Tech
Roadrunner and Wile E Coyote
The never-ending chase (BSR Agency/Getty Images)
MEEP-MEEP

YouTuber runs a Tesla through a fake wall to poke a hole in its camera-only, no-lidar strategy

The Tesla ran straight into a wall painted to look like its surroundings. A non-Tesla with lidar stopped easily.

Jon Keegan

In the classic “Looney Tunes” cartoons, the Road Runner is constantly evading capture by the tireless Wile E. Coyote, who sets elaborate (and fruitless) traps to snare the clever bird. One of the most famous tricks Wile E. conjured was to paint a fake tunnel on a rock wall, hoping the Road Runner’s eyes would be fooled and he would smash into the wall.

It turns out this trick appears to work on Teslas.

Former NASA engineer turned YouTuber Mark Rober published a new video over the weekend that tests a camera-only Tesla against a lidar-equipped vehicle to see how the Tesla’s “autopilot” braking system responds to the optical illusion of a wall thats camouflaged to look like a seamless view of the road ahead.

About 10 years ago, as CEO Elon Musk was seeking to cut down on the costs of Teslas, he make a radical decision to use only visible-light cameras for monitoring objects in the road ahead, as opposed to the more costly, but more accurate, lidar systems. Lidar (light detection and ranging) — which is used on Alphabet’s Waymo self-driving cars — shoots out infrared lasers that can detect objects in the dark, through fog, and in other conditions that would obstruct visible light.

At the time, Musk defended his decision to go all in with visible light cameras by pointing to the power of AI to detect pretty much any object on the road, thanks to its firehose of training data, which it collects from millions of connected Tesla vehicles on the road.

But recent reports have called attention to what appear to be failings of the camera-only system in a series of accidents that have led to multiple injuries and even a death.

Rober’s video presents an easy-to-understand, powerful illustration of Tesla’s potential limitations when it encounters such obstructions in the road. Rober’s tests also showed that the camera-only system had a hard time seeing a dummy of a child in heavy fog and a deluge of water (which, to be fair, you would probably never experience unless you were driving under a waterfall, similar to the idea that you probably won’t encounter a wall thats painted to look like its surroundings).

Rober’s Tesla did stop safely in a few of his tests. It stopped when a stationary child dummy was in the center of the road in plain daylight, when the dummy popped out from behind a vehicle, and when the dummy was backlit with extremely bright floodlights.

Tesla has been on a downswing of late, as the stock has lost half its value in the past three months, sales are dropping alarmingly in Europe, and Musk’s DOGE side quest appears to be consuming all of his time and attention.

Tesla shares were down more than 5% today.

More Tech

See all Tech
tech
Jon Keegan

EPA: xAI’s Colossus data center illegally used gas turbines without permits

The Environmental Protection Agency has ruled that xAI violated the law when it used dozens of portable gas generators for its Colossus 1 data center without air quality permits.

When xAI set out to build Colossus 1 in Memphis, Tennessee, CEO Elon Musk wanted to move with unprecedented speed, avoiding all of the red tape that could slow such a big project down.

To power the 1-gigawatt data center, Musk took advantage of a local loophole that allowed portable gas generators to be used without any permits, as long as they did not spend more than 364 days in the same spot. That allowed xAI to bring in dozens of truck-sized gas generators to quickly supply the massive amount of power the data center needed to train xAI’s Grok model.

The new EPA rule says the use of such portable generators falls under federal regulation, and the company did need air quality permits to operate the turbines. xAI is also using dozens of such generators to power its Colossus 2 data center just over the border in Alabama.

To power the 1-gigawatt data center, Musk took advantage of a local loophole that allowed portable gas generators to be used without any permits, as long as they did not spend more than 364 days in the same spot. That allowed xAI to bring in dozens of truck-sized gas generators to quickly supply the massive amount of power the data center needed to train xAI’s Grok model.

