AI Leaders Reveal the Next Wave of AI Breakthroughs (At FII Miami 2025) | EP #150

Key Concepts:

• Stable Diffusion: Open-source image generation AI model.
• Ultra-narrow AI: Fine-tuned AI models for specific tasks in content creation.
• Liquid Neural Networks: Warm-inspired AI architecture requiring minimal compute, enabling private AI.
• Large Quantitative Models (LQMs): AI models trained on scientific data for applications like drug discovery and materials science.
• Tensor Processors: Hardware designed for efficient tensor operations in AI.
• Open Source AI: AI models, software stacks, and infrastructure with publicly available code.
• AI-Native Companies: Startups built from the ground up with AI at their core.

Stability AI and the Future of Content Creation

• Stable Diffusion's Impact: Stability AI created Stable Diffusion, a market-leading image generation model with over 270 million downloads. This is significantly more than the second most popular model, which has only 9 million downloads.
• Ultra-Narrow AI for Film and TV: Stability AI is focusing on fine-tuning AI models for professional content creation workflows in film, TV, gaming, and advertising. This involves creating specialized models for tasks like rig removal, paint and rotoscope, and camera match plate construction. They are working on approximately two dozen models now, with a goal of 50-60 models.
• Creating Reality: The technology to create realistic video is already here for certain workflows. Full video generation on the fly is expected within 6-12 months.
• Prem's Perspective: Prem emphasizes that AI should be viewed as a tailwind, not a headwind, in the entertainment industry. He draws parallels to the transition from silent to talking movies and the adoption of digital filmmaking. He envisions Stability AI becoming a creative agent, empowering individuals to become creators.

Liquid AI and Private AI Capabilities

• Liquid Neural Networks: Liquid AI is building generative AI systems powered by liquid neural networks, a new AI architecture invented at MIT. This technology requires minimal compute, enabling AI to run locally on devices like phones and laptops.
• Private AI for Enterprises: Liquid AI provides private AI capabilities for enterprises with sensitive data or security concerns. This allows companies to use generative AI without relying on cloud-based solutions or GPU-based infrastructure.
• Applications of Private AI: Private AI can power various applications, including autonomous navigation for fighter jets, education tablets, cars, and satellites.
• Revenue and Investment: Liquid AI raised a significant round with G42 as one of the leads and is experiencing spiking revenues.
• Alexander's Perspective: Alexander believes that the world needs to make AI useful and that AI can bring value at every scale. He emphasizes the importance of "machine learning done right," which means consuming less energy to build and host AI systems. He envisions a future where every device is AI-enabled.

Sandbox AQ and Large Quantitative Models

• Focus on Quantitative AI: Sandbox AQ focuses on quantitative AI (LQM) rather than language AI (LLM). This involves creating AI models trained on scientific data for applications like drug discovery and materials science.
• Quantum Equations on GPUs: Sandbox AQ uses quantum equations on GPUs to model molecules and atoms. This is achieved by converting quantum equations into the language of GPUs, enabling them to perform complex calculations.
• Applications of LQMs: LQMs are used to develop new medicines for diseases like cancer and Alzheimer's, as well as to create advanced materials for various industries.
• Partnerships: Sandbox AQ has partnerships with companies like Aramco to convert hydrocarbons into higher-order chemicals.
• Jack's Perspective: Jack emphasizes the power of small teams armed with AI tools to solve complex problems. He also urges people to embrace AI to address the big challenges facing society, such as diseases and energy storage.

Toren and Democratizing AI Hardware

• Native Tensor Processor: Toren builds a native tensor processor that is simpler and easier to program than GPUs. This hardware is designed for efficient tensor operations, which are fundamental to AI.
• Open Source Software Stack: Toren open-sourced its software stack, allowing developers to see exactly how the hardware works and how it runs. This is intended to unlock AI applications that are currently hard to program with GPUs.
• Democratizing AI Hardware: Toren aims to democratize the hardware stack and the software stack, making it accessible to more people.
• Cost Reduction: Toren's systems are targeted to be 5 to 10 times cheaper than current systems.
• Jim's Perspective: Jim believes that AI doesn't have to be unbelievably expensive, big, or proprietary. He wants to make computational hardware available to lots of people so they can use it. He also encourages people to start small with AI and iterate on how their systems work.

Quantum Black and AI Transformation in Enterprises

• Bridging the Gap: Quantum Black aims to bridge the gap between the hype around AI and its actual impact on businesses. They work to increase the success rate of AI use cases from 11% to 100%.
• Global Presence: Quantum Black has 5,000 people working in 50 countries, transforming nations and companies.
• Leadership Challenge: Alexander emphasizes that AI transformation is a leadership challenge that requires buy-in from the top of the organization. He also stresses the importance of data quality, architecture changes, and talent development.
• Competing with AI-Native Companies: Companies need to compete with AI-native startups by becoming AI-first organizations.
• Alexander's Perspective: Alexander believes that AI should be used to make the world a better place. He encourages people to educate themselves about AI and to embrace it fully to transform their enterprises.

Conclusion:

The panel discussion highlights the transformative potential of AI across various industries. It emphasizes the importance of open-source models, private AI capabilities, and specialized hardware to democratize access to AI and enable innovation. The panelists also stress the need for leadership buy-in and a strategic approach to AI transformation in enterprises. The key takeaway is that AI is not just a technological advancement but a fundamental shift that requires a new mindset and a willingness to embrace change.