Modern AI both shortens the useful lifespan of software and increases the importance of development speed. Waiting around doesn’t seem optimal right now.
By whom? He seems highly credible to me, and his credentials check out, especially compared to hype men like Sam Altman. All youre doing is spreading FUD by an unnamed "they"
He only criticizes ai capabilities, without creating anything himself. Credentials are effectively meaningless. With every new release, he clamors for attention to prove how right he was—and always will be. That’s precisely why he lacks credibility.
He started and then sold a machine learning startup to Uber. He's also written multiple books about the construction of the human mind and he has a PhD from MIT. I would hardly call that creating nothing. He's not clamoring for attention, he's asking that AI be regulated and pointing out a lot of the glaring issues with the field.
This is literally just the scaling laws, "Scaling laws predict the loss of a target machine learning model by extrapolating from easier-to-train models with fewer parameters or smaller training sets. This provides an efficient way for practitioners and researchers alike to compare pretraining decisions involving optimizers, datasets, and model architectures"
"Raising visibility on this note we added to address ARC "tuned" confusion:
> OpenAI shared they trained the o3 we tested on 75% of the Public Training set.
This is the explicit purpose of the training set. It is designed to expose a system to the core knowledge priors needed to beat the much harder eval set.
The idea is each training task shows you an isolated single prior. And the eval set requires you to recombine and abstract from those priors on the fly. Broadly, the eval tasks require utilizing 3-5 priors.
Putting a starship in low earth orbit right now would be a bit reckless, because if the engines fail to relight it's going to come down at some point in a completely random spot along the trajectory, and it's quite a large piece of debris that is designed to not burn up. By contrast this test (which involved the first 0g relight of a raptor) was designed so that if that failed the ship would still come down in a designated keep out zone in the ocean.
Even ignoring the safety risk, the value SpaceX gets from this flight is largely testing the re-entry (heatshield, flaps, etc) of starship. Putting it in an orbital trajectory risks a failed engine relight making it impossible to test that, because the ship will be dead (out of power) by the time it comes back down. Whatever low-cost payload you could put up there for "free" (the cost of risking the payload being destroyed if the test flight goes wrong too early) might not be sufficient to pay for the risk stopping in an orbital trajectory imposes on the test objectives.
Now that they've successfully lit a raptor in 0g, I imagine they're a fair bit more likely to make the subsequent flights orbital.
An FTS does not prevent 100t of debris of which a significant amount is designed to survive re-entry from impacting the surface. If you blow up 100t you still have 100t of debris, just in lots of bits (and honestly, still big chunks - an FTS does not atomise or even close - it punches a hole and then the vehicle collapses structurally)
All it really does is remove the explosive potential of the fuel.
So no, it's not designed to do this.
Once in orbit the FTS system is usually deactivated (safed).
It's designed to survive re-entry in a very particular, and repeatedly adjusted, orientation. It's unclear how much would survive re-entry in an uncontrolled tumble. But better safe than sorry.
Debris in orbits that nearly intersect the atmosphere are practically harmless. The issues arise when you have an orbit that will intersect other orbits a billion times before decay.
While the reality is very complicated, you can make a rough handwave model for orbital debris in a circular orbit: At 100km it lasts 1 orbit (90 minutes) at most, and every ~100km you add after that increases orbital lifespan by ~10x.
> Putting a starship in low earth orbit right now would be a bit reckless, because if the engines fail to relight it's going to come down at some point in a completely random spot along the trajectory, and it's quite a large piece of debris that is designed to not burn up.
More relevantly—it's harder to make a marketing stunt out of a fallible mission.
And don't say launch services, 'cause you're using "it's a marketing stunt" to explain why they aren't taking payloads. "It's R&D" makes a heck of a lot more sense.
> More relevantly—it's harder to make a marketing stunt out of a fallible mission.
SpaceX is well known for making a marketing stunt out of their failures (see "How Not to Land an Orbital Rocket Booster"). In fact, Elon Musk is known to make marketing stunts out of anything.
But that's not it, they could have launched a dummy payload if they wanted to, they have already done it for other rockets, and yeah, they made it a marketing stunt. Famously, they launched Elon Musk's Tesla Roadster in the direction of Mars. But here, launching a payload at this stage of development is just not the best thing to do.
The utility of test flights doesn't come from delivering a payload; but from collecting data. For example today's test involved intentionally weakening starship's heat shield to see if previous estimates of required shielding had been too conservative.
You can't deliver a payload to orbit if you're not planning on going to orbit. They have a launch licence for a particular mission plan.
The license for Falcon 9 and Falcon Heavy specifically says “Authorization: SpaceX is authorized to conduct flights of
launch vehicles: […] (c) Transporting Dragon 2 to low Earth orbit or a payload
to orbit;”
The license for Starship/Super Heavy Launch Vehicle does not include such an authorization.
"The payload is data." Elon
The data they get from the test and they focus on getting the most and most valuable data.
This is an explicit choice.
A physical payload at this stage would reduce the overall value that they are able to get from a flight.
