TL;DR:
- Real AI secures ISCRA project at Data Science Conference 2023.
- LEONARDO, Europe’s 4th largest AI cluster, to power the development of a Human-Centered Large Language Model (LLM).
- LEONARDO, with 14,000 Nvidia Ampere GPUs, ranks 4th globally and 2nd in Europe in computing power.
- Real AI’s CEO, Tarry Singh, envisions responsible AI development with UNINA and LEONARDO.
- Project HOMINIS aims to create ethical, bias-free AI systems using open datasets.
- HOMINIS goals include diverse dataset compilation, ablation studies, model optimization, and ethical alignment.
- Real AI’s partnership with UNINA and NVIDIA signifies a convergence of excellence in AI innovation.
Main AI News:
In a landmark announcement at the Data Science Conference 2023 in Belgrade, held on Thursday, November 23, Real AI emerged victorious in securing the prestigious ISCRA project. This momentous achievement grants Real AI the privilege of spearheading the development of Europe’s inaugural Human-Centered Large Language Model (LLM), harnessing the computational might of the world’s fourth-largest AI computer cluster, ‘LEONARDO.’
LEONARDO – Europe’s Computational Powerhouse
Situated within the CINECA data center in Bologna, LEONARDO stands as an epitome of high-performance computing prowess. This technological marvel is constructed around the formidable Atos BullSequana XH2000 computer system, boasting an impressive ensemble of nearly 14,000 Nvidia Ampere GPUs. As of its grand inauguration in November 2022, LEONARDO has solidified its position as the fourth fastest supercomputer globally and the second most formidable in Europe. Such technical supremacy places LEONARDO at the forefront of AI advancement within the continent.
Real AI’s Unprecedented Opportunity
“We aspire to furnish Europe with a definitive solution for responsible AI advancement. Collaborating with UNINA and the AI supercomputer cluster ‘LEONARDO,’ we are poised to turn this vision into reality,” remarked Tarry Singh, CEO of Real AI BV.
Real AI has embarked on a groundbreaking mission to construct the inaugural Human-Centered Large Language Model, leveraging LEONARDO’s substantial computational resources. This audacious endeavor propels Real AI to the vanguard of AI innovation, underpinned by a steadfast commitment to environmental sustainability and technological excellence.
Project HOMINIS: A Paradigm Shift in AI
Project HOMINIS, an acronym for “Human-centered Open-source Model for Intelligent, Neutral and Inclusive Systems,” aims to redefine the AI landscape. It endeavors to create ethical, bias-free AI systems by employing open and scrutinizable datasets. As part of its core objectives, HOMINIS seeks to address toxicity in web-scale datasets while proposing sustainable alternatives to conventional AI foundation models, thereby mitigating their environmental impact. The four pivotal pillars of this visionary project encompass:
- Aggregating invaluable and diverse datasets from various sources, including scientific papers and knowledge repositories, while eradicating any semblance of bias.
- Conducting comprehensive ablation studies on transformer models and exploring alternative architectural paradigms.
- Optimizing the foundational model through cutting-edge techniques, culminating in a preliminary release for collaborative community refinement.
- Implementing instruction tuning to achieve ethical alignment, ultimately leading to the release of a responsibly trained AI model.
Furthermore, HOMINIS endeavors to reduce energy consumption through innovative methodologies such as Flash Attention and routing, thereby enhancing data processing efficiency and elevating model inference and knowledge integration to new heights.
Real AI’s Collaborative Endeavors with UNINA and NVIDIA
The partnership between Real AI, UNINA (The University of Naples Federico II), and NVIDIA marks a convergence of excellence. UNINA boasts a distinguished track record of spearheading groundbreaking initiatives, including the establishment of the first Apple iOS Developer Academy in 2016 and the Cisco Digital Transformation Lab in 2018. Most recently, in tandem with Real AI, UNINA has assumed the role of founding member in Europe’s inaugural Human-Centered AI Masters program.
On the other hand, NVIDIA, the world’s preeminent AI company, has revolutionized the GPU landscape by transforming it into a potent computational juggernaut. NVIDIA’s influence spans across various domains, encompassing virtual reality, high-performance computing, and AI, making it an indispensable force in the technological panorama.
Conclusion:
Real AI’s success in securing the ISCRA project and its collaborative efforts with UNINA and NVIDIA signal a promising future for Europe’s AI market. The development of a Human-Centered LLM and the focus on ethical AI and sustainability will likely drive innovation and competitiveness in the region’s AI sector, making it a hub for responsible AI development and technological advancement.
