlsu 2y ago
Oh wow, that's wild. It's like a cuckoo that has evolved to mimic Steve Jobs, just enough to feed on VC and beta product dollars. I'm sure it'll take off out of the nest, fat and happy, once its brood parasitism is compl
486.
by mlsu 2y ago
this is great and it does make perfect sense to me, someone with no stat thermo background. but it makes me wonder how is it then, that things are becoming more spread out over time? what is the property about the past, that it seems have a
487.
by mlsu 2y ago
Broadly, what marketing is saying is “AI.” There is huge value being created with deep learning today on internal systems. Recommenders, machine translation, computational photography… it is huge, improves people's lives, drives revenu
488.
by mlsu 2y ago
there is a lot of consumer surplus, but it’s very unclear if any one of these companies will be able to capture it. Especially if mostly good enough commodity LLMs like Llama 70b are in the mix. Where is the differentiation?
489.
by mlsu 2y ago
Put a number on it. How much value of this will they capture from you personally (we'll assume, very very charitably by the sound of it, that you represent an "average" user of AI products) when this market matures? Exactly h
490.
by mlsu 2y ago
I would almost invert that statement. Sorry if this comes off ranty, but what exactly are people doing in the "AI space" currently that isn't "undifferentiated spam/chatbot" being sold to non-techies who hear
491.
by mlsu 2y ago
> Hardware accelerated video blocks often see plenty of reuse across a GPU lineup. I’m surprised that Nvidia spent extra engineering resources to make a completely different video engine for Tegra X1, rather than reusing the one in mains
492.
by mlsu 2y ago
I love that turn of phrase. Insects or near-automata. Describes it perfectly. LinkedIn -- like a floodlight in a swamp.
493.
by mlsu 2y ago
The labor market is a market like any other. Some markets are competitive and require sales (and sales looks very different in different markets). Some markets have differentiated products or services and some markets have fungible commodit
494.
by mlsu 2y ago
This might be a cynical view of things, but I think it's planned, rather than a happenstance result of dysfunction. Gosh gee Willikers, the fellers in congress just can't get anything done ¯\_(ツ)_/¯ It's no coincidence t
495.
by mlsu 2y ago
"OpenAI is an AI research and deployment company. Our mission is to ensure that artificial general intelligence benefits all of humanity." Is this like EA where they have to make a trillion dollars first before getting to the actu
496.
by mlsu 2y ago
the cost of solar/wind depends on how much solar/wind is actually deployed. 1kWh of solar delivered midday, when there is 20% penetration? easy peasy. 1kWh of solar delivered at 2AM, when there is 65% penetration? much much more d
497.
by mlsu 2y ago
That note about manufacturing is very interesting. It is indeed very difficult to make 10s, 100s, or 1,000's of things. Let alone millions. At my company, we (software group) took a tour of the hardware R&D lab. They have a mechani
498.
by mlsu 2y ago
I suppose not. His expeditions to the Antarctic failed their objectives. His business ventures were never successful. And he died in debt. BUT. He successfully sailed one of the greatest and most perilous ocean journeys ever in recorded his
499.
by mlsu 2y ago
I'd love to ask them why they made so much of it, and why we are making the same mistake.
500.
by mlsu 2y ago
Yes, indeed. It took a while, but not that long in the grand scheme. A few million years; plastic is more concentrated and more delicious, energy-wise. I forgot to mention my theory, which is that we will be the ones who dig up the plastic
501.
by mlsu 2y ago
This lead me to google QQ's emoji set. Is this the one? https://www.emojiall.com/en/platform-qq If so, absolutely. Kinda seems like every East Asian emoji is more... expressive, cute, dynamic (?) than its western
502.
by mlsu 2y ago
Yes, certainly. The value of plastics is incalculable. (Personally, if there were a plastic-eating bacteria introduced tomorrow, I would be dead within the week, rather than dead within a month or two like most people). I don't really
503.
by mlsu 2y ago
Plastics are a store of energy. They could be used to sustain life, once evolution catches up. That's why I don't feel particularly bad about throwing plastic into the landfill. I'm taking an energy loan out against evolution
504.
by mlsu 2y ago
If that's the only thing LLMs are good for, my money for improving software productivity is in good old fashioned developer tools. A better language reduces boilerplate. A better compiler helps you reason about errors. Better language
505.
by mlsu 2y ago
I recommend reading Lynn's story, written in her own words: https://ai.eecs.umich.edu/people/conway/RetrospectiveT.html It is amazing, tragic, and triumphant in so many ways.
506.
by mlsu 2y ago
Plenty of people were skeptical of the iPhone when it came out. On launch, it was very very expensive, carrier locked, and relatively lacking in features compared to other phones. The app store was not in place. You had the Apple apps and m
507.
by mlsu 2y ago
This is interesting: > Autoencoder family > Note: Only 65536 features available. Activations shown on The Pile (uncopyrighted) instead of our internal training dataset. So, the Pile is uncopyrighted, but the internal training dataset
508.
by mlsu 2y ago
It stems from a more fundamental misunderstanding about how allocation of resources works. If a space program costs $10b and you suddenly defund it, you don't get $10b to spend on other stuff. You get a bunch of unemployed people, most
509.
by mlsu 2y ago
I have been speculating on Nvidia stock. It has been following an exponential curve. I have curve fitted the stock price, and determined that: - in 2027, the market cap of Nvidia will be approximately $10 trillion - approximately half of US
510.
by mlsu 2y ago
I read Pale Fire on the plane recently. Picked it up randomly, knew almost nothing about it. It is absolutely riotous. Most of the references probably went over my head, but I was immediately annoyed, then grinning about how annoyed I was,

More ›