Phones, laptops, GPUs and many other modern highly integrated electronic products are repairable. Look it up, there are tons of videos on YouTube of people doing component level repairs. Leaked schematics make a huge difference here.
> Look it up, there are tons of videos on YouTube of people doing component level repairs. Leaked schematics make a huge difference here.
I've watched these videos — and the people you're referencing doing this are rarely using any kind of schematics at all to repair modern digital-logic boards. And not for lack of accessibility!
Modern logic-board designs consist of a few proprietary ICs, plus generic labelled support components (e.g. VRM caps, surface-mount resistors and diodes, etc.)
You can repair these boards, but these repairs fall into three categories:
1. bad solder joints or broken traces — which mostly just requires looking at the board carefully to notice these.
2. bad generic support components — which in theory you can determine the need for by testing with various multimeter modes across the individual component's legs; but more often you just notice that what the component is in line with isn't working, and "swap out to see if that fixes it." And which where such a swap-out can be done by just looking the part to figure out what it identifies itself as; then de-soldering it and soldering on a replacement.
3. bad proprietary ICs — which you determine by tapping the signal lines leading from/to the IC on an oscilloscope; and which you "fix" by buying other for-parts copies of the board, de-soldering those ICs off of the sacrificial parts boards, and soldering them onto the "almost good" board.
In none of these cases would referencing a schematic help! They're all effectively "context free" repairs — see, probe, think, do.
(A schematic can in theory help you to find test points to differentially diagnose 1 vs 2 vs 3 in the case where a board is failing mysteriously... but once you have some experience in board repair, you can get 80% of the same information by just staring at the board for a minute.)
---
Of course, if you're repairing a power supply, or an audio receiver, or some still-half-analogue electronic appliance from the 1970s — then yeah, schematics help. But these types of systems do still come with those schematics! (You just need to buy the thing directly. You aren't getting a schematic for an "embedded" PSU with a computer; but you do get the schematic if you buy that same PSU at retail at a Shenzhen parts-mall. And you get forwarded that same schematic [and more] if you make an industrial order of 10000 of them, too.)
What I meant is, most likely assuming that you are using pytorch / jax you could code down the model pretty fast. Just compare it to llama, sure it is far behind, but the llama model is under 1000 lines of code and pretty good.
There is tons of work, for the training, infra, preparing the data and so on. That would result guess in millions lines of code. But the core ideas and the model are likely thin I would argue. So that is my point.
Importantly, it’s lack of trade barriers but all the rest of the mix, too. Ie if one part of a bilateral trade deal doesn’t honor IP rights of the other side there’s problems. Also linking countries with hugely different wage levels can be destabilizing to the more developed. Those factors probably could have been mitigated.
Can you give me some examples of SpaceX innovations that disrupted the industry?
Here are some of my issues with SpaceX marketing:
1) Cost to NASA/DoD seems to be similar to ULA.
2) SpaceX continues to fly expandable boosters for missions above LEO and for heavier payloads. Reusable boosters have lower lift and range and require extensive refurbishment between missions. Upper stage is not reusable.
3) SpaceX hasn't flown a single mission into deep space or Mars despite this being the stated goal of the company. In the same time, others have flown multiple rockets to Mars and to other planets. We have multiple rovers roaming the surface of Mars and satellites orbiting other planes all while SpaceX has contributed nothing to these efforts.
I personally see SpaceX contributions to the industry are mostly incremental:
1) Booster reusability to LEO and for light payloads.
3) Human rated capsule for space station resupply missions.
3) High quality video content from launches.
4) Increased interest in space exploration at the expense of unrealistic expectations in relation to cost and timelines.
Depends on how much demand for those low orbit launches there is. A non-zero portion of those launches is Starlink, so basically in-house demand. Yes, I'd call that incremental. Which doesn't mean it is not impressive, because it is.
The main point so, IMHO, is that prices for comparable launches are not so much cheaper using SpaceX. Kind of like those Ryan Air 19.99 tickets. Sure, there are some, but on average the price difference is much smaller than people think.
Calling SpaceX SpaceRyanair seems unexpectedly accurate. O'Leary and Musk have a lot in common.
NASA's shuttle would have solved reusability in the 70s/80s, but a combination of DoD interference and severe funding cuts effectively crippled the project. This was politically fatal because it killed NASA's reputation as an effective manager of innovation. Which is why every project from the 90s onwards has been supported by slightly overwrought rapturous PR - to try to restore that reputation, because funding relies on it.
SpaceX has copied this.
So yes - SpaceX is incremental. A non-incremental company would be developing blue sky alternatives to chemical engines and/or looking for new propulsion physics. I can't see that happening without a government-run academic R&D program.
Booster reusability is massive, and if starship flies that will be quite a bit beyond that.
They are charging not much less than ULA for some launches because they can, and to recoup R&D expanses. The cost per launch to them is much lower.
I do however agree that technologically it’s incremental. Look up the DC-X. NASA and Lockheed had flown VTVL test boosters before. Werner von Braun talked about them. Starship is basically what he wanted to evolve Saturn V into.
(Not at all belittling the achievements of SpaceX engineers. Space flight is truly brutally hard even if you are doing what’s been done before let alone innovating at all.)
But none of that went anywhere because by the 1990s you had a stolid conservative industry happy with expendable rockets and cost plus contracts. It didn’t move until a lunatic swaggered in and turned over the table.
Bureaucracies and stolid industries generally have to be kicked and beaten and dragged into the future. They are happy to cash checks, moo, and chew cud.
Same goes for EVs. Marketable EVs have been possible since the GM EV1. There was demand too; GM had to pry those prototypes away from people. But the car industry was set in its ways. It must go vroom vroom or it’s not a car, and then there is the huge oil industry with all its lobbyists and PR. Again it took an unhinged lunatic to say “fuck you all we are doing this” and basically shame the industry into moving by showing them up.
If Tesla fails it will be because Tesla succeeded. The big car companies are probably better at making cars and may outcompete Tesla once they actually decide to move.
The same personality willing to offend millions to bring trolls back to Twitter is the same sort of personality it apparently takes to whip conservative bureaucracies and industries into action.
The only other force I am aware of that can do it is government mandate, and that carries a more literal whip. That wasn’t going to happen for rockets or EVs due to regulatory capture and conflicts of interest.
Edit: I guess the other force is war, which is why war has tended to bring innovation. But I’ll take Elon Musk types with all their negative traits over war any day.
As a side note I suggest that the issue is one of operational fidelity.
They do it in a reasonable time frame for a fraction of the cost, aka, a bit of '1st era' commercialization of Space - still relatively government related, but mostly private though nowhere near consumer or even general industrial.
They're just making what was previously 'research' to be economically feasible.
I don't see how his ventures could have made the progress they have with a business/finance/sales person calling the shots. Elon is adept at engineering; you can see it whenever he is interviewed and goes down rabbit holes about metallurgy, simulations, manufacturing, etc. The achievements of Tesla, SpaceX, Starlink and others would not have happened without thousands of correct evaluations of and decisions on technical matters, and Elon is the singularity around which all of that orbits.
