Jolla, once known for its mobile manufacturing, has pivoted to focus on privacy-driven AI technologies through its sister startup, Venho.ai. They have unveiled a new AI assistant that promises to be a “fully private” alternative to data-hungry cloud companies that sift through your personal data.
This AI assistant is engineered to work seamlessly with applications such as email, calendars, and social media accounts. It provides users with a conversational interface that delivers information and performs tasks on their behalf, including summarizing emails and documents, scheduling meetings, filtering social media content, and conducting web searches.
The tool is also capable of generating new AI agents on demand, provided the required API keys are accessible. An AI agent marketplace will also be launched next month as part of their offering.
Additionally, they are emphasizing the role of the AI assistant as both a shopping companion and a personal memory aid. This allows users to research potential purchases or leave notes for the AI to remember, ensuring that this personal information remains secure and is not shared with large data-mining companies.
However, this AI assistant software is not an isolated development; the “Jolla with Venho” team has spent the last year developing proprietary AI hardware, previously described as ‘AI agents in a box,’ designed to realize this vision of personalized AI convenience without compromising privacy.
Their strategy leverages smaller AI models that can be hosted locally, enabling the effective management of various data tasks along with a vector database that gathers information from linked accounts, enhancing the user experience during queries.
The forthcoming AI assistant software will function atop this backend data framework, facilitating seamless transitions between different AI agents based on user requirements.
Navigating Through the Noise
During an exclusive preview at the Mobile World Congress (MWC) in Barcelona, co-founders Antti Saarnio and Jolla veteran Sami Pienimäki divulged their plans to establish a decentralized AI operating system that could challenge dominant cloud firms as generative AI shaping software usage evolves.
They cite their extensive experience in mobile hardware and OS development (including their work with Sailfish) and Android app integrations, suggesting this history positions them favorably in the competitive landscape.
“The cognitive overload due to AI is substantial right now, and we need assistance to filter through the noise,” Saarnio affirms, emphasizing that the AI assistant will be an indispensable tool for simplifying information. “It’s a personal tool — you fully own this.”
The AI assistant, branded as Mindy, features a customizable avatar with a female aesthetic and will soon be available as a subscription service hosted on a private cloud managed by Venho.ai.
In a demonstration, TechCrunch observed Saarnio engaging with the AI assistant on his laptop through a chat interface, asking questions about his emails, creating tasks from email-related activities, and scheduling meetings. (The assistant also supports voice queries.)
There was a slight delay before the assistant responded initially, but subsequent replies were more prompt. Saarnio indicated they are working toward further accelerating response times, aiming to achieve responses in under one second.
A web search query, where he requested information about a specific company, did not yield results. However, Saarnio clarified that this was due to his laptop not being connected to the Internet, causing the Google Search API to be inactive.
Beyond being hosted on their private cloud, the AI assistant will become accessible on the Mind2 Jolla AI device next month. This hardware has been distributed to early adopters since January.
This means that owners of Mind2 can self-host the AI assistant on their personal server-like hardware, eliminating the need for third-party cloud services to store their data (even the “private cloud” provided by Venho). Especially for queries that utilize the small AI models integrated into the device, which includes DeepSeek’s 1.5BN parameter model and Meta’s 1B parameter Llama model, according to Saarnio.
“It’s capable of processing and providing accurate answers — for instance, detailing your emails and their content — with minimal mistakes. However, it can get confused if given an extensive context, like uploading a large document for queries,” he explains, offering insights into when the device might need to utilize external cloud resources.
More complex queries may require accessing larger language models (LLMs), which could compromise privacy since user data would be subjected to external terms and conditions. Yet, with a conversational assistant in place, users will have the ability to specify how they want the AI to operate, such as prohibiting the sharing of health information with LLMs, allowing further personalization to their comfort level.
For early adopters (the first 1,000 users), the pricing for the AI Assistant is set at $10 per month following a 14-day free trial, with the final price expected to be around $20 per month, according to Saarnio. (For those purchasing the Jolla hardware, the 16GB RAM, 128GB memory Mind 2 device is priced at €699, although there are still discounts available for early purchasers.)
Since our last encounter with Jolla’s hardware, the now-available gadget has expanded in size. Pienimäki noted that the increased dimensions (now roughly equivalent to a small, thick paperback) are primarily due to enhanced heat management requirements, as it now includes a larger heatsink and additional thermal management components alongside the processing hardware.
The final assembly of this product takes place in Finland, at a previous Nokia manufacturing site in Salo, hinting at the potential revival of notable Finnish technological achievements. (It’s worth noting that Sailfish OS was created as a spinoff of an abandoned Nokia software project by former employees.)
So far, around 500 Mind 2 devices have been shipped to early adopters, tapping into interest from the Sailfish enthusiast community. Saarnio mentioned that they have received valuable feedback and bug report assistance from these early users via a Discord channel. “I recommend this approach to any startup,” he remarked.
While they focus on selling this hardware and its associated services, they also see promising B2B prospects emerging.
In fact, Saarnio noted that some telecommunications companies recognize the potential of this hardware as a home hub solution beneficial for multiple users in shared living spaces. “It’s akin to having a private Amazon Alexa,” Pienimäki adds, which encouraged the team to expedite the development of Mindy “to demonstrate the conversational capabilities of AI,” as stated by Saarnio.
Compiled by Techarena.au.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
