During the debut of the first MIT Future Fest, Tony Fadell displayed a slide showing three once highly sought-after AI devices that are now no longer in production: Rabbit R1, Humane AI Pin, and Limitless pendants.
"These can be considered as the 'first generation' of AI products. Based on my background, these companies all called me to ask for my help," said Fadell, "but I didn't accept it."
It's not surprising that these companies want to hear the advice of Fadell. He is hailed as the "father of iPod", one of the co-creators of iPhone, and the founder of the smart thermostat company Nest, which was later acquired by Google.
"You must truly understand what you want to do and what pain points you want to solve," he said. "In these cases, none of them meet any real need; they're just interesting technologies for geeks. You might think, 'Well, that's cool, but it doesn't really integrate into my life.'"
These early AI devices promised to provide the convenience of a personal assistant, but in reality, they were not very user-friendly. Even if they had been useful, most people had never had an assistant before, so they didn't know how to use one – or how to gradually develop trust in a personal robotic assistant.
"Less than 0.01% of the global population has actually had a human [assistant] to do things for them, so when [companies] say 'Oh, we want an assistant,' it's often people who already have one," he said. "Most consumers don't even know what an assistant is."
Fadell Please further explain to us, the 99.99% of people, that hiring a trustworthy personal assistant is not as simple as picking the most outstanding resume from a pile of job applications.
“It took me several years to figure out how to use [assistants] most effectively, before I entrusted them with the most sensitive data and even had them act as my agents to arrange meetings and deal with banks for me,” he said. “I understand that we have such a vision for the future, but there are still many steps to be taken before ordinary consumers, or even ordinary businesses, can follow this path.”
Discussing how to build trust in personal assistants can easily seem a bit self-centered, but the viewpoint of Fadell indeed makes sense. When hiring a human assistant, you wouldn’t give them your bank account login information on the first day. We are moving towards a world where anyone can have a personal AI assistant, but these tools only have value if we can trust them to be completely secure.
So far, it seems even harder to build trust in an AI assistant than in humans. Meta recently launched a universal AI assistant Muse, but users have reason to remain cautious. A security researcher quickly discovered a serious vulnerability in Muse shortly after its release, and a recent report by 404 Media stated that some employees of Meta identified security issues before its release, prompting multiple teams to engage in a "crazy sprint" to fix these problems, working overtime.
“No matter what we entrust to some form of intelligence, trust and security will be of paramount importance,” said Fadell. “The only company that comes to mind right now – perhaps there’s another as well – is Apple; they are capable of achieving this. Apple owns all the hardware, all the chips, all the components, but they don’t have a complete AI ecosystem.”
Even if in the end, it's not Apple that creates the most popular AI proxy, Fadell predicts that a successful proxy must run only on the device side. This is for both privacy reasons and to keep the technology more lightweight.
“When you hear all sorts of claims, like ‘Oh, data centers will conquer the world’… I don’t believe it,” he said. “I’ve seen this before. I’ve seen such development paths during the internet era, and similar things. We already have such powerful computing power on the device side now, and it will only get stronger, but they’re still powered by batteries.”
By keeping sensitive data on the device side rather than transmitting it through the cloud, Apple has earned users' trust in sharing biometric data thanks to features such as Face ID. As Fadell states, Apple seems to be more trusted by consumers in terms of privacy compared to its competitors; however, unlike other top tech companies, Apple does not have its own world-class AI model. The new version of Siri AI runs on a customized version of Google's Gemini.
Although Apple lags behind in its own AI, it is far ahead in terms of hardware. Fadell speculates that companies like Meta and OpenAI have turned their attention to hardware devices because they do not possess billions of active devices like Apple does.
“Because they can’t access the sensors on your phone, they’ll say, ‘Oh, we need video.’ Then you have to confirm, ‘Okay, I’ll provide you with video.’ After that, they say, ‘We need audio.’ You check that box. ‘We need your GPS location.’ You check that box too. Before you know it, you’ve agreed to let them access 20 or even more types of data,’ he said. “So what do they do with it? They make a device that has all these capabilities but no screen… Then they connect to your phone via Bluetooth or Wi – Fi, and from there, they send the data back to the internet, or to a 5G network. In this way, they can obtain all the sensor data.”
Given that he has been involved in creating iPhone and iPod, one might think that Fadell still views the company where he once worked through a filter. Although he is willing to give Apple its due when due, he is not shy of criticizing them when he believes they have made mistakes.
When the audience asked about the role of luck in finding a product market fit, Fadell pointed out that it is even more difficult for startups to “create” such luck, as a single failed product could mean the end of the company.
“Since it’s a startup, you only have one chance,” he joked, “it’s not like Apple when they did Vision Pro.”
