Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Search results

  1. Ben W

    FSD V12 (end to end AI)

    The common low-hanging-fruit problems we're seeing, like weird stopping behavior at typical stop signs, and driving speeds on typical streets, I fully expect to be solved this year. And agreed that solving these issues alone may be enough to make it a useful L2 system for straight-line...
  2. Ben W

    FSD V12 (end to end AI)

    Surprisingly, this is not always the case. An example: AlphaGo developed world-class capabilities from training on an enormous library of professional human games, but it was eventually leapfrogged by AlphaZero, which started completely from scratch. In other words, the pre-existing structure is...
  3. Ben W

    FSD V12 (end to end AI)

    This is almost certainly an emergent property of the neural net training. They'll need to retrain the model with fewer examples of being close to the curb (and more examples being further away) to properly address this. Unclear whether there is supervisory C++ logic that explicitly monitors curb...
  4. Ben W

    FSD V12 (end to end AI)

    These parameters are only modified by retraining the neural net. (There are two semantic definitions of "parameters": the one described here is neural net weights. The other sense is user-facing parameters outside the model, such as assertiveness or auto-speed or "minimal lane changes".) So I'm...
  5. Ben W

    FSD V12 (end to end AI)

    Yes, I believe each dot-release is fully retrained. The amount of C++ code and explicit control logic is rapidly diminishing; I'm not sure what sort of controls and parameters are still explicitly coded, other than the user-facing settings.
  6. Ben W

    FSD V12 (end to end AI)

    Until FSD reaches true L4 reliability (less than one safety mistake per million miles), the shpilkes (love the word) will regrettably be necessary. There is going to be a vast uncanny valley where the reliability is enough that you REALLY want to just relax and stop paying attention, but not...
  7. Ben W

    FSD V12 (end to end AI)

    In general, by gathering a whole bunch of video examples of the "correct" behavior, and using that in the next training set for the neural net (which tries to mimic the training set), and rolling it out in the next version. It is sort of a black art, though.
  8. Ben W

    FSD V12 (end to end AI)

    No doubt they can improve specific common cases by explicitly bolstering the training set for those cases. The question is whether that approach might inadvertently underweight something else important, or also whether it will help at all the with the long tail of edge cases that don't fit...
  9. Ben W

    FSD V12 (end to end AI)

    Good idea, worth a try. The car is a 2017 Model 3, but I would have expected that miscalibrated cameras would have affected v11 as well? V12 is now downloading on my 2022 Model Y as well, we'll see if it's significantly different there too.
  10. Ben W

    FSD V12 (end to end AI)

    I'm taking that into account. For L4 reliability, the car will need to be able to drive at least a million miles on average between necessary interventions (where the car makes a clear mistake that any skilled human driver would avoid). Right now it seems to require such an intervention about...
  11. Ben W

    FSD V12 (end to end AI)

    My first drive experience with 12.3 was similar. Even on the 'assertive' setting, it wanted to drive painfully slowly on unmarked two-lane roads, like less than half the posted speed. (Much slower than any reasonable human.) It double-stopped at most stop signs; first stopping far short of the...
  12. Ben W

    I don't expect any automated driving system to actually reach full level 5.

    BACC. (Burger-Aware Cruise Control.) With the ability to place the order, and pop the frunk so the order can be placed in it.
  13. Ben W

    Wiki Super Heavy/Starship - General Development Discussion

    If it's used for raising a depot to high orbit, the "cargo" would be 1200 tons of methalox. (Or 3000 tons of liquid xenon? One can dream!) And if the power source is continuous, it may make sense to just keep accelerating even at non-optimal times, depending whether ion fuel consumption is the...
  14. Ben W

    Wiki Super Heavy/Starship - General Development Discussion

    True, supplying non-LEO depots really does call for a Spaceship-only design, perhaps with ion thrusters for propulsion instead of methalox. (But with methalox as the payload.) How long would it take for an ion-propulsion kick stage to boost a full methalox depot from LEO to GEO? Napkin math...
  15. Ben W