It's exactly because they hadn't demonstrated engine relight in space yet. That means they couldn't guarantee they would be able to deorbit Starship in a controlled way. An uncontrolled deorbit would be bad because it could come down literally anywhere and large chunks would hit the ground.
Because they couldn't guarantee precise deorbiting, they never put it in orbit to begin with. Starship was on a ballistic trajectory that falls back to Earth. Any payloads they deployed would fall back too, unless they carried their own rockets to reach orbit themselves.
> Starship was on a ballistic trajectory that falls back to Earth
Actually Starship on this flight wasn't quite on a ballistic trajectory, the periapsis was actually above the ground so it counts as an orbital trajectory. Without the atmosphere it would keep going.
Lots of technical speculation in the sibling comments. I suspect the real answer is that any real payload would require at least FCC approval, and possibly modification of their FAA flight plan. The income from the payload would not be worth the delay.
From what I followed, its because:
They already send payload to space with the falcon rockets (reliable cash cow) and
the only point of starship system is proof of concept of full reusability. There is no point to the whole project if they dont achieve this.
They are constrained in launches, either through regulatory (FCC) or engineering timeline
and so far, every launch counted. They achieved a substantial improvement.
If they dont catch the starship (upper stage), the whole project is pointless so its important keep things simple (details like payload door, cargo dont matter) until they achieved the whole loop (land & catch starship)
They already have a cheap way of launching. Cheap for current market rate. Whole point of this project is to improve cost by / 1000.
They achieved substantial progress with every flight.
1. It flies, 2. hot staging 3. reach orbit 4. land both & heat shield, 5. catch booster, 6. extreme conditions & improve heat shield.
No. This and all previous flights have intentionlly been barely sub-orbital, with less than one orbit. Launch in Texas, re-entry over Indian Ocean. A full orbit at that altitude takes about 90 minutes; this was over in a little over an hour.
The reason was safety. If it was orbital, then controlling the re-entry would require a sucessful relight of the engines. If that failed then the re-entry point would depend upon the vagaries of orbital decay from residual atmospheric drag. That's no doubt why today's relight was so brief; they didn't want to significantly alter the reentry point.
I don't believe this is quite correct. The last few trips are actually orbital, just not of the correct elliptical shape to do more than a half orbit as the perigee is less than the radius of earth. If earth was a point mass, it would have orbited.
This means you don't have to do anything to deorbit while proving you could have made a full orbit if you wanted to.
If the earth suddenly became a point mass except for all the humans, we'd all already be in orbit. At the "top" of a highly elliptical orbit that passes relatively close to that point mass, yes, but an orbit. Everything that was within the Earth's sphere of influence, not moving so fast to be on an escape trajectory, and not with 0 horizontal velocity relative to the point mass would be in an orbit.
Orbital means "on a trajectory that doesn't intersect (or escape from) the body you are orbiting", otherwise the word is meaningless.
I think it would also have to be at an equinox. Otherwise the pole would be tilted towards or away from the sun, meaning that it orbits at a slightly different velocity, so you would have some velocity relative to the center of mass.
Every time you throw a rock, it ends up on a trajectory that "would have orbited if the earth was a point mass". That's just not a very useful definition of "being in orbit".
Why not? If your starting point is completely static compared to the point mass and your aspect area isn’t zero, you’re going to fall directly down towards the point mass and are going to hit it.
If it was a point mass, and you had exactly zero horizontal motion relative to it, you'd go right through and out the other side.
Well, except for relativity turning it into a black hole with a Schwarzschild radius of 8.87 mm so it won't be "point-like".
But most of the disintegrate sheen of plasma that used to be your body would have had some horizontal motion compared to it, even if only due to you starting off as an extended body.
Orbit is not about distance, it is about speed (perpendicular to gravity). Starship deliberately did not reach orbital velocity on this flight, so although they are fairly certain it is capable of doing so (esp after testing engine relight on this flight), they have not done it yet. Not being at orbital velocity is one reason why they are not delivering payloads.
The reason for choosing not to go orbital at this point is indeed as another commenter said: safety. A craft which has not reached orbital velocity is therefore traveling on a parabola, which has a predictable flight path. If they lose control of all systems, they can still be confident that it will crash where they want it to (in this case, the Indian Ocean). If instead you boost enough to get orbital and then lose control, it is much harder (almost impossible) to know where the craft will end up.
SpaceX (and the FAA) wants to make sure everything works properly by doing sub-orbital flights, then when they are fairly confident nothing will completely break on them they can start testing in an orbital regime, by powering on Starship's engines long enough to achieve the requisite speed.
They don't need to relight to reach orbit, they just performed SECO a bit early. For reference at 190 km up which Starship reached, it needs to go 28,000 km/h to stay in orbit. Today it was flying at 26,000 km/h.
Scaling The Turk to OpenAI scale would be as impressive as agi
"The Turk was not a real machine, but a mechanical illusion. There was a person inside the machine working the controls. With a skilled chess player hidden inside the box, the Turk won most of the games. It played and won games against many people including Napoleon Bonaparte and Benjamin Franklin"
reply