Self-driving cars have proven harder than expected, but given the rest of Musk's tale, I wouldn't bet against him. SpaceX is already launching satellites, and Starlink is already providing service. T-mobile owns the frequencies for it, it's "just" a matter of building the satellites and putting them into space. If the claim were more outlandish, it would be harder to buy, but the technology is eminently doable.

Self driving is one of the hardest, if not the hardest software problems ever to be solved. I think people underestimate just how challenging the problem is. Must says a lot of stupid stuff but in reality he has delivered on the vast majority of it, though as he says it's not always in the time frame he expects.

>I think people underestimate just how challenging the problem is

Is this the shifting of the narrative?

"People" aren't underestimating the problem. It's widely known to be an extremely difficult problem.

Musk said he'd have it done by now, and actually sold the technology. People have been skeptical of it for the better part of a decade.

I remain skeptical Tesla is anywhere near let-the-driver-sleep self driving (except perhaps in tunnels), but I’m very glad they’re trying to do it, and do it at scale for consumer cars, not just company-owned hyper-sensored vehicles that cost as much as a house.

Literally every company working on the problem a decade ago was over-optimistic, but many pursued a strategy that would keep the technology only for company-owned vehicles, which I find less satisfactory. The whole point of personal cars is independence. We already have buses and transit vans if we want mass transit. If you want self-driving to actually save all the lives it supposedly could, it has to be scalable to everyone.

Musk is terminally optimistic on AI generally, including paranoia about AGI. I don’t expect let-the-driver-sleep self driving for a good 10 years, but it’ll be never if no one tries. The technology does not advance by itself.

People are already sleeping while the Tesla drives, albeit illegally. Highway driving it seems to do really well, around town though sleeping seems way off.

People underestimate it all the time, just because you don't doesn't mean other don't. If you read any forum they talk about how it should have been solved by now.

Wouldn't something like Westworld road system easier? Basically, almost all the cars are self-driven, and it's more like a public transit than anything. Essentially, a closed system. (Easier in the technical sense, not the implementation (politics, ...)).

Even if that were easier, there’s basically no way to get there without full self driving capable of working around human drivers. The reason why cars are convenient is due to the existing road network, and any deployment of self driving cars is going to have to function on existing road. If you tried to segregate the system, you would have to close some of the existing roads but this would mess with regular people and result in a very small network of pure self driving roads.

Just a couple days ago they announced that they will also be using Falcon 9 to launch the new Gen2 Starlink satellites, instead of the original plan to launch all of them using the still-in-development Starship. So it seems like Musk is pretty committed to doing this in a reasonable timeline, with smaller technology risk.

Starship's orbital test may happen somewhere in the next 6 months, so until they succeed with that Falcon 9 is all they have (it's plenty though).

It could be earlier than that, but they won’t deploy Starlink satellites to a permanent orbit until later flights, so I agree.

And Falcon 9, with droneship landing and fairing recovery now routine, is really a workhorse. Remarkable how much they fly it.

F9 is the backup plan, but they're still planning to launch within a few months on Starship.

You'll know Starship is about ready to launch after they've completed static fires for both stages.

Building and launching them is the easy part.

SpaceX will need to get FCC/ITU approval for the G2 satellites. T-Mobile is seeking agreements with international carriers. I think those are probably required for the proposed system to be viable. Spectrum management bureaucracies in other countries, and opposition from competitors will slow things down. It will require a lot of money to get worldwide operating agreements in place, and some of the money will be bribes to the right people.

Yes, I expect this to be one of many announcements going forward, with Amazon Kuiper and OneWeb coming into the party -- eventually. However, SpaceX has the advantage of vertical integration (they launch their own rockets) which substantially helps with costs.

Also the benefit of time. Kuiper hasn't launched anything yet, and all their launchers are still in development - BOs own rocket New Glenn, Arianne 6, and ULA Vulcan. Vulcan probably isn't too far off, 2023, but they will need time to ramp up production. It'll also be a lot more expensive than for SpaceX.

Amazon will probably have to heavily subsidize Kuiper.

I reckon they hope the market is big enough for multiple providers.

Musk has had some failures (FSD, battery replacement, solar roof, satellite laser communication), but also notable successes (supercharger network, landing a rocket).

While this particular idea seems theoretically possible, as far as I know there haven't really been proper tests of it yet. So I could imagine it theoretically working, but being really bad (which may honestly be enough).

There's probably a reason Apple seems to be doing hardware adjustments for their satellite communication.

