By on April 22, 2019

Elon Musk

Tesla celebrated “Autonomy Day” on Monday, broadcasting a web presentation to investors touting its self-driving strategy — part of CEO Elon Musk’s attempt to prove that the manufacturer’s longstanding commitment to the technology will bear fruit. While the mood was undercut by news of one of its vehicles spontaneously combusting in a Shanghai parking garage, the company stayed the course, focusing on autonomy instead of electrification.

Still, it wasn’t the fire that had investors seeking comfort. Tesla’s promise of fully self-driving vehicles is years old, with Musk initially pegging 2018 as the target date. That didn’t happen, with 2018 turning out to be the year the company faced elevated scrutiny over the effectiveness of its Autopilot system after a series of high-profile crashes. 

The broadcast began about 30 minutes later than anticipated but, once up, Tesla hit the ground running. Musk claimed (once again) that LiDAR is not the way to go for totally autonomous driving, calling it a “fool’s errand.”

He instead suggested that the key to success involves a proprietary neural processor that the company has tasked Samsung to build.

“It seems improbable. How could it be that Tesla — who has never designed a chip before — would design the best chip in the world? But that is objectively what has occurred,” Musk said. “All Teslas being produced right now have this chip … All Tesla cars right now have everything necessary for FSD. All you have to do is update the software.”

FSD is the computer that’s said to enable full-self driving via future over-the-air software updates, thanks to help from from a small and robust CPU. But that’s only half the battle.

Tesla showcased the problem of identifying objects by showcasing a lizard, boat, pair of scissors, and a cello — noting that, while it’s easy for us to differentiate between them, a computer just sees a mass of pixels with different brightness values. However, the myriad of angles, colors and shapes practically every object can assume makes the process infinitely more complicated.

The solution? Artificial neural networks. While much simpler than what exists inside your own skull, Tesla says it can feed images into a vast network and effectively let its system learn what’s what over a period of time. The entire presentation wobbled between being overly technical for average folks and insultingly simplistic. But the gist is that the company sees another hardware upgrade as the path to self-driving glory.

Musk claimed that the new chip would be about three times better than the current one and take another two years to get here. He also noted that FSD has built-in redundancies that should make it bulletproof, saying the system was less likely to fail than a human brain. But it will need to be updated remotely and fed a lot of data if it’s to function properly, meaning Tesla will continue to snatch fresh data from its vehicles for years to come — as simulated data has been deemed insufficient by the company.

“Using simulators, it’s like grading your own homework,” Elon joked.

While this paints a clearer picture of how Tesla plans to tackle self-driving in the future, it doesn’t sound terribly different from its current strategy. The company has amassed driving data from its customers for years; all this amounts to is even more robust hardware — which it has released before and promised would result in complete vehicular autonomy. That’s not to say Tesla isn’t on the cutting edge of neural network processing or that it won’t work, just that we’ve heard this song before.

Investors are expected to take a ride in vehicles equipped with the company’s latest self-driving system later today, so we’ll see how it’s all coming together. Naturally, it’s doubtful any of them will have anything overtly negative to say. At worst, the company’s share price maintains its current trajectory. Tesla’s still making headway in the world of autonomy, even if that progress is not quite as swift as Waymo’s.

[Image: Tesla]

Get the latest TTAC e-Newsletter!

28 Comments on “Promise Hour: Tesla Broadcasts Self-driving Tech to Perk Up Investors...”

  • avatar
    SCE to AUX

    No way, thank you very much.

    And what lawyer at Tesla said this was a good idea? I’d like to know if any have quietly jumped ship over this.

    Level 2 autonomy (like Autopilot) doesn’t even have to work in order to comply with its definition. But Level 4 or 5 is no joke.

    It won’t be burning cars, poor build quality, or missed production targets that ruin Tesla; it’ll be this.

  • avatar

    This is all such BS.

    Just STFU and make good cars that people enjoy driving.

  • avatar

    The comments he made about LIDAR and HD maps make sense.

