Written by: Chloe, PANews
Tether CEO Paolo Ardoino revealed Tether's latest local artificial intelligence Local AI development kit at the Plan ₿ event in Lugano, Switzerland, on October 26. This new AI development kit focuses on user privacy and is supported by peer-to-peer technology.
A prominent feature of Local AI is its versatility, allowing developers to run written code on various devices, from low-cost phones to high-end smartphones, laptops, or large systems equipped with H100 processors. Moreover, this technology can be applied to multiple devices, including smart refrigerators and automotive entertainment systems.
In March this year, Tether announced the establishment of a new artificial intelligence department called 'Tether data', dedicated to advancing its entry into the AI field with the goal of building the next generation of open AI models. Under the leadership of Paolo Ardoino, the newly established AI department Tether data focuses on developing 'open-source, multimodal AI models'. This significant initiative aims to enhance the accessibility and efficiency of AI and underscores Tether's commitment to promoting AI and peer-to-peer technology.
The launch of Local AI aims to break through the application scope of traditional computing devices.
At the recent Plan ₿ event in Lugano, Switzerland, Ardoino demonstrated how Local AI transcends the application scope of traditional computing devices. The software development kit (SDK) could ultimately be implemented in smart appliances, including refrigerators and automotive entertainment systems, highlighting its wide potential applications.
Local AI places special emphasis on 'local' computation, meaning that AI computations are completed on users' own devices, using P2P (peer-to-peer) technology to allow direct data exchange between devices.
Traditional network architecture has every image, website, or application stored on a server, each with its own IP address. When we browse the web, we send requests to the server through the IP address, indicating that the server controls all ownership of the user.
P2P means that each user on the network is responsible for storing all or part of all data. In addition to being able to make requests to other IP addresses, users also need to handle requests from others, effectively making themselves a server. This is also the goal of Tether to promote AI and peer-to-peer technology.
A key feature of Local AI is its ability to load and store model weights and fine-tuning data within a P2P framework. This approach ensures that users' personal sensitive information remains secure and reduces reliance on central servers. At the Plan ₿ event, Ardoino showcased a fully local and privacy-focused translation application that can run on any device, avoiding the sharing of personal data with third parties and being compatible with various models (e.g., Marian and Llama). Additionally, the foundation of Local AI is based on the open-source project 'holepunchto/bare', which plays a crucial role in its architecture.
Assuming a user is using a translation application developed based on the Local AI SDK, the application can directly obtain updated language models from other users. The user's personal device can directly share optimized translation results with other devices, and all this information exchange is encrypted and does not require going through a central server.
Finally, Tether plans to release the Local AI SDK as open-source software after completing comprehensive testing. This move will benefit developers and communities interested in AI and privacy-related technologies, marking Tether's continued expansion of its AI technology products beyond just stablecoin services.
Last year, after accumulating computing power capital by purchasing chips, a department was established this year to expand AI business.
In 2023, Tether's annual net profit reached $6.2 billion. Last September, Tether made a significant investment in AI hardware, spending $420 million to purchase 10,000 Nvidia H100 GPU chips, each valued at around $40,000, accounting for approximately 2% of Nvidia's total shipment of 550,000 H100 GPU chips for the year.
At that time, this transaction was part of Tether's collaboration with Bitcoin miner Northern Data Group, where Northern Data could obtain a large number of H100s, and Tether would acquire a 20% stake in Northern Data. Forbes also mentioned that Northern planned to rent these graphics cards to AI startups in urgent need of computing power, which could make Northern Data the largest GPU operator in Europe, apart from Amazon, Microsoft Azure, and Oracle.
In March this year, Tether announced the establishment of the Tether data department, which focuses its AI business expansion on several key areas. The first is to develop open-source multimodal AI models, establish new industry standards, and promote innovation and accessibility in AI technology. In an interview with WIRED, Ardoino mentioned, 'The concept of decentralization can be applied to different fields, such as AI. We have seen AI being heavily politicized. I also believe that having a role independent of traditional tech companies (such as Amazon, Microsoft, and Google) is very important.'
Secondly, Tether data will seamlessly integrate AI solutions into products on the market, using technology to solve real-world challenges. Finally, by actively participating in community contributions, Tether data is committed to advancing the field of open AI while showcasing its innovative projects.
With ample cash, Tether is targeting Amazon, Microsoft, and Google.
In June this year, Tether released a report stating that for the first half of 2024, Tether's profits were $5.2 billion, with total reserves of $118.5 billion. Ardoino took over as CEO in December last year after serving as CTO for six years, and under his leadership, Tether has made significant investments in energy development, Bitcoin mining, and communication technology to diversify its investments.
Ardoino stated that part of the funds would be used to establish a buffer for USDT reserves, while the rest would be invested in the company's new venture capital department, Tether Evo. The company has already acquired a majority stake in the neural implant technology startup Blackrock Neurotech. 'The technology of brain-computer interfaces (BCI) will become very important in the future, and it is essential to build brain and computer interfaces that respect personal privacy, ensuring that data remains local and is not collected by a single company operating the platform,' Ardoino said in an interview.
Additionally, Ardoino also mentioned in an interview that AI is another important area for Tether's business expansion. Tether's VC evaluates hundreds of proposals each month, most of which come directly from startups, and has invested about $2 billion in the fields of AI and alternative financial infrastructure. For Tether, investing in AI not only means technological innovation but also ensures that the company remains at the forefront of technology.