Chip-making company, Intel, in partnership with Argonne National Laboratory has announced plans to develop Intel Aurora genAI, a set of generative artificial intelligence (AI) models with a staggering 1 trillion parameters, built for the world’s scientific community.
Aurora Supercomputer to Power GenAI
During this year’s International Supercomputing Conference(ISC) High-Performance Conference(HPC) being held in Hamburg, Germany since May 21, Intel presented leadership performance for high-performance computing (HPC) and AI workloads and also shared its portfolio of future HPC and AI products united by the one API open programming model.
The company also unveiled an international initiative to use the Aurora supercomputer to develop generative AI models for science and society.
“The project aims to leverage the full potential of the Aurora supercomputer to produce a resource that can be used for downstream science at the Department of Energy labs and in collaboration with others,” said Rick Stevens, Argonne associate laboratory director.
The generative models will be powered by the Aurora supercomputer which has 2 Exaflops of double precision compute performance meant to power scientific and engineering endeavors.
This supercomputer hosts close to 64,000 GPUs and 21,250 CPUs as well as more than 1,000 Distributed Asynchronous Object Storage (DAOS) storage nodes providing enough power to accelerate the process of developing the models.
Sitting at the heart of the Intel Aurora genAI models will be two essential frameworks, NVIDIA’s Megatron and Microsoft’s DeepSpeed.
These two powerful frameworks, which were used to develop Megatron-Turing NLG 530B, provide data, tensor, and pipeline parallelism which play a specific role in improving memory and compute efficiency.
The Intel Aurora genAI models will be trained on massive datasets of general and scientific texts, code, and structured scientific data from biology, chemistry, materials science, physics, and medicine.
Biology, chemistry, energy materials, climate science and more… @Argonne, @HPE and Intel are creating a gen AI model for scientific research.
Aurora, with over 60K Intel MAX GPUs, a fast IO system and all solid-state mass storage, will train #GenAI models at scale. #ISC23 pic.twitter.com/YWVROfsufO
— Intel News (@intelnews) May 22, 2023
With this, Intel expects to create models with 1 trillion parameters, which in comparison to the free and public versions of OpenAI’s Chatbot, ChatGPT with 157 million parameters, creates a remarkable 5.7-fold increase in the number of parameters.
Generative AI Models for Science and Engineering
The AI models will be employed for scientific tasks “in a variety of scientific applications, from the design of molecules and materials to the synthesis of knowledge across millions of sources to suggest new and interesting experiments in systems biology, polymer chemistry and energy materials, climate science, and cosmology,” said Intel.
Additionally, Aurora genAI models, which will also involve the collaboration of Hewlett Packard Enterprise, are expected to speed up the identification of biological processes linked to cancer and other diseases and offer potential drug design targets.
Aside from science, the models also have the potential to be used across other sectors like financial modeling, natural language processing, machine translation, image recognition, and speech recognition.
Intel aims to develop and complete the Intel Aurora genAI models by 2024 making it a notable achievement in the world of generative AI models.
Argonne National Laboratory’s Aurora 2+ ExaFLOPS System will be utilized in many other projects in the science and engineering space. For instance, the supercomputer will power the AI model built by Dr. William Tang, Principal Research Physicist at the Princeton Plasma Physics Lab to predict unwanted disruptions in reactor operation.
“Aurora will ingest massive amounts of data from present-day reactors to train the AI model. The model may then be deployed at an experiment, to trigger control mechanisms that prevent upcoming disruptions,” Intel wrote.
Related Articles:
- Best AI Text Generators
- Meta Drops Open-Source AI That Can Recognize and Generate Speech in Over 1,000 Languages
- What You Need to Know Before Telling the ChatGPT App Anything Personal
What's the Best Crypto to Buy Now?
- B2C Listed the Top Rated Cryptocurrencies for 2023
- Get Early Access to Presales & Private Sales
- KYC Verified & Audited, Public Teams
- Most Voted for Tokens on CoinSniper
- Upcoming Listings on Exchanges, NFT Drops