Manabu-eo

How likely this problem is already on the training set by now?

Related to overbuilding, vertically mounted solar panels can help flatten the generation curve during the day, and may perform better than "optimally tilted" panels on winter, especially where snow might otherwise be a problem.

"I believe in empathy, like, I think you should care about other people [...] empathy is good" — Elon on Joe Rogan

"I believe in empathy, like, I think you should care about other people [...] empathy is good" - Elon Musk

I saw the argument that the source code is the preferred base to make changes and modifications in software, but in the case of those large models, the weights themselves are the preferred way.

It's much easier and cheap to make a finetune or LoRA than to train from scratch to adapt it to your use case. So it's not quite like source vs binary in software.

> the basic controls of fully automated vertical landing were directly demo'd in a real flying 1/3 scale test bed (though important to note not an orbital one) with the DC-X in 1993.

Surveyor 1 was the first automated vertical soft-landing rocket AFAIK, in 1966. DC-X was the first using turbo-pump engines.

Framework entire brand is to let you upgrade parts on your own, and they said CAMM was technically impossible because signal integrity.

You just need to use two CAMM2 to get 256-bit bus, just like what you do with regular DIMMs when you need more channels.

You are making a big logical jump here. I only gave one company as example because that is enough to disprove your previous post.

Also, before you thought he knew very little about his many companies, implying no distinction, but now you adjusted up his knowledge about two companies, but inexplicably down for the others.

You also imply he should give equal attention to all of them, ignoring some of them are bigger, more important, or simply more interesting to him. Is equal attention the optimal strategy here, or you would be getting an F grade if you suggested that?

He didn't need to invest a lot of time to make a good investment in DeepMind, that was then bought by Google, for example. Investing in what you know and understand is a good investment advice, but so is to diversify your portfolio and to not spend too much time optimizing your investments in lieu of everything else.

Some of his "investments" are more like spending on a hobby (as destructive as it can be, in the case of twitter for example... or constructive like SpaceX), so not even bound by those rules...

He has a degree in Physics, that is like half of any engineering curriculum. Before funding SpaceX he hired several industry consultants to educate him, indicate aerospace engineering textbooks to study, etc. And then he had about 6 years of experience as almost full time CTO and CEO of SpaceX, until he had to divide his attention with Tesla. And somehow, after he and the SpaceX team achieved what dozens of other teams with more funding failed, he "understands nothing"? No need to be "the faster learner in the history of mankind".

Someone being capable in one field doesn't means he isn't a insufferable jerk or a moron in other fields. I don't understand this impulse to paint someone as completely black or completely white.

While not the main focus, see Section 6.1 and Figure 10 for a simple adaptative exit strategy for inference.

I imagine that they choose a fixed number of recurrent iterations during training for parallelization purposes. Not depending on the previous step to train the next is the main revolution about transformers vs LSTM (plus the higher internal bandwidth). But I agree that it might not be the most efficient model to train due to all that redundant work at large r.

Or how a lot of time is wasted. For example on perpetual motion machines and infinite data compression.

He is talking about Falcon 1, not CCDev. There was no close call at CCDev, nor any grant money for Falcon 1.

Most of that $5 billion in the LA times article was not received by Musk's companies at the time, and I'm not sure if they ever received the full amount as it was contingent on reaching several milestones in the span of 20 years.

A big part of those $5 billion was not having to pay sales taxes on possible future Gigafactory production, for example, but that gigafactory never reached the initally planned size AFAIK.