    Wiki Super Heavy/Starship - General Development Discussion

    Interestingly, LEO to GEO requires about 3.9 km/s of delta-V, and GEO to the moon's surface also requires about 3.9 km/s of delta-V. So if there were refueling depots/stops in both LEO and GEO, this could allow a single Starship to get 400 tons to the moon. Or having an intermediate depot in an...
  16. Ben W

    Wiki Super Heavy/Starship - General Development Discussion

    Having a T:W ratio above 1 is only critical for liftoff, when ascending vertically. For a second stage, most of the acceleration is already horizontal and gravity losses are small, so there are diminishing returns to adding thrust. The Falcon 9 second stage (with a single Merlin) has only a ~1:2...
  17. Ben W

    SpaceX Starship - IFT-3 - Starbase TX - Pre-Launch Preparations Thread

    At a mutual spin rate of 0.3rpm (about one revolution every three minutes), and transferring the propellant very slowly (say 10kg per second), it should be very manageable. Transferring 100T of propellant at that rate would take about 3 hours, so the change in center of mass would be very...
  18. Ben W

    SpaceX Starship - IFT-3 - Starbase TX - Pre-Launch Preparations Thread

    My expectation is that they could make the two ships slowly spin around each other, and use the small centrifugal force to settle the propellants laterally. This could be done with thrusters, or with flywheels/gyroscopes; unclear if the mass of the required flywheels would outweigh (literally)...
  19. Ben W

    Mars and Off Planet Colonization - General Possibilities Discussion

    At the pace at which Optimus (and AI in general) is progressing, I fully expect that the ISRU plant could be built robotically, before humans are sent. Especially with NASA involved on the safety side, and inevitable delays in getting an actual crewed mission sent, this seems like the most...
  20. Ben W

    Mars and Off Planet Colonization - General Possibilities Discussion

    "Detach the Saucer Section!" I don't think it's realistic to split Starship in half like this, since (among other things) it would involve splitting the heatshield in half, as well as the downcomers from the header tanks. Sending crewed Starships in pairs has the advantage of avoiding this...
  21. Ben W

    Mars and Off Planet Colonization - General Possibilities Discussion

    Dyneema / Zylon is typically sold as a 12-strand braided rope, so presumably it could be un-braided. Or the full 6cm tether could be made by braiding twelve 2cm themselves-braided ropes together. (Or two Starships could be connected via multiple smaller tethers, rather than one large tether.) I...
  22. Ben W

    Mars and Off Planet Colonization - General Possibilities Discussion

    The required tension would be far less than this. Starship's dry mass is ~100T, plus ~100T payload, plus (wild guess) ~50T fuel after trans-Mars injection. So even at a full 1g acceleration, the tether would only need to support ~250 tons tension. (It wouldn't matter if Starship is tethered to a...
  23. Ben W

    SpaceX F9 - Starlink Group 7-9 - SLC-4E

    6:31pm PST launch. With local sunset at 4:59pm, that may be just slightly too late for a spectacular sky show, but I'll be watching from my backyard anyway!
  24. Ben W

    Boring Company Hyperloop

    The primary reason for Hyperloop's infeasibility (per Hyperloop One) seems to have been the cost of tunneling. It wouldn't completely surprise me if The Boring Co. ends up acquiring Hyperloop One's assets for pennies on the dollar, and keeping the project on the back burner, in case the...
  25. Ben W

    SpaceX Starship - Integrated Flight Test #2 - Starbase TX - Including Post Launch Dissection

    All true. There are lots of factors at play here. Booster needs to keep its acceleration high enough to keep the propellant settled, which requires less thrust if Starship is also at reduced throttle during separation. Conveniently (and counterintuitively), halving Starship's acceleration...
  26. Ben W

    SpaceX Starship - IFT-3 - Starbase TX - Pre-Launch Preparations Thread

    Earth is almost net zero in terms of its total thermal effect relative to it not being there. Its radiation warms up the spacecraft of course, but that's balanced by the spacecraft being in Earth's shadow half the time, which blocks the much hotter sun. It does change the geometric distribution...
  27. Ben W

