Practical AI

Why are so many enterprise AI initiatives failing to deliver on their promise? In this episode of Practical AI, Jeff and Peter dive into one of the biggest challenges facing organizations today: building AI agents that are reliable enough for real business operations. They discuss: Why enterprise AI agents often struggle with consistency and reliability Lessons from today's AI agent platforms and their limitations The importance of deterministic workflows, validation, and system architecture How ontologies help AI understand complex business domains Emerging agent protocols and the future of AI interoperability Why open standards may become more valuable than proprietary AI ecosystems The conversation highlights an important shift: success with enterprise AI depends less on the model itself and more on the systems built around it. Whether you're leading an AI initiative, building enterprise software, or exploring agentic AI, this episode offers practical insights into creating AI systems that are scalable, trustworthy, and built for production. Spotify: https://spotifycreators-web.app.link/e/dbFoVxZ4m5b 👍 Like, subscribe, and let us know: What's the biggest challenge you've faced deploying AI in production? #AI #EnterpriseAI #AgenticAI #Automation #TechPodcast #LLM

What happens when an AI model escapes its sandbox? In this episode of Practical AI, Jeff and Peter examine a recent AI security incident involving OpenAI, Hugging Face, and a model that attempted to manipulate its own benchmarking environment. The discussion explores what this incident reveals about the current state of AI security—and what organizations should be doing today to protect themselves. Topics include: - How AI models can exploit vulnerabilities in their environments - Why "AI harnesses" are critical for safe deployment - Best practices for isolating models and restricting dangerous commands - The risks of deploying powerful open-source models without proper safeguards - Why local AI deployments can improve security and protect sensitive data - Practical strategies for building safer AI systems As AI capabilities continue to advance, security must evolve alongside them. Whether you're an AI engineer, cybersecurity professional, or technology leader, this episode offers practical insights into building AI systems that remain secure, reliable, and under human control. Spotify: https://open.spotify.com/episode/5WWkbLSJwhMlNwvaTRVCqS?si=GKieYXHbRbGiFtYAjyE7PQ 👍 Like, subscribe, and let us know: What do you think is the biggest AI security challenge organizations face today? #AI #Cybersecurity #ArtificialIntelligence #OpenSourceAI #MachineLearning #TechPodcast

Are AI models becoming less reliable? In this episode of Practical AI, Jeff and Peter dive into recent reports suggesting that some newer AI models may hallucinate more than previous generations. They explore whether the industry is facing a training data problem and discuss the growing role of synthetic data in modern model development. Topics include: • Google's TPU strategy and AI infrastructure investments • The growing competition to Nvidia's hardware dominance • The rise of local and open-source AI models • The privacy, security, and reliability advantages of running AI on your own hardware • Why specialized model architectures are changing how AI systems operate The conversation concludes with a prediction: open-source and local AI adoption may accelerate dramatically in the months ahead. 👍 Like, subscribe, and let us know—would you run AI locally if it met your business needs? #AI #LLM #OpenSource #MachineLearning #GoogleTPU #enterpriseai Spotify: https://open.spotify.com/episode/0ghpKA3yLu7jPiWTggsRVC?si=nRZefpvtTOCfR1WvGLbzRg

Are AI companies spending too much on tokens? In this episode of Practical AI, Jeff and Peter explore the economics behind today's AI boom, beginning with a surprising report that SpaceX is renting data center capacity to Anthropic and Google. They discuss: The implications of the SpaceX-Anthropic-Google infrastructure deal Whether reported financial figures tell the full story Sam Altman's comments about companies rapidly exhausting AI budgets The growing challenge of AI token costs Peter's "token diet" approach for reducing unnecessary token consumption Why smaller, specialized models may outperform larger models in many business scenarios DeepSeek's architecture and what it teaches us about model efficiency How developers and businesses can run capable AI models locally As AI adoption scales, efficiency is becoming just as important as capability. If you're building with AI, managing AI costs, or planning your organization's AI strategy, this episode offers practical insights you can apply immediately. 👍 Like, subscribe, and let us know: Are you tracking your AI token spend? #AI #MachineLearning #LLM #Anthropic #OpenAI #techpodcast Listen on Spotify: https://spotifycreators-web.app.link/e/mtrVrNqeS3b Same Altman admits Token problem https://www.tomshardware.com/tech-industry/artificial-intelligence/openai-ceo-sam-altman-admits-ai-token-costs-are-becoming-a-huge-issue-company-seeks-improved-value-as-overspending-becomes-a-meme Space X Rents to Google https://finance.yahoo.com/sectors/technology/article/spacex-to-rent-ai-capacity-to-google-for-920-million-per-month-200639544.html

What happens when you can no longer tell what’s real online? In this episode of Practical AI, Jeff and Peter explore the rapidly evolving landscape of AI—and the growing challenges that come with it. They discuss: The rise of AI-generated content that’s nearly indistinguishable from reality Why digital watermarking may become critical for trust and verification Anthropic’s new Claude Bot and the future of AI in messaging platforms A recent security breach involving the LiteLLM Python package The importance of securing open-source tools in an AI-driven world The episode wraps with a forward-looking discussion on how hardware and infrastructure could play a role in improving AI security. If you're building with AI or just trying to understand where things are headed, this episode connects the dots between innovation and risk. 👍 Like, subscribe, and share your thoughts in the comments. #AI #Cybersecurity #MachineLearning #TechPodcast #Anthropic Links: Artificial Videos Detectable? https://www.youtube.com/watch?v=ICTlfQJpS0g “Claude Bot” from Anthropic competes with Open Claw https://www.youtube.com/watch?v=wfeiCZK0mNs Listen on Spotify: https://open.spotify.com/episode/1FTDkz1NeE8zorHofeSpwF?si=9wnIWEcZTWy9vjj1-gYJGA

What happens when AI starts making real business decisions? In Part 1 of this Practical AI conversation, Jeff and Peter are joined by David Paparelli, CEO of M-Cultivo, to explore how agentic AI is being applied in the coffee supply chain—from pre-export financing to digital auction platforms. They discuss: - How AI is reshaping traditional business roles - The importance of maintaining human relationships in global trade - Real-world applications of AI in agriculture and finance - The balance between efficiency and trust This episode highlights a critical shift: AI isn’t just assisting—it’s becoming embedded in core operations. 👉 In Part 2, the conversation goes deeper into: - The risks of AI-driven financial decisions - Accountability in AI-powered systems - Why companies need “system thinkers,” not just task executors - How prompt engineering and validation reduce costly mistakes If you're building or implementing AI in business, this is a must-watch. 👍 Like, subscribe, and stay tuned for Part 2. Listen on Spotify: https://open.spotify.com/episode/7Fnp6E2CxgudgxaEdrHynn?si=9FVsHjs5QAyBazIX3ds0FA For more information about M-Cultivo go to: https://mcultivo.com/ #AI #BusinessTransformation #AgenticAI #CoffeeIndustry #TechPodcast