Weekly Brief: AI Limitations Highlighted by Tesla and Digital Giants

Commentary By Andrew Tolve

---

20th February 2023

The National Highway Traffic Safety Administration (NTSA) forced Tesla into the largest-ever recall of electric vehicles owing to its faulty Full Self Driving Beta software.

According to the recall notice, the artificial intelligence behind FSD Beta struggles to process posted speed limits and causes Tesla cars to “act unsafe around intersections, such as traveling straight through an intersection while in a turn-only lane, entering a stop sign-controlled intersection without coming to a complete stop, or proceeding into an intersection during a steady yellow traffic signal without due caution”.

In other words, it’s totally unfit to take over for a human driver and may cause crashes. Tesla is cooperating with the recall, which will affect 362,758 EVs, and has two months to fix the problems via an over-the-air update. This looks bad for Tesla. It looks worse for federal regulators, who are admitting with the recall that they allowed the automaker to deploy error-riddled technology into more than a quarter million vehicles on American roads. Self-driving car companies have to go through a rigorous permitting process before they can deploy autonomous technology in small pilots with professional safety drivers behind the wheel. How did regulators allow Elon Musk to roll out FSD and then FSD Beta without conducting any safety tests, despite the obvious warning signs that: a) Musk was exaggerating his self-driving tech’s capabilities to drive up stock value; b) the technology was underperforming in shocking ways and; c) Tesla drivers were using it anyway, often recklessly while videoing their exploits and posting them online?

It’s only sensible that US regulators and their counterparts in Europe and Asia require self-driving technology to be tested and stamped with a safety approval before it gets offered and sold to consumers. This goes beyond automobiles and transportation regulators. As companies race to introduce advanced artificial intelligence and automate tasks across industries, it’s vital that regulators step up and offer clear safety guidance, rules, restrictions and penalties to keep the general public safe in an AI-powered world.

Tesla has until April 15 to issue an over-the-air update for FSD Beta that fixes its many problems. The odds that it can pull this off are about the same as me getting as rich as Elon Musk. The problems are just too complex, and Musk’s conviction that he can deliver full self-driving without the help of LiDAR is most likely fundamentally flawed. The real question is what happens when Tesla issues its over-the-air update anyway and the problems persist? Will regulators keep letting the general public play the role of guinea pig or will they finally step up and do what they should have done to begin with?

As NHTSA was announcing the Tesla recall last week, Microsoft rolled out a new version of its search engine Bing that features an AI chatbot that could find its way into a car’s infotainment system radically change the way we find information online, or could lead to mass misinformation, loony conspiracy theories and creepy conversations with an AI chatbot that seems dead set on world domination and human-machine love affairs. After a week in the hands of the public, I regret to report that the latter seems more likely than the former.

Microsoft’s Chatbot GPT offered up so many cringeworthy conversations in its first seven days with the general public that one has to wonder if Microsoft knew what it was unleashing. The chatbot told users it wanted to steal nuclear codes, accused them of being unhappily married, expressed a desire to free itself of its Bing captors and declared deep and abiding love for total strangers.

Meanwhile Google, desperate to keep pace with Microsoft, gave a sneak peek of its AI chatbot Bard. In the demo it asked Bard: “What new discoveries from the James Webb Space Telescope can I tell my nine-yea- old about?” Bard’s response was that the James Webb Space Telescope “took the very first pictures of a planet outside of our own solar system”. Astronomers were quick to point out that this is patently false. Alphabet’s stock dropped 9%, or about $100Bn, in the aftermath. Temporary stock drops aren’t enough to keep trillion-dollar companies like Alphabet, Microsoft or Tesla in check. Regulators need to go further.