    SpaceX Starship - IFT-3 - Starbase TX - Pre-Launch Preparations Thread

    A different way to think about it: at thermal equilibrium, the "average temperature" as seen in all directions from the Depot's point of view must be the same temperature as the fuel. (About -180 °C.) Ironically, the boiling point of LOX (-183 °C) is slightly lower than the freezing point of CH4...
  28. Ben W

    SpaceX Starship - Integrated Flight Test #2 - Starbase TX - Including Post Launch Dissection

    Just pointing out that they could. To the extent it allows them to make the hot staging ring less massive, it may turn out to be a performance win. But mostly just observing that there are multiple ways for SpaceX to mitigate the hot staging failure without requiring a hardware redesign. As soon...
  29. Ben W

    SpaceX Starship - Integrated Flight Test #2 - Starbase TX - Including Post Launch Dissection

    Good analysis. Another way to look at it: SpaceX's simulations for the hot staging must have had Starship initially separating from Booster at some particular rate, say 0.6 g. "All" that's needed (in quotes because it's an enormously difficult problem) is to adjust the thrust profiles to more...
  30. Ben W

    SpaceX Starship - IFT-3 - Starbase TX - Pre-Launch Preparations Thread

    If you think reflective Starships might upset astronomers, just wait til you add several acres of solar panels to each one! Nuclear may be better suited as a power source for that reason, though that may make it a greater environmental hazard than just launching LOX+LH2 in the first place. And...
  31. Ben W

    SpaceX Starship - IFT-3 - Starbase TX - Pre-Launch Preparations Thread

    At first, yes. But if Elon's dreams come true of sending 1000 Starships to Mars every couple years, it will be far more than that. Of course, by that time we'll be able to place enormous telescopes in orbit equally well, so maybe the astronomers will be ok with it after all.
  32. Ben W

    SpaceX Starship - IFT-3 - Starbase TX - Pre-Launch Preparations Thread

    Indefinitely, if you can deflect enough incoming radiation (from sun and Earth; the Moon's contribution is measurable but probably insignificant). A reflective Mylar sunshade or equivalent would do the trick, although it might also severely irritate astronomers. A potential approach is to keep...
  33. Ben W

    SpaceX Starship - Integrated Flight Test #2 - Starbase TX - Including Post Launch Dissection

    Starship's exhaust certainly reduces Booster's forward acceleration. The question is whether it's strong enough to cancel it out entirely. Booster's three center Raptors combined can produce anywhere between about 270MT and 675MT of thrust, depending on throttle. Starship's six Raptors can...
  34. Ben W

    Deorbit ISS. A Job for Starship?

    I certainly hope it's uncrewed when it deorbits!
  35. Ben W

    SpaceX Starship - Integrated Flight Test #2 - Starbase TX - Including Post Launch Dissection

    Deceleration is fine, as long as it doesn't decelerate below zero. ("Deceleration" is not technically the correct term; more "reduced acceleration".) To find the balance point, all that's needed is to total up the forces acting on the booster: the forward thrust provided by Booster's center 3...
  36. Ben W

    Wiki MASTER THREAD: Actual FSD Beta downloads and experiences

    First drive on 11.3.3 in the books, in my MY2022. 10-mile drive in West LA. Three disengagements. First was at the end of a freeway onramp, when the car insisted on staying exactly parallel with another car as the two lanes narrowed and merged. I had to manually brake and pull in behind. Second...
  37. Ben W

    LEO Space Station with Artificial Gravity (w/Discussion of effects on the human body)

    Revisiting this thread again, with some high-level observations and notes. Also, some details regarding my estimates for the tether mass for connecting two Starships. Starship has a dry mass of about 100t, and a payload capacity of 100t. I'm guessing that it would contain about 50t of fuel...
  38. Ben W

    LEO Space Station with Artificial Gravity (w/Discussion of effects on the human body)

    Thanks for the chat as well, I've enjoyed it. Will just comment a final time on each point and where we differ; we may have to agree to disagree. My contention is that the word should not be "slightly", but "drastically". Another not-to-be-overlooked upside would be the tether's ease of...
  39. Ben W

