vlovich123

Why is it so surprising that continuous measurement is likely to catch a failing system? The more surprising belief is that sporadic random measurements at random times give you good enough predictive powers. Like all the blood work and vitals we do at the doctor’s office - I’m sure it captures some problems but more subtle deviations from baseline are much more difficult to suss out.

While I don't agree TigerStyle inherently actually helps you with this, design and architectural mistakes don't usually get solved in the way you're describing. You sometimes have to rethink how to structure your entire codebase. Sometimes you can evolve an architectural improvement but sometimes it requires a complete rewrite and there's no way to get outside of that path dependency.

The problem is that “Python” the language includes native poorly isolated FFI interfaces exposed by CPython. So the language isn’t “Python” but CPython. That’s why alternate implementations like PyPy haven’t seen adoption despite having a very mature JIT while languages like JavaScript have JITs that put CPython’s performance to shame.

Training costs haven’t plummeted like they have for inference afaik and the efficiencies are just translated into running larger and longer training. Basically Jevon’s paradox in action just like on the inference side.

The problem is that clearly documenting all details as required in the PEP is at odds with complicated new features like a JIT where all the details can’t be known and have to be decided on as they are realized. Process / bureaucracy is at direct odds with experimentation.

Funny you mention probabilistic algorithms. If I’m not mistaken one of the OpenAI results is that most of the probabilistic methods (specifically those that use a logarithmic amount of space)-4 have non-probabilistic versions that are as efficient in time and space (L = RL = BPL)

Short term benefit for long term harm. You’re handing the power and clout to a hostile nation that is literally underwriting the competition’s ability to employ this strategy. Its economic warfare and you’re rooting for the away team. Even if you’re not an American, I’m not sure I’m convinced the scenario of Chinese dominance around AI is good for other countries.

Training is really really expensive and open models are not getting funded by VCs. Distillation also doesn’t improve performance. Until training gets exponentially cheaper so that you can do it for a few hundred grand to train a competitive model, open models are labs trying to flood the market to get market share / undercut more successful labs, not a sustainable strategy that will extend indefinitely necessarily. It’s like claiming Uber and DoorDash are fantastic because their prices are so low - once they eat the market the prices become higher than ever.

I feel like people complain about “food” but it’s a stand-in and that the cumulative increase in rent, transportation and food probably rose faster than wages. So Cato reply “but food hasn’t grown as fast as wages” is a myopic retort (admittedly one they acknowledge) that tries to fight the narrow factual argument ignoring the broader one where the emotional narrative probably remains true.

Didn’t know this but glad I’m in such distinguished company.

Many worlds is a philosophical interpretation akin to god. It doesn’t add anything to our understanding and instead invokes some kind of supernatural untestable phenomenon.

I never said anything about quantum gravity. String theory is also trying to bridge things. But many worlds is pointing to a gap in the math and claiming god. All I’m saying is we don’t have any model or math and many worlds is the more fantastical explanation.

And so as a result there’s multiple worlds? That’s the simpler more testable prediction?

I could be wrong but my understanding is that’s purely a function of our technological limitations in how we create antimatter, not in the fundamental energy required. Our tech of a particle acceleration is ~1e9 times less efficient than the actual energy required because most of the energy is lost, not converted. A gram would be $~6.2M/gram if we could get to 1% energy conversion efficiency. No one bothers of course to even try because it’s not worth it - expensive, difficult, and huge energy losses. Fusion reactors for space ships will probably come before antimatter gets cracked as a battery source.

“Starts to play a role” means when the effect of gravity becomes significant to the quantum equations which currently don’t have any placeholder for gravity.

Just like Newton’s equations are accurate to describe gravity and motion but only because the full equations Einstein worked simplify to them at small speeds.

So similarly it would make more sense that quantum equations and relativity are two sides of the same coin and there’s some missing mathematical bridge where at small particle sizes it simplifies to quantum and at large particle scales gravity matters and we get Newton and then further relativity.

Using Occam’s razor that’s at least my prior vs some undefined mechanism of action where all possible interactions of every particle actually creates an infinite superposition of universes where that did happen and you get an infinite number of such superpositions ad infinitum. Much easier to believe the huge gaping hole between two very accurate theories explains how quantum mechanics turns into classical mechanics, especially since we know gravity exists, what it is, and quantum mechanics can’t explain gravity.