This small phone with artificial intelligence has some surprising ideas.
I saw it in action, and I'm still not sure if what I saw was real.
One year ago, the first demonstration of the Humane AI pin took place, which also turned out to be the last. At this year's Mobile World Congress, another artificial intelligence device emerges, attempting to fill the void left by Humane, and it does so in an even more surprising way. This new phone collects a substantial amount of personal information, both past and present, to create an artificial intelligence avatar that acts as a virtual assistant. Its conception combines elements from the Rabbit R1, Gemini Assistant, and science fiction, also incorporating aspects of blockchain.
The company behind Newnal AI is based in Korea and previously gained attention for developing a blockchain-based vaccine verification method, widely used in the country. In my conversation with founder YT Kim, he emphasized, “We never sold cryptocurrencies,” demonstrating they are aware of the negative perceptions surrounding blockchain today.
The Newnal phone trains an artificial intelligence model based on the user's characteristics, which involves considerable collection of personal data. The concept revolves around the user downloading their personal data history from platforms like Meta and Google, as well as medical and financial documents, to then provide this information to the Newnal system. Kim assures that the company encrypts this “personal knowledge graph” and fragments it for storage across multiple third-party cloud servers, so that only the data owner can access it.
The devices I observed were not final versions, but Newnal plans to launch the phone at the end of spring. There are significant privacy implications in all this, no doubt. In an effort to be transparent about their operations, Newnal has published an unusual amount of technical documentation on their website. While I cannot verify the security of their methods, this is just a first glimpse.
The operation of the device is equally intriguing: the information collected, along with the user's appearance and voice, is used to create a virtual version that can be seen and move on a small screen located above the main phone screen. To activate it, a button on the side must be pressed, allowing for various actions to be requested. During the demonstration, it was shown how the assistant helped purchase auto insurance. The AI appeared to compare policies and ultimately completed the necessary forms on Geico's website, even buying a used car to validate the demonstration, which is remarkable. I watched how the model completed each step, showing its sources of information while doing so, and in the end, the policy payment was made.
The device, which Kim describes as inspired by the iPhone 5S, has an appealing design, in a black and silver rectangle, with a dedicated upper screen for the AI. Newnal claims it operates on a hybrid operating system, combining its own platform and Android. Although the versions I saw were prototypes, Kim communicated that they have global launch plans for May 1 at a price of $375 each, shipping to customers two months after pre-orders.
While the context could evoke the launch of the Humane AI pin, the device strongly resembles the Rabbit R1, promising a lot at a relatively low price. Previously, the launch of the R1 had generated skepticism, although the demonstrations conducted by Newnal were more convincing, fulfilling all the proposed requests. However, the question of trust in having an AI complete insurance forms raises doubts. The company must prove that its approach is not only functional in controlled demos but also viable in practice.
In a landscape where other devices at the Mobile World Congress seem to follow established formulas, Newnal presents a bold idea that challenges convention, although the validity of this idea will need to be demonstrated outside of a showcase environment.
