throwAGIway

Blue Origin is newspace, isn't it? It's definitely a big outlier among LM, NG, Boeing etc...

Yeah, a day job. I'm not in the US and there are no such quant positions that would accept me remotely and/or pay more than the remote-only US corporate job I have. But I'd probably avoid it even if I was in the US, I don't like long hours and stress.

I started to make much more money by just spending the same amount of time writing enterprise software. HFTs made it very hard to find a moat in technical trading, I don't believe you can beat them at their own game today; I especially don't think you can just make something and get passive income from it - that was never the case, it was always a job, though very flexible one.

I switched to stock picking based on the premise that the market at large might be momentarily irrational because it doesn't have the same insight into the tech industry as I do - e.g. I made good profits on AMD between $30 and $150, Nvidia $100 to $400, some medium term trades on Supermicro and TSMC.

It's influenced by the population wave caused by a communist president in early 1970s. The results are subpar if you normalize for that. There are a lot more adults in the age to have children right now than before, but it's going to drop hard very soon.

https://en.m.wikipedia.org/wiki/Hus%C3%A1k%27s_Children

Is there a single place anywhere on Earth where it doesn't suck for somebody there?

As I said, nowadays it's not working so well. I wouldn't do it - saying this as a former trader that used it successfully (15-20% returns p.a.) for more than a decade, but 20 years ago. Today I don't daytrade at all, just some stock picking based on my knowledge of the tech industry.

It's called Drive Pilot and the user experience of activation and usage is no different from a Tesla. It's just much less capable.

Technical indicators work because they work.

Really - it works because other traders decided there are price points of significance based on mathematical equations and act on it. Traditional technical analysis could be seen as looking for a set of technical indicators that optimally describe the amalgamation of TA and the capital it controls on the market at the moment.

But nowadays it's very hard to make it work.

It's called Drive Pilot.

I drove the Mercedes with officially regulated L2 self driving capability and it was much worse than the old Tesla model I tested years before... The current version is supposedly much further based on the online feedback, while the Mercedes is still bad.

> keep Musk out of things

Never heard that from people at Tesla or SpaceX.

I don't want to counter it. They have their opinion and that's fine, I don't agree nor disagree.

I'm trying to say that we don't need any further advances in the science of ML or computers - we could already do it with a monumental lot of engineering. ChatGPT is what I meant by saying "the dumbest system - raw input to raw output based on conversations".

Several engineering steps are necessary IMHO:

- go through the training data and clean it up - likely what OpenAI does right now, but it's going to take a lot more than they have done so far

- make it a complex system with thoughts, memories, reasoning loops, planning, goal setting, self introspection... - this is what many amateurs are trying but the professionals seem to think their time is better used elsewhere right now

- careful machine code-level optimization of the software for purpose-built hardware

So, no new technologies or science are strictly necessary - "just" man-centuries of engineering.

Right now companies and researchers seem to think that there are more gains to be made at the fundamental scientific/technological level. But if that doesn't pan out, it doesn't mean AGI will never exist - just perhaps later, more energy consuming and less powerful than they thought.

Just imagine how dumb you would look if everybody heard the first thought your brain came up with, you couldn't stop yourself from saying it, think about it to consider the implications and meaning - or at least be aware you're saying it... I'm a sleep talker and the bullshit I say without any recollection of it is just insane, and frankly very similar to a hallucinating ML model.

I truly don't think it does. AGI - a human level generally intelligent agent - is possible with today's technology. It'd take a lot of classical programming work - definitely man-centuries, but even if nothing new comes out and microchips stop progressing, it's possible. Look what can be done with the dumbest system imaginable - just raw input to raw output based on unfiltered conversations of internet idiots. I just can't see how it isn't possible if somebody invested the necessary man-centuries, which isn't that much - a mid level software house with few years of time.