In what marks the company’s latest stride in the world of AI, Amazon has acquired the AI wearable startup Bee.
Both the CEO and CTO of the startup announced the news on LinkedIn, while Amazon confirmed it to several media outlets.
“When we started Bee, we imagined a world where AI is truly personal, where your life is understood and enhanced by technology that learns with you,” said Maria de Lourdes Zollo, the CEO and co-founder of Bee.
According to Crunchbase, the company has raised a total of $7 million across two funding rounds. For those unfamiliar with the company, reading Zollo’s LinkedIn post raises questions about how the Bee AI wearable differs from existing devices, such as the Humane AI Pin or the Rabbit r1.
This naturally leads to broader questions: What exactly does Bee do, and what sets their AI wearable apart?
The $50 Device That Listens to Everything Around You
Bee offers a $49.99 wearable device that listens, summarises, and records sounds from a user’s daily life. Worn as a bracelet, it records and transcribes conversations, which users can review later on their iOS app.
Using the app, users can ask questions to interpret their recordings and gain insights. Bee can also be linked to Google accounts like Gmail and Google Calendar, to help it recognise who the user is communicating with and their location, enabling it to offer more relevant insights. It can also be used as a voice assistant, and the results can be heard on the iOS app.
Source: Bee
According to the company, the device lasts for seven days on a full charge and supports 40 languages. Their website states that it is currently available for shipping only in the United States.
Before discussing how well the device works, or not, it certainly raises eyebrows when it comes to privacy. As tech YouTuber Becca Farsace puts it, “This device, that is designed to help me, creates far more questions about other people’s privacy than my own.”
The company’s privacy policy states that it does not store audio recordings and that all processing is done in real time, adhering to high security standards, with data being immediately deleted afterwards. Moreover, Bee explained that it also provides users with the opportunity to delete their data at any point in time, permanently.
Addressing the question of whether users’ data will be used for AI training or sold elsewhere, Bee said, “absolutely not.” She added that it does not train an AI model with users’ data, nor does it sell, monetise, or share it with third parties.
That said, the company plans to address various concerns regarding its privacy, including those raised by Farsace.
Bee said it is actively developing a ‘fencing’ feature, where users can set boundaries around what topics Bee is allowed to learn. “You’ll have the ability to define topics that should never be processed, and designate physical locations where Bee will automatically pause learning,” the company stated.
Zollo mentioned to The Verge earlier this year that the company is working on a “liveness detection” update, which aims to prevent Bee from mistakenly interpreting broadcast messages as conversations.
Moreover, these concerns aren’t limited to just the Bee but also extend to devices like Meta’s Ray-Ban glasses and Google’s upcoming Astra glasses.
Ethan Sutin, CTO of Bee, said on a podcast interview, “We kind of adapt to these new technologies all the time. When the Ring [doorbell cameras] came out, it was quite controversial, but now, people understand that lots of them have cameras on their doors.”
Sutin indicated that with more such devices coming out, users will be more aware of the devices they are subjected to and will change both their behaviours and expectations.
He stated that regardless of local laws concerning consent for voice recording, the company takes extensive precautions to avoid capturing personal information of nearby individuals through speaker identification and similar features.
Never Forget Again
Farsace, in a YouTube video review, stated that while the device allows one to manually tag all speakers in a conversation recorded by Bee, it faced difficulties in distinguishing them in the first place.
“This confusion causes errors in its answers,” she said, highlighting Bee’s struggle in answering questions specific to a person in the conversation. “If it’s not 100% accurate on its own, is it useful at all?”
She also recounted her experience of being part of a conversation that involved a maximum of 15 speakers. Yet, Bee showed 34 people in the recorded conversation.
Despite these occasional difficulties, Farsace pointed out that the device was indeed useful in some situations. For example, when she bought a wine she enjoyed but later forgot its name, Bee helped her find it based on transcripts of her conversations.
“Bee also creates an ongoing list of suggested to-dos that are actually quite good and surprising,” she said, speaking about how the device picked up on her travel plans in her conversations, and wrote them down in a to-do list. “These are the things that are important to me, but I would have never formally written down,” Farsace added.
Similarly, a review from The Verge stated that the device is ‘most helpful’ when it comes to summarising meetings and helping the user remember to do random tasks.
Although the public reviews are a few months old, it is reasonable to expect several updates to enhance the user experience following Amazon’s acquisition. The company has been regularly updating its app on the App Store, with numerous variants that include “minor bug fixes”, as noted in the update descriptions.
Besides, it is worth noting that the device is the company’s first-generation model.
AIM reached out to Bee and its CTO, Sutin, to understand if the company had improved the user experience and what the product roadmap looks like following the acquisition, but did not elicit a response at the time of publication.
The post Amazon Acquires Bee and Its $50 Device That Turns Conversations Into To-Do Lists appeared first on Analytics India Magazine.
