The news that Apple has shelved its autonomous car plan–Project Titan–has sent ripples across technology and automotive industries. The company has pumped untold resources into Titan for almost a decade.
It is now apparently relocating some personnel in the Titan team to artificial intelligence research and development and probably letting the others go.
Specific reasons for the strategic switch remain unclear, but some inferences can be made. On the face of it, the foray into Titan made sense.
Apple has made its reputation and earned its profits by single-mindedly following through with a strategy of developing integrated, consumer-facing devices. Products like the Mac, iPad, iPhone, Apple Watch, and even the iPod have a design principle in common: They are built around tight integration of software and hardware, and superglued to the iOS ecosystem.
This means that besides the fact that the devices are individually best-in-class (or close to it), they secure high margins for the company. Apple has tight control over the hardware supply chain, and the smooth interoperability of the devices creates a very loyal following.
If you own an iPhone for example, and you’re in the market for a laptop or desktop, it’s a no-brainer to opt for a Mac. The operating system will be familiar and easy to navigate, and your data and content can be smoothly transferred between devices and stored on the cloud. Moreover, every Apple user shops for new apps, accessories, etc. on the Apple iStore where again, interoperability is a plus.
To move from consumer electronics to a self-driving electric car may seem ambitious. But any autonomous vehicle is essentially about synthesizing software with hardware.
Indeed, Google’s Waymo autonomous ride-hailing fleet consists of operating systems installed on modified assembly-line vehicles. Elon Musk of Tesla also says that the big deal with the world’s most valuable electric car company is the software.
So the world’s most valuable consumer electronics company decided to enter autonomous electric vehicles. But people have been trying to build autonomous vehicles for a long time – the better part of two decades and nobody has really succeeded yet. It’s an intractable problem, except in very controlled environments, even after recent huge breakthroughs in AI capability.
An autonomous vehicle must identify and understand everything that can happen on the road. It needs to see pedestrians at zebra crossings, identify lights and police barriers, and steer clear of mad cyclists and aggressive drivers.
This may be fine for an autobahn with law-abiding German drivers, or in a small town in California or Arizona (the environments where Waymo operates) with wide roads and little traffic. But in the chaotic mixed traffic of a large metropolitan area, it would require hugely capable AI. Nobody has been able to demonstrate a grasp of this yet to the satisfaction of sundry traffic authorities, and importantly, vehicle insurers.
Moreover, an autonomous vehicle must respond to natural language prompts. If a passenger says “I want to get a haircut at Khan Market –collect me an hour later” this has to be understood by the car the same way it would by a human chauffeur.
The car must convert “hair cut at Khan Market” into GPS coordinates, and drop the passenger at the nearest reasonable spot to the preferred hairdresser. It must make a “fuzzy” assumption of what “an hour” means (maybe 50 minutes, maybe 70), and return during that time slot, or wait in parking and contact the user at that time. Again, that requires high-quality AI.
The projected timeline for Project Titan’s launch was reportedly 2028 with meaningful revenues and profitability expected to take even longer to come through. In the meantime, peers like Tesla and Google would have locked in many patents and achieved far higher adoption.
This lag in development could have been a good reason to shelve Project Titan. As to repurposing the Titan team to work on AI, that also does make sense. The team would have been heavily focussed on AI anyway.
Given recent breakout developments in large language model, or LLMs, (ChatGPT was only unveiled 15 months ago), that’s an area of great interest. People are already developing use cases around ChatGPT and Gemini.
Apple could introduce “iAI” or whatever it chooses to call its own version of an LLM to market quickly. It has a dedicated user base that will be happy to give any new product or service from Apple a fair trial. Revenues and return on investment could come quicker and could be higher than the autonomous vehicle venture.
#Apples #switch #focus #selfdriving #cars #sense