    LEO Space Station with Artificial Gravity (w/Discussion of effects on the human body)

    Right. What I was getting at was that in discussing the Starship tether you seemed hugely concerned about all of the stability and control issues, ostensibly because it wouldn't be connected to any stationary object, when all those same concerns apply just as strongly to aircraft. Yet you didn't...
  40. Ben W

    LEO Space Station with Artificial Gravity (w/Discussion of effects on the human body)

    Not to put too fine a point on it, but you're saying that the _only_ concern of engineers regarding powered flight was about finding a powerful enough engine to get off the ground, and _nothing at all_ about stability or control or vibration or wobbling issues while in the air not rigidly...
  41. Ben W

    LEO Space Station with Artificial Gravity (w/Discussion of effects on the human body)

    Let's back up a bit. Rigid points of external contact are only required for damping out absolute "DC" inertia; linear or rotational. For the "AC" components of inertia (e.g. vibration, wobble), a rigid point of external contact is not needed, only a damper that converts the vibrational motion to...
  42. Ben W

    LEO Space Station with Artificial Gravity (w/Discussion of effects on the human body)

    But then again, computers are exceptionally good at "extremely difficult" control problems like this. (Such as launching and landing Falcon 9 boosters.) If all the required degrees of freedom are present, with redundancy, then it should be doable. "Springy" was a poor choice of word. The...
  43. Ben W

    LEO Space Station with Artificial Gravity (w/Discussion of effects on the human body)

    The appeal of the tether is that the load-bearing attachment points already exist in the Starship design, in the form of the catch arms. A "suspension" to make the tether very slightly springy would certainly be necessary; it wouldn't just be a "dumb" rope. An alternative design would be to have...
  44. Ben W

    LEO Space Station with Artificial Gravity (w/Discussion of effects on the human body)

    Presumably the damping structure could also be attached to Starship, or at least to the catch arms, and not just floating loose on the tether. But even floating loose, it would have some damping effect, absorbing some of the vibrational energy via friction and converting it to heat. The amount...
  45. Ben W

    LEO Space Station with Artificial Gravity (w/Discussion of effects on the human body)

    Piano strings are also under constant tension, yet each string has its own damper, which very effectively stops vibration when applied. This would work in space, too. (The mechanism of action of the damper is frictional, and is not dependent on gravity or atmosphere.) I think this...
  46. Ben W

    LEO Space Station with Artificial Gravity (w/Discussion of effects on the human body)

    The tether would have a "suspension" or other means to actively damp this vibration. The fundamental frequency of this vibration would be fairly high (the speed of sound in Zylon is about 9000m/s), so the fundamental vibration frequency for a 300m tether would be around 15Hz. It's not clear...
  47. Ben W

    LEO Space Station with Artificial Gravity (w/Discussion of effects on the human body)

    Well, landing a rocket [on a droneship] also requires active synchronization between both systems with no redundancy, so I'd call that a similarity, not a difference. (In this context it's an inconsequential detail that the booster and droneship don't synchronize directly with each other, but...
  48. Ben W

    LEO Space Station with Artificial Gravity (w/Discussion of effects on the human body)

    Why should the tether itself need to control the system? At each side of the tether will be a fully functional Starship with independent (and mutually synchronizable) active 6DOF maneuverability, so it's perfectly fine for the tether itself to be completely passive. You don't need an active kite...
  49. Ben W

    LEO Space Station with Artificial Gravity (w/Discussion of effects on the human body)

    Agreed that mass is the primary difficulty with making a rigid connection. Even a tether is not insignificant mass; a 300m Zylon tether strong enough to support two mostly-empty Starships at 1g would mass a few metric tons. A useful rigid connection would likely add another order of magnitude...
  50. Ben W

    LEO Space Station with Artificial Gravity (w/Discussion of effects on the human body)

    I certainly didn't mean to imply that it would be easy (because nothing in spaceflight or rocketry is); just that it should be solvable. One advantage of testing a space-based artificial-gravity system, unlike a rocket launch, is that it can be eased into arbitrarily gradually, starting at (say)...