Jobs aren’t distributed out of some cookie jar. They are needs and wants and obligations that other people will pay to have fulfilled or taken off their hands. Figure out how to solve those problems and you’ll have all the work you could ever ask for.
Move to a higher level of abstraction and architecture. We’re leaving the era of hand wiring data structures and program logic the same way we left behind the era of hand wiring ICs and discrete components. Different skills will be needed.
I’d rather be in the textile industry post industrial revolution than before it. The fortunes made during the age of mechanization make all of history’s kings and merchants paupers by comparison.
Everyone thinks they'd be the king and not the pauper. The luddites starved on the street because they were kicked from their properties with nowhere to live and no way to earn a living. The next generation of kids worked on the textile machines and commonly got turned to hamburger, all while the robber barons made obscene wealth. It mostly worked out over time because the populace fought things like unions and social safety nets. But hey, don't worry, the modern day tech barons are telling us we don't need those pesky 'expensive' social safety nets, I'm sure out of the kindness of their blackened hearts they'll provide for us all when robots replace our jobs.
What makes you so sure that problem solving and invention aren’t just engineering challenges that we can solve by combining LLMs with well designed algorithms? The way I see it, we’ve just discovered something fundamental like steam power or electricity and we’re currently in the very inefficient stage of brute force solutions like mine pumps driven by condensing engines that needed tonnes of cheap readily available coal and arc lamps running off an entire room full of galvanic piles. In other words, we’re just throwing shit at the wall and seeing what sticks; but stick it will and then we’ll quickly be on to locomotives and lightbulbs.
You guys buy too much into the marketable fuzz. In the last 15 years we've had multiple other technologies that were supposed to change the way we live (3D printing, Crypto, VR, EV, self-driving, now AI).
It's just the VC scheme: Over-promise/under-deliver = Profit
AI is and will continue to be a search on steroids.
You're buying into too much market fuzz too. That internet thing going big never happened and cell phones turned out to be bust, all that hype for nothing..... See how cherry picking works.
Not saying they are a bust, either of them. Just to scale down expectations because AI, even the generative type won't be coming up with novel solutions.
EV aren't a bust either, but case in point... manufacturers are already anouncing scale downs because expectations were too high. Combustion will stay around for quite a bit given the battery production constraints.
Rome wasn’t built in a day, all of those technologies will still be around 20 years from now and will likely be powering a lot of everyday stuff. Short timelines are hype, the technology itself is anything but.
Tongue in cheek but this has been considered and has resulted in experiments like tree of thought and various check your work and testing approaches. Thinking step by step is really just another way of saying make a plan or use an algorithm and when humans do either they need to periodically re-evaluate what they’ve done so far and ensure it’s correct.
The trick is training the model to do this as a matter of course and to learn which tool to apply at the right time which is what the paper is about wrt interspersed thoughts.
an MVP is the smallest subset of useful features to take the product to market in order to prove or disprove its viability
and the reason you want to keep it to the minimum is to minimize the time and money invested in it so that you can throw it away if it doesn’t work and then try a different product. You’ll probably have to do this several times before succeeding so your budget for each should be about 1/10th of your available time and money.
Exactly, they’ve reached feature parity with large expendable launch systems so they can piggyback paying customers with a high risk threshold (starlink) on flights they’d be doing anyways. Given their cadence this phase won’t last long, they’ll likely achieve at least one successful landing next flight.
They have also been eager to launch version 2.0 starlink satellites, but they don't fit in the falcon fairing. The first couple batches of those would be test articles as well, so I'd be surprised if the next starship launch doesn't a few on board.
They've been launching a version of the v2 sats that do fit in the Falcon 9 fairing and supposedly have all the functionality of the larger versions. The issue is that F9 can only carry ~24 of those at a time, which slows down the pace of expansion a lot.
Yes this would all be true, if it were true. It is likely to become true at IFT-4 but they are very demonstrably not quite where you say they are.
This was still a suborbital flight and they cannot do much of anything that is commercially practical on suborbital flights (like launch satellites, even if they raise their apogee). They appear to have not had good control authority in coast and reentry. They did not do a relight/deorbit burn test that is likely an obstacle to tackle before they can make orbital flights. I assume we'll get some confirmation about these things soon enough, but please, you can be optimistic without being hasty.
> This was still a suborbital flight and they cannot do much of anything that is commercially practical on suborbital flights
If they had flown a slightly steeper ascent and burned for a little longer (possibly a minute if not less), they would have ended in a stable orbit. Not doing that was intentional.
They do not need engine relight capability to reach orbit - plenty of orbital rockets exist that cannot relight their final stage.
Yes, but the point was that they can't launch starlinks or just about any commercially meaningful payload until they are reliably in orbit, and they can't reliably get into orbit until they demonstrate at least one relight, because they need to reliably re-enter the atmosphere for the reusability tests.
So they are at least one more launch away from launching starlinks.
> commercially meaningful payload until they are reliably in orbit,
Which then can do without a relight.
> and they can't reliably get into orbit until they demonstrate at least one re-light
And part of testing deorbit/landing capability includes testing that they can relight the engine.
So they could launch the next one with Starlinks (possibly test articles of those as well since no full-size V2 satellites have been laucnhed yet). Get it into orbit and include a deorbit burn/re-entry as part of the flight plan. If the latter part somehow still does not work out ... they still got Starlinks into orbit. And they now have more data to fix it on the next flight. They already have several vehicles lined up for static fires and flight tests.
They won’t put starship into orbit until they can test relight. They won’t risk, nor would they be allowed to risk putting it up there without a demonstrated ability to bring it back down in a controlled manner.
They can test relight, but my point was that they wouldn't do it with Starship in orbit, because if it doesn't work, then they have no idea where it's coming back down. But maybe that's not as much of a problem as I had assumed given what's stated in the FCC application.
> They do not need engine relight capability to reach orbit - plenty of orbital rockets exist that cannot relight their final stage
I’m sure they will have this thing plez dispensing cybetrucks all the way to mars in short order. But I’m also pretty sure that the FAA and whoever SpaceX buys their insurance from are gonna need pretty good assurance that they can control where that thing comes down. Expendable orbital stages are designed and engineered for safe end of life operations and uncontrolled reentery. Starship is not.
> They do not need engine relight capability to reach orbit - plenty of orbital rockets exist that cannot relight their final stage.
I don't think SpaceX is interested in having uncontrolled Starship re-entries. It's large enough that (even without a heat shield) debris will very likely make it down to surface.
If they passed a law requiring US companies operating in Europe to divest their ownership in those subsidiaries or be banned from operating there then that would be their right as a government. Isn’t the EU rather famous at the moment for forcing foreign businesses to comply with their laws and regulations? e.g. GDPR
That said, so long as you don't have data centers in the country, what can they realistically do to stop you if you have a VPN?
There's nothing illegal w.r.t a private citizen trying to circumvent censorship afaik. If you're trying to use illegal content or services (hitmen, drugs, child porn, non us compliant crypto such as finance etc), then I can see the retribution, but unless you go full middle east/CCP people will still get access to and repost it
You need a data centre if you want to run the bleeding edge SOTA models but the weights of these models are fixed which means it should be possible to instantiate them in hardware and once you have a model smart enough to do most things you’d be crazy not to build a factory to churn out little boxes that you can connect to literally anything to give it a natural language interface and useful levels of intelligence. Mark my words: your doorbell will have an LLM in it in less than 10 years.
