Unlock Idle GPUs: The Tetris Game of AI Resource Allocation
Related Episodes
AI Reshapes Engineering: From Code to Architecture
AI reshapes software engineering from coding to architecture and expands software's capabilities into human-like tasks, potentially automating significant portions of US labor by 2030.
View Episode Notes →
User Experience Trumps AI Power for Agent Adoption
monday.com reveals how shifting focus from tech capabilities to user experience, managing expectations through naming and entry points, and embedding interfaces solves AI adoption challenges.
View Episode Notes →
AI Hype Obscures Non-Determinism Risks and Undermines Open Knowledge
AI's non-deterministic nature and anthropomorphic interface create significant risks, obscuring limitations and undermining honest evaluation. Understand AI as a tool, not magic, for responsible innovation.
View Episode Notes →
AI Bots Reshape Internet Traffic -- Commerce Benefits, Publishers Threatened
AI bots are transforming internet traffic, growing 400% annually and blurring lines between bots and agents. Manage this shift by understanding intent, not just identification, to navigate new commerce and content challenges.
View Episode Notes →
Speed, Infrastructure, and Evals: The AI Agent Trifecta
Unlock AI agent potential: discover how lightning-fast inference, evolved evals, and smart tooling engineer agents that deliver speed, quality, and real-world effectiveness.
View Episode Notes →
AI Mediates Community Dialogue, Navigates Generative Risks
AI mediates dialogue and reveals innovation gaps, but rapid advancements demand societal readiness to prevent manipulation and ensure human intent.
View Episode Notes →Resources
Resources & Recommendations
Books
- "Just What Makes Your Beer Taste Better" - An analogy often used in the early days of cloud computing to explain that the cloud would handle infrastructure complexity, allowing users to focus on their core product.
Videos & Documentaries
- AlphaGo from DeepMind (2015) - A documentary about DeepMind's AI program AlphaGo, which inspired the guest to focus his energies on AI due to its general and extensible approach to problem-solving.
Tools & Software
- Codex - Mentioned as a background coding agent, indicating its use in asynchronous throughput and economically sensitive workloads.
- Cursor - Mentioned as a coding company utilizing mini-models and ensembles in an agentic context.
- Cognition - Mentioned as a coding company utilizing mini-models and ensembles in an agentic context.
People Mentioned
- Razi Abusa - Awarded a Populist badge on Stack Overflow for an answer that outscored the accepted answer for "how to find last merge in git".
- Jared Quincy Davis - CEO and founder of Mithril, who can be found on X at @jaredq_ or on LinkedIn.
Organizations & Institutions
- DeepMind - The Google-owned AI research company behind AlphaGo, which inspired the guest's career path.
- Mithril - The guest's company, which focuses on providing infrastructure, GPUs, and abstractions for AI development with better economics.
- OpenAI - Mentioned as a large customer that receives raw capacity allocations and a company using mini-models and ensembles.
- Google - Mentioned in the context of its background indexing work for classical search and as a major cloud provider.
- Microsoft - Mentioned as a major cloud provider that eventually entered the cloud market after AWS.
- Amazon Web Services (AWS) - Highlighted as establishing the original value propositions of the cloud, particularly elastic capacity and handling infrastructure complexity.
- Oracle - Mentioned as a partner cloud for Mithril's omnicloud strategy.
- Nebius - Mentioned as a partner cloud for Mithril's omnicloud strategy.
- Arm - The chip designer that discussed moving some GPU workload to the CPU in resource-constrained environments.
- DataBricks - Mentioned as a company using mini-models and ensembles.
Websites & Online Resources
- Stack Overflow - The platform where Razi Abusa answered the question "how to find last merge in git".
- X - Where Jared Quincy Davis can be found at @jaredq_.
- LinkedIn - Where Jared Quincy Davis can be found.
Other Resources
- AlphaFold - Mentioned as an example of what AlphaGo-like approaches can achieve in areas like scientific progress.
- Neural GCM - Referenced as an example of rewriting simulation systems built for CPUs to be "neural" (GPU-native) for scientific use cases.