Satellite laser links have been deployed haven't they? And all new satellites going up have the laser links on them, I thought.

Not sure how battery replacement is a failure? They got some money to do it, decided it was not worth doing, so they didn't. I mean that's not really a technological failure, but simple a good business choice.

Laser links exist.

>Musk has had some failures (FSD, battery replacement, solar roof, satellite laser communication), but also notable successes (supercharger network, landing a rocket).

Yeah, of all the projects of Musk's to be skeptical of, this one seems the most plausible and mundane.

No, self driving cars are as hard as expected.

Even in its current state with FSD beta my car drives itself 95% of the time.

I'm not sure what software was present on this random Lyft I took a couple months ago, but I was seated in the front, watching Tesla's infotainment screen the entire ride, which on one side was displaying a sort of wireframe (not the right word, but they weren't realistic drawings) of objects in front of and around the car as we were driving.

I was alarmed to see how often pedestrians, and even cars passing while waiting at a traffic light, would just randomly wink out of existence while still in the car's path, or nearby. Sometimes it would show cars parked at the curb to the sides, and other times nothing, even with our car being the same distance away. And yet somehow at one point it consistently displayed a traffic cone that was on the sidewalk against a building. Didn't really give me much confidence in Tesla's self-driving software if the object detection bits can't even get that right. But again, I don't know what software was running on that car, may not have been the latest-and-greatest.

I don't think I'd trust even the FSD beta to drive my car 95% of the time. Or 1%, even.

The display is not a great indicator of what the car "sees". Ideally the display would reassure the passenger, but you could totally solve self driving without solving this UX problem.

It's a beta so you shouldn't trust it. I'm just saying, I've driven thousands of miles and 95% of the time no issues. The other 5% I almost always know where it will have issues and manually take over before it even gets into a bad situation. A lot of these bad situations I think are easy fixes that will be resolved with future updates. I agree the object recognition is a bit jumpy, but it hasn't really affected the driving afaik, and I'm sure over time that will stabilize. It is really cool Tesla shows you exactly what the car sees. In other self-driving systems the user usually has no idea what the car is seeing. Tesla could do that as well, but this way it is much more honest. I'm looking forward to getting the latest release which is able to recognize the 3d model of unknown objects and avoid them.

>I've driven thousands of miles and 95% of the time no issues. The other 5% I almost always know where it will have issues and manually take over before it even gets into a bad situation.

This is not Full Self Driving, or remotely close to it. Having to take over for any tricky situation is antithetical to FSD's purpose. Besides the fact that "thousands of miles" is a microscopic scale of driving. That's a vacation for some people.

>A lot of these bad situations I think are easy fixes

Ah yes, easy fixes. Just have to catalog all the "bad" situations!

Are you a developer? Everyone knows the last 10% takes 90% of the time.

Though in this case I would say the remaining problems in FSD become exponentially more difficult the closer to 100%.

Why am I critical? Because Elon Musk reminds me of bad bosses and product managers in my career. They act like PT Barnum to the public and throw tantrums internally.

I also strongly dislike Musk's rejection of Kanban and Toyota Production System principles.

After all, Musk was fired as CEO by the PayPal board after Engineering mutinied against his ill-advised plan to migrate to .net and Windows.

Err the best engineers in the world work for him and want to work for him. Engineers run his companies. They’re given crazy ambitious goals and the resources to go do it. Engineers love challenges and making the impossible possible. Robots, fast cars, and rockets.

You keep saying things like that, but from your comment history you seem to be a huge Musk fan, so color me dubious. So what data do you have on 1) the best engineers and 2) their preference for Musk as boss?

Yet more tantrum like behavior with Twitter and Tesla return-to-office has definitely soured many of my colleagues recently.

Glassdoor reviews give Elon a high rating and many of the reviews mention how the people are the smartest they’ve ever worked with.

That sounds like it was probably the standard Tesla display, rather than the FSD Beta display? My understanding is that it's a very different system. But yeah, the standard display has pretty laughable object persistence.

I was interested in FSD but autopilot is so shitty and unreliable it has convinced me to not buy the upgrade. If they want to sell an expensive upgrade from autopilot to FSD they need to make the autopilot flawless at its job.

Self driving cars are indeed still not there. But the right question to ask about those is who has the highest chance of getting those working first? There are a few companies working on this problem. Tesla is by far the biggest of those companies and they are throwing some serious money at the problem. It's not a safe bet to make but the odds are still pretty decent that they might get there before anyone else does. Their strategy for this seems pretty sound. But obviously it's taking longer than they hoped. A much less safe bet to make would be betting it will never happen.