    If we can drive with two eyeballs, theoretically there is no reason the car can’t drive with eight cameras.

    • 0 avatar

      The cutting edge of neural-net based image and video processing is still in its infancy (check out the bug-eyed nightmare fuel from Google’s programs) and the best systems need massive amounts of computing power to achieve high accuracy. There are people working to put that computing power into vehicle-sized packages, but there’s a reason they’re using a whole huge host of sensors. Using radar and lidar in conjunction with your cameras gives you a far higher chance of correctly identifying stuff, and helps ensure you’re not getting spurious results from your image processing.

      Humans get short shift a lot, especially when it comes to discussions of autonomous vehicles. Our vision processing is incredible when we’re paying attention.

      • 0 avatar

        I’m not a fan of LIDAR at all. I’m using cameras + stereoscopic FLIR. LIDAR adds nothing.

        • 0 avatar

          “I’m not a fan of LIDAR at all.”

          for a particular reason?

          I mean, other than “Elon says it’s not needed?”

          • 0 avatar
            SCE to AUX

            @JimZ: mcs does this stuff for a living.

          • 0 avatar
            Matt Posky

            During the post-presentation Q&A, Elon said there weren’t practical applications for lidar if Vision works as intended. But he did remind investors that Tesla uses radar, which he said should be sufficient for road travel, and admitted spaceX uses lidar for ranging objects in space.

    • 0 avatar

      It isn’t the image input that fails. It’s the image processing. Musk’s FSD chip is no HAL 9000 (which went nuts under stress). Accidents with autonomous test vehicles (and not just Teslas) show that the technology can’t be trusted above level 1.

    • 0 avatar

      “If we can drive with two eyeballs, theoretically there is no reason the car can’t drive with eight cameras.”

      LOL, what a load of crap.

      your eyeballs (and those cameras) don’t actually DO anything other than sense light and send information off to something for processing. Your eyes send what they sense to your brain, an incredibly complex organ we don’t completely understand. a Tesla’s cameras send what they sense to a really fast, glorified calculator.

      I mean, I have two legs. theoretically there’s no reason I can’t run as fast as Usain Bolt.

  • avatar

    Someone I know woke up to loud explosions at around 3am a few days ago. Their Chrysler Pacifica, thankfully parked outside, exploded and burned to the ground. No viral video of this one because it is not a Tesla.

    Cars burn all the time. We just don’t hear about it because only Tesla fires are worth money (to journalists and short sellers.)

    • 0 avatar


      “Someone I know” isn’t the most compelling argument. While you’re busy calling this clickbait for journalists and short sellers, the flip side is perhaps this was a dog and pony show prior to the earnings call. He deflected any questions that weren’t about FSD.

      He also said they would have to update 250k cars. And if the “Tesla Network” will be powered by low orbiting satellites, well he better start getting them up there. According to him, it’s only 1 year away. He was even vague about insurance, “probably Tesla” was the answer.

      Oh, and the stock isn’t moving after hours.

    • 0 avatar

      “Someone I know woke up to loud explosions at around 3am a few days ago. Their Chrysler Pacifica, thankfully parked outside, exploded and burned to the ground.”

      yeah, sure it did.

  • avatar

    I didn’t realize Samsung was fabricating complex processors? I can’t seem to find anything they seem to make processors for that is more powerful than a mobile phone, and it looks like the Galaxy phones are using boards partially designed by suppliers. I thought AMD and Intel were the principle players in processors, though I know nothing about whatever “neural processors” are supposed to be. I’m mostly a consumer level electronics person though, so what do I know.

    • 0 avatar

      The processor landscape has changed – the single giant CPU era for computing could be coming to a close. Examples:

      There’s some amazing processing power in modern GPU – the processors responsible for 3-D video graphics. They’ve been re-purposed (and now designed for) for really complex things like neural networks.

      Samsung could be making large chips with tens of smaller processors than the typical AMD or Intel processor – a chip filled with type of processor in their cellphone. That combination is surprisingly powerful now.

