Tesla’s EV Broth: Elon Musk brings AI to autopilot, forced to take over after car goes bonkers

After nearly ten years of unfulfilled commitments and deceptive marketing, Tesla’s Full Self-Driving feature remains far from capable of safely transporting you to your destination in a Tesla vehicle. This is especially true if it requires your intervention to avoid accidents.

For instance, CEO Elon Musk used the now-renamed X-formerly-Twitter to live stream a test drive using an unreleased version of the EV’s Full Self-Driving software, a feature that they call Autpilot. However, the results were less than impressive. Reports indicate that the software, which Musk touted as “mind-blowing,” caused him to abruptly enter an intersection and even tried to jump a red traffic light.

Musk’s intervention
Musk’s response during the livestream was telling. He exclaimed, “Ohhh, intervention!” followed by laughter, admitting, “That’s why we’ve not released this to the public yet.”

This specific version, labelled v12, relies on neural networks to recognize elements like traffic signals, pedestrians, and other vehicles. Earlier, Musk had assured that “v12 is reserved for when FSD is end-to-end AI, from images to steering, brakes and acceleration out.”

In simpler terms, Musk rejects the use of radar, a technology adopted by Tesla’s competitors, and instead believes that artificial intelligence can rely solely on the vehicle’s cameras.

However, as numerous individuals have pointed out, this approach renders the technology more unpredictable and harder to troubleshoot when malfunctions occur.

Reinventing the wheel
Musk also pledged that v12 would be the initial release of the software not labelled as a “beta,” implying that even loftier claims about its capabilities should be expected.

Following Musk’s recent dicey experience, he presented a seemingly straightforward solution: to train the deep learning models with a more extensive dataset.

Musk stated during the livestream, “So with that intervention we just had, the solution is essentially to feed the network a bunch more video of traffic lights. That was a controlled left turn where there was a green light for the left turn but not a green light to go straight.”

In essence, Musk is postponing any resolution by making further assurances regarding future iterations of the software, which was initially introduced in beta form to a limited group of drivers in October 2020.

The question arises as to whether Musk’s words should still carry weight at this stage. Since at least 2014, Musk has been making grand promises about achieving complete autonomy for Tesla vehicles.

Persistent Issues
And yet, numerous glaring issues continue. These range from vehicles colliding with motorcycles or median barriers on highways to losing control in snowy conditions and even colliding with stationary emergency vehicles. The latter situation prompted a formal investigation by the US National Highway Traffic Safety Administration.

The scrutiny isn’t confined to the NHTSA alone. Even the Justice Department has initiated an investigation into Tesla due to allegations of deceptive marketing related to its driver assistance system.

And this doesn’t even touch on the multitude of lawsuits, including a class-action suit, that Tesla faces over accidents and fatalities reportedly linked to its software.

Nearly ten years down the line, Full Self-Driving continues to pose risks on public roads, as Musk’s own livestream demonstrated.

While his ardent supporters on social media were quick to emphasize that the remainder of Musk’s test drive went off without significant problems, the severity of running a red light or navigating intersections at the wrong time cannot be understated.