As for SpaceX, their reputation for getting things done is rock solid right now. If they manage to get starship going, that will be a major leap in launching capabilities. But even just getting the star link network operational, or delivering people to the space station have been enormous accomplishments already.

The only thing that's remarkable about this announcement is how unremarkable it is relative to Elon Musk's other accomplishments. It's a radio on a satellite that connects to a phone. Doesn't sound like rocket science. I'm sure the antennas etc. are amazing but that is essentially commodity technology that he's simply launching to orbit. Doing that cheap enough that stuff like this makes sense to even consider doing is the true accomplishment.

> proven harder than expected

Expected by whom? People who didn't understand the problem presumably.

After hyperloop and tesla semi, I don't trust Elon that much.

Curious that the Apple event next month is themed "Far Out" with space imagery. I imagine this is the modems that were widely reported last year as being part of the 13/14 SOCs

I think this is very interesting and a smart move for Starlink to improve profitability options in the long term.

The rule of thumb for Apple Event invitation themes is to just assume they mean nothing—or that they’re just screwing around with people who like to read them like tea leaves.

The technology has already been demonstrated in a similar form by Lynk. The large gen2 satellites have already been built, and you can see the deployed fitted into the Starship prototype in Texas, which is planning an orbital flight test shortly (low single digit months is the desired plan, but don’t get too excited until the static fire tests have been completed on both the booster and the upperstage). They have several other Starships being built as we speak, you can watch them on livestreams on YouTube right now, and these will be the ones that start deploying the new satellites (the first Starship will not enter a permanent orbit as they want to ensure that if there’s a problem with Starship it doesn’t stay in orbit as debris… like Shuttle External Tanks, it’ll splash down in another ocean).

Garmin inReach but works on any phone and a bit more bandwidth so not restricted to emergency messages. Also free with most T-Mobile plans (inReach is currently expensive, but may drop in response if that isn't a capacity only issue).

> Garmin inReach but works on any phone

That's because it's an app that connects to an Iridium phone via BT.

No he's asserting that this is like Garmin inReach, but is higher bandwidth, magically works with any existing phone, and is free.

Which we all know isn't going to happen. 80% of the commenters here have fallen for the banana in the tailpipe. Again.

> Also free with most T-Mobile plans

I can't find anything suggesting a bundling or partnership between garmin inreach (iridium) and T-mobile. inReach has a separate subscription.

Doesn't this require specific hardware?

They claimed it will work on most existing phones, but they did caveat with "aspirationally" all over the place.

If it’s send-only than a satellite receiving a ping isn’t insane - they’re only 340 miles up.

It's not send only. It's a standard cell phone service.

I didn't hear any claims of "aspirationally".

No, it's will be using standard cell phones with no additional hardware on the cell phone.

Seems like it requires Starship to launch unless I heard it wrong? Its too big for Falcon 9. Starship is still not space ready so it seems like it really depends on that program moving forward before we see these satellites in space. Given Musk's track record, I expect probably a year extra delay before it finally is opened to users. He did mention that they will probably produce a mini version of the satellite if Starship is delayed too much. Would they need to develop this mini satellite?

I would bet money there will be some availability of the service by 2023, when they said it'd be available.

I know the meme is that Musk's track record is to delay, but Starlink has been regularly underestimated and has kept pace pretty successfully. Much more than industry insiders expected.

At the event, Musk said if Starship isn't ready they'll try to launch a slightly smaller version of the Gen2 sats on Falcon 9.

This isn't just SpaceX, this is T-Mobile as well.

Why do people keep comparing to Theranos when Starlink is already being used by hundreds of thousands of people? Starlink exists. In fact, SpaceX/Starlink now operates more satellites in Low Earth Orbit than every other country and business in the world, combined. Theranos had no proven track record.

Like, what's enough to satisfy you?

Minor pedantic observation

Starlink will be able to claim that achievement of having more than everyone else combined soon but isn’t there yet.

There are around 4000ish non starlink satellites in orbit today , starlink has 3k today .

I don't know where you're getting "4000ish non starlink satellites". This seems to be the current number at the start of the year inclusive of starlink. It's growing fast!

Many sources on the web seem to quote the Union of Concerned Scientists and here is their database: https://www.ucsusa.org/resources/satellite-database

It's currently at 2219 starlinks and 5465 total, but Starlink has launched more since then.

I explicitly said low earth orbit (below 2000 km), not all satellites in orbit.

This is an egregious comparison, not even remotely close.

Touche