      Another category are FPGA chips – chips that you can program their inner circuitry like code. Once locked into a particular configuration designed around the task they need to do – they can be quite efficient and speedy.

  • avatar

    Preston Tucker on steroids…

  • avatar

    “All Tesla cars right now have everything necessary for FSD.”

    That’s what he said about the two previous hardware versions.

  • avatar

    Self-driving, self-immolating… what’s the difference?

  • avatar

    I just hope this new miracle chip is fireproof.

  • avatar

    Tesla Q1 deliveries. 63,000 cars.
    31% lower than Q4.
    Mix shifted to Model 3 from high priced S and X. (those models declines 56%).
    This is real BAD.

    • 0 avatar
      SCE to AUX

      Tesla has consistently stated that its future depends on the Model 3 (and eventually the Y), and that the S and X were merely stepping stones. The volume mix shifted to the Model 3 last year.

      They need margin and volume, and it has to be on the Model 3.

      Frankly, I pay no attention to the latest EV supercar concepts from other mfrs, because their volumes will never matter. Those are easy to build, but it’s much, much more challenging to produce something affordable for a profit.

  • avatar

    Meanwhile, despite all the doom and gloom, Tesla continues to accomplish amazing things:

    17,500 sales of a US-made car in a single month in Europe is unheard of. Best selling car in Norway, Netherlands, Switzerland for the month of March. Yes, these are built up demand deliveries, it will level off, but it’s a $70k electric car, not a $25k Golf.

    When was the last time a US made car was a best seller anywhere outside of North America, let alone an expensive electric car?

    This latest automated driving stuff is typical Musk–he’ll promise the moon in 2 years, and in 2 years it won’t happen and people will call BS on him. But in 5 years we’ll start to see it regularly, there will be some crashes that people will go nuts over but statistically it will be safe. In 10 years they’ll be the leaders and the other manufacturers will be trying to play catch-up.

  • avatar

    Elon Musk are really taking people for a ride here, including the Tesla fanbois at TTAC. The reality, though, is that Elon Musk is incompetent, because even after all this time, he hasn’t even managed to do something as basic as ensuring that it doesn’t take longer to charge a Tesla than it takes to fill a fuel tank! That is simply outrageous. When Elon Musk isn’t even capable of fixing this simple defect, there’s no reason to think he’s got what it takes to develop self-driving cars.

  • avatar

    “The fundamental message that consumers should be taking today is that it’s financially insane to buy anything other than a Tesla, ” the CEO of the electric auto maker said after Monday’s robotaxi announcement. “It would be like owning a horse in three years. I mean, fine if you want to own a horse. But you should go into it with that expectation.”

    So all you insane dawn and dusk nosebag oat-fillers out there, remember, Dobbin’s days are numbered! No more kegs of Bud Lite! Teslas will run the world and we’ll have a million robotaxis by next year, Uber is toast, all it takes is an over-the-air update and the Muskies will come alive! Alive I tell you! Ah, ha ha ha.”

  • avatar

    There is always room for a new wave of investors….

    I think what used to be called Tesla investors are now referred to as Millionaires.

Read all comments

Back to TopLeave a Reply

You must be logged in to post a comment.

Recent Comments

  • ToolGuy: Pssssst – hey, kid – you there. Stay on the pavement — your life will be better. Now go.
  • ToolGuy: Not to rub it in, but let’s review: • ToolGuy advised you not to buy a Honda • ToolGuy advised you of...
  • ToolGuy: As long as you have two fuel tank straps (and you should), the chances of both of them giving way due to...
  • ToolGuy: “It’s like talking about the space under your passenger seat.” The correct approach to the right...
  • thehyundaigarage: Is that the best they could do for panel alignment on a press vehicle? That hood and fender, damn!

New Car Research

Get a Free Dealer Quote

Who We Are

  • Adam Tonge
  • Bozi Tatarevic
  • Corey Lewis
  • Jo Borras
  • Mark Baruth
  • Ronnie Schreiber