Semiconductor giant Intel has thrown its hat into the ring of the generative AI competition with the announcement of “Aurora genAI,” a revolutionary generative AI model specifically designed for scientific research.
On the same day, Intel also introduced the Aurora supercomputer, boasting an impressive 2 exaflops of computing power, which will serve as the backbone for the “Aurora genAI” model. Unveiled during the ISC23 keynote, the Aurora genAI model will be trained on a wide range of data, including general texts, scientific texts, scientific data, and domain-specific information related to various fields such as system biology, cancer research, climate science, cosmology, polymer chemistry/materials science, and more.
At the core of the Intel Aurora genAI model are two key foundations: Megatron and DeepSpeed. However, what truly sets this model apart is its staggering size, aiming to reach a massive 1 trillion parameters. To put this into perspective, the popular GPT-3.5 model, which powers the free version of ChatGPT, has approximately 175 billion parameters. The Aurora genAI model boasts over 5.7 times that number, promising unparalleled capabilities and potential.
The development of this generative AI model was a collaborative effort between Intel, the Argonne National Laboratory, and Hewlett Packard Enterprise (HPE).
Intel’s Aurora genAI will undergo extensive training on a vast array of data, including general texts, codes, scientific texts, and structured scientific data across fields such as biology, chemistry, materials science, physics, medicine, and more. The resulting model, armed with a trillion parameters, will find applications in diverse domains, ranging from molecular and materials design to various scientific experiments. Excitingly, the model will be instrumental in accelerating the identification of biological processes associated with cancer and other diseases, as well as suggesting potential targets for drug design.
The unveiling of Intel’s Aurora genAI marks a significant milestone in the field of generative AI and scientific research. With its unparalleled scale and the powerful Aurora supercomputer behind it, this groundbreaking AI model has the potential to reshape various scientific disciplines and open doors to novel discoveries and advancements in fields such as systems biology, polymer chemistry, energy materials, climate science, cosmology, and more.