Scaling Real-World AI with AMD: Insights from Industry Leaders

Diverse Applications of AI Across Industries

The panel discussion highlighted the diverse ways in which AI is being applied across various industries, showcasing the transformative potential of this technology:

Computational Biology and Genetics

• Arjun Raj, a professor of genetics and bioengineering, discussed how AI is enabling breakthroughs in understanding the complex biology of cells and developing predictive models to control cell behavior.
• The team at Somite Therapeutics is using large amounts of biological data and machine learning techniques powered by AMD GPUs to solve the challenge of turning biological signals into actionable instructions for cells, with the potential to unlock new therapeutic approaches.

Enterprise AI for Workflow Automation

• Casey, the head of US operations at Upstage AI, shared how the company is using AI to automate tedious, repetitive workflows for enterprises, particularly in industries like finance, healthcare, and the public sector.
• Upstage AI has developed a proprietary OCR and language model called Solar, trained on AMD GPUs, to enable template-free document extraction and validation, unlocking significant efficiency and cost savings for their customers.

Vision-Language Models for Physical AI

• Roman Hassenbach, the CEO and co-founder of Rambler AI, discussed the company's focus on building an end-to-end platform for training and deploying vision-language AI agents to understand and support human tasks in the real world.
• Rambler's technology aims to enable a granular understanding of the physical world, allowing AI systems to confirm if tasks are performed correctly and support workers in industrial and manufacturing settings.

The Importance of Hardware Choice and Ecosystem Partnerships

The panel emphasized the critical role of hardware and ecosystem partnerships in enabling the development and deployment of these diverse AI applications:

Hardware Flexibility and Efficiency

• The panelists highlighted the importance of having access to powerful, flexible, and efficient hardware like AMD's GPUs, which can run large language models and computer vision workloads on a single device, even at the edge.
• This hardware flexibility and efficiency is crucial for enterprises looking to deploy AI solutions in a cost-effective and scalable manner, transitioning from reliance on massive cloud-based models to more targeted, domain-specific models that can run on smaller hardware footprints.

Ecosystem Collaboration and Open Platforms

• The panelists praised AMD's approach of fostering an open ecosystem, where companies can build on top of AMD's hardware and software platforms, integrating their proprietary models and solutions.
• This open and collaborative approach, combined with AMD's commitment to providing technical support and partnership, enables companies like Upstage AI and Rambler AI to focus on their core competencies and innovations, rather than spending time on low-level infrastructure challenges.

Key Takeaways and Future Outlook

• The panelists highlighted the immense potential of AI to transform a wide range of industries, from computational biology and genetics to enterprise workflow automation and physical AI applications.
• Overcoming the "complexity barrier" in fields like biology and enabling granular understanding of the physical world through vision-language models were identified as critical frontiers for AI research and development.
• The availability of powerful, flexible, and efficient hardware like AMD's GPUs, coupled with an open and collaborative ecosystem, was seen as a key enabler for companies to bring these AI innovations to life and scale them in real-world applications.
• Looking ahead, the panelists emphasized the importance of open data sets, standardized models, and continued hardware and software advancements to unlock the next generation of AI capabilities and drive widespread adoption.