The new EPA rule says the use of such portable generators falls under federal regulation, and the company did need air quality permits to operate the turbines. xAI is also using dozens of such generators to power its Colossus 2 data center just over the border in Alabama.

tech
Rani Molla

Trump to push Big Tech to fund new power plants as AI drives up electricity costs

President Donald Trump is expected to announce a plan Friday morning that would require Big Tech companies to bid on 15-year contracts for new electricity generation capacity. The move would effectively force companies to help fund new power plants in the PJM region as soaring demand from AI data centers pushes up electricity costs across the US power grid.

Earlier this week, Trump called on tech giants to “pay their own way,” arguing that households and small businesses should not bear the cost of power infrastructure needed to support energy-hungry data centers.

Microsoft quickly responded, saying it would “pay utility rates that are high enough to cover our electricity costs,” along with committing to other changes aimed at easing pressure on the grid. Other major tech companies are expected to follow suit, though Wedbush Securities analyst Dan Ives warned the added costs could slow the pace of data center build-outs.

As we’ve noted, forcing tech companies to shoulder higher electricity costs is likely to hit some firms harder than others. Companies like Microsoft, Google, and Amazon can pass at least some of those costs on to customers by selling data center capacity downstream. Meta, in contrast, does not have a cloud business, meaning its AI ambitions lack a direct revenue stream to offset rising power costs.

So far tech stocks don’t appear to be affected much in premarket trading. However utility companies most levered to the AI boom certainly are, with Vistra, Constellation Energy, and Talen Energy deep in the red ahead of the open as analysts at Jefferies warn that these firms face risks from this plan.

Earlier this week, Trump called on tech giants to “pay their own way,” arguing that households and small businesses should not bear the cost of power infrastructure needed to support energy-hungry data centers.

Microsoft quickly responded, saying it would “pay utility rates that are high enough to cover our electricity costs,” along with committing to other changes aimed at easing pressure on the grid. Other major tech companies are expected to follow suit, though Wedbush Securities analyst Dan Ives warned the added costs could slow the pace of data center build-outs.

As we’ve noted, forcing tech companies to shoulder higher electricity costs is likely to hit some firms harder than others. Companies like Microsoft, Google, and Amazon can pass at least some of those costs on to customers by selling data center capacity downstream. Meta, in contrast, does not have a cloud business, meaning its AI ambitions lack a direct revenue stream to offset rising power costs.

So far tech stocks don’t appear to be affected much in premarket trading. However utility companies most levered to the AI boom certainly are, with Vistra, Constellation Energy, and Talen Energy deep in the red ahead of the open as analysts at Jefferies warn that these firms face risks from this plan.

tech
Jon Keegan

OpenAI working to build a US supply chain for its hardware plans, including robots

When OpenAI purchased Jony Ive’s I/O, it entered the hardware business. The company is currently ramping up to produce a mysterious AI-powered gadget.

But OpenAI plans on making more than just consumer gadgets — it also plans on making data center hardware, and even robots.

Bloomberg reports that OpenAI has been on the hunt for US-based suppliers for silicon and motors for robotics, as well as cooling systems for data centers.

AI companies are looking toward robots as a logical next step for finding applications for their models.

OpenAI told Bloomberg that US companies building the AI brains of robots might have an edge against the Chinese hardware manufacturers that are currently making some impressive humanoid robots.

Bloomberg reports that OpenAI has been on the hunt for US-based suppliers for silicon and motors for robotics, as well as cooling systems for data centers.

AI companies are looking toward robots as a logical next step for finding applications for their models.

OpenAI told Bloomberg that US companies building the AI brains of robots might have an edge against the Chinese hardware manufacturers that are currently making some impressive humanoid robots.

Latest Stories

Sherwood Media, LLC produces fresh and unique perspectives on topical financial news and is a fully owned subsidiary of Robinhood Markets, Inc., and any views expressed here do not necessarily reflect the views of any other Robinhood affiliate, including Robinhood Markets, Inc., Robinhood Financial LLC, Robinhood Securities, LLC, Robinhood Crypto, LLC, or Robinhood Money, LLC.