What happened
Two government-backed startups in India, Soket AI and ZenteiQ, have released sovereign AI models that can run in air-gapped environments. Soket AI launched EKA-1, a 26.3 billion model built for maths and coding, while ZenteiQ released two open-source models, Dhīr 2B and Dhīr 5B, which are multilingual assistants that support English and 10 Indian languages. These models are designed for secure, on-premise and offline use in critical sectors.
Soket AI launched EKA-1, a 26.3 billion model built for maths and coding.
ZenteiQ released two open-source models, Dhīr 2B and Dhīr 5B, which are multilingual assistants that support English and 10 Indian languages.
The models are designed for secure, on-premise and offline use in critical sectors.
Soket AI's EKA-1 model was trained on about 7.5 trillion tokens covering all 22 scheduled Indian languages and English.
ZenteiQ's models were trained on Tensor Processing Units (TPUs) due to the scarcity of GPUs.
Source details: hindustantimes.com
Why it matters
The release of these sovereign AI models is significant because they can operate in air-gapped environments, which means they do not require a network connection to function. This makes them ideal for use in critical sectors such as defence and cybersecurity, where data security is paramount. The models have been trained on a large and have been designed to be secure and reliable. The safeguards in place, including protections against content on suicide and self-harm, are also a positive development. The use of TPUs for training the models is also noteworthy, as it demonstrates the ability to train AI models on alternative hardware.
The release of these sovereign AI models is significant because they can operate in air-gapped environments, which means they do not require a network connection to function.
This makes them ideal for use in critical sectors such as defence and cybersecurity, where data security is paramount.
The models have been trained on a large and have been designed to be secure and reliable.
The use of TPUs for training the models is also noteworthy, as it demonstrates the ability to train AI models on alternative hardware.
The safeguards in place, including protections against content on suicide and self-harm, are also a positive development.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
A route planner searches possible journeys using explicit rules. What does this illustrate about AI?
What to watch next
The impact of these sovereign AI models on the Indian economy and the potential applications in critical sectors such as defence and cybersecurity will be significant.
The impact of these sovereign AI models on the Indian economy will be significant.
The potential applications of these models in critical sectors such as defence and cybersecurity will be substantial.
The development of similar models in other countries will be worth watching.
The use of alternative hardware for training AI models is a notable trend.
The implementation of safeguards to prevent the misuse of AI models is crucial.