Why Sam Altman and Elon Musk are WRONG and Ontology is the next step in AI

Summary

Michael, a developer and AI enthusiast with a philosophy degree, discusses the misconceptions surrounding AI. He argues that current AI is narrow and lacks true understanding, leading to misguided paths. Michael suggests that instead of scaling AI from a basic level, we should start with a clear ontology, defining what things are and their relationships, to create more efficient and accurate AI solutions. He believes this approach will reduce energy usage and prevent AI from making unintuitive decisions.

Takeaways

• 🎓 Michael, with a background in B2B development, Caltech AI studies, and Philosophy, aims to address misconceptions in AI by emphasizing the importance of understanding how to think about thinking.
• 🤖 The current AI conversation is misguided by focusing on scaling AI to super-intelligence without considering the limitations of narrow AI capabilities.
• 🚀 AI's narrowness is highlighted by its ability to perform specific tasks but lack of true understanding or meaningful thinking, exemplified by issues with LLMs like GPT.
• 🔋 Energy usage in AI is a significant problem due to the vast amount of power required to run complex calculations and train AI at higher levels.
• 🚨 AI's lack of understanding of 'what things are' leads to failures, such as Tesla's Vision misunderstanding human driving relying solely on vision.
• 🛠️ Michael suggests a pivot to a software system that knows 'what things are' and adjudicates tasks based on defined relationships and logic, rather than intuition alone.
• 📚 The concept of 'ontology' in computer science is introduced as a solution, which provides a framework for defining things and their relationships within an AI system.
• 🔗 Organizing data around defined 'things' and their properties is essential for creating a structured approach to AI problem-solving, reducing the need for excessive power consumption.
• 🌐 Michael points to 'Palet' as a sleeping giant in AI, potentially serving as a data and AI orchestration tool that could revolutionize enterprise capabilities.
• 💡 AI should be seen as a piece of the solution within a broader context, not as the sole solution, and it requires specific knowledge and understanding of logical relationships to be effective.

Q & A

What is Michael's background in relation to AI?

Michael has a background as a developer working in B2B applications and has studied AI through Caltech. He also has a degree in Philosophy from a university, which he believes is important for understanding AI.

Why does Michael think philosophy is important in AI conversations?

Michael believes philosophy is important because AI involves computers doing more 'thinking,' and philosophy is essentially about thinking about thinking, something he feels is lacking in many AI professionals.

What misconception does Michael identify about AI?

Michael identifies the misconception that AI is being approached from the idea of super-intelligence, starting with AI and trying to scale it up, rather than starting with a clear understanding of what things are and then applying AI.

How does Michael describe the limitations of AI?

Michael describes AI as very narrow in what it can do and think. It can perform incredibly specific tasks but does not mirror human thinking in any meaningful sense.

What does Michael see as a problem with the current approach to AI?

Michael sees a problem with the current approach to AI because it relies on back intuition and requires an enormous amount of energy usage, which is not sustainable.

Why does Michael argue that AI's understanding is misguided in terms of Tesla Vision?

Michael argues that AI's understanding is misguided in Tesla Vision because it misunderstands how humans drive. AI focuses on vision alone, but human driving involves understanding what things are and their meanings in society.

What does Michael propose as a solution to the current AI approach?

Michael proposes using a software system that knows what things are and then adjudicates tasks based on those definitions, rather than relying on AI to intuit information.

What is the role of ontology in Michael's proposed AI solution?

Ontology, in Michael's view, provides the necessary guardrails for AI by defining what things are and their relationships, allowing AI to leverage data in a clean, logical way.

How does Michael relate ontology to object-oriented programming?

Michael relates ontology to object-oriented programming by suggesting that just as object-oriented programming organizes code, ontology organizes data, defining both the individual pieces and their relationships.

What does Michael see as the advantage of using ontology in AI?