This is the George Hotz MO: I trust him to create a lot less than I trust him to optimize, and I trust him with optimization only in a very narrow technical sense. "Optimizing" Twitter by reducing headcount to 50, regardless of the social or revenue consequences, is actually stupid. Optimizing a whole mess of software that exists between your tensor and the hardware is a decent idea.
The math isn't super difficult. Some books will try to throw a mess of differential equations at you, but some simple calculus is all you need for backpropagation.
I have been through the math thanks to the youtube videos by A. Karpathy. Deriving some of the differentials, e.g. for batchnorm seems fairly hard (hard as in slogging through something with many steps where you can't make a mistake at any step). But the principles are quite simple - I think by design. If they were hard to compute or reason about then the neural net wouldn't work very well!
Define "fix." Twitter is a tool used by extremists to reach massive audiences. Some of those extremists have socially destructive and violent objectives. Responding to harassment and hate speech involves understanding the psychology of harassers and racists so that incentives can be designed not to reward those behaviors. IMO, "fixing" Twitter isn't just about fail whales and 500 errors.
It is only "painfully clear" to those who know nothing about it. Asking anyone who worked with him says otherwise. Examples include engineers like Tom Mueller and other early SpaceX people, astronauts like Garrett Reisman and many more. Or don't believe them, watch any in-depth technical interview with him. There's three-hour long videos on Raptor engines on YouTube where he speaks off-the-cuff on in-depth details about rocket engine design and hypersonic aerodynamics, among other things.
It's incredible to me that people try to insist (contrary to all evidence) that a serial entrepreneur who has become the richest man on the planet somehow lucked into his success and is really just a "showman."
Is this why VCs aren't investing in the area? Investment has been historically quite low for a new technology area, and it's so obviously the next big wave of technology. I've been looking for some explanation or series of explanation to adequately explain it.
Came to say this. The point seems not to be give an "all sides showing" but specifically to remove language that's designed to trigger our subcortical response systems.
It was always possible to do 18 pop feudal, but today's focus on luring deer is perhaps the biggest change to the meta?
There wasn't any real balance patch. Just people who have gotten much better at luring deer to improve early build orders. Like 3 years ago, you'd just assume 2x Boar + 8x sheep as your early food source. But today, people are adding 2x or even 3x deer to that.
As deer + boars collect food at the fastest rate, you can more consistently go up at lower villager counts. There's also much more precision: killing boars / deer / sheep under the town center more consistently today than before.
--------
TL;DR: Skill improved, people can do 18 pop today even though its always been possible. So people are just getting faster and more optimized.
Hera made a video about this recently and attributes it more to all the QoL updates like shift queueing. Sheep actually spawn next to your TC now. Before DE you had to find your first 4 as well.
Oh, one more thing. April patch changed Portuguese Organ Guns.
The goal of the April patch was to make organ guns more of an anti-unit shotgun, and less of a wall destroyer. So you'll see fewer organ guns on Arena (one of the few maps that start off fully stone walled).
To the degree these "solutions" (in all their varying degrees of effectiveness) did not solve the problem they intended and caused new harms, then yeah, those values and mechanisms of social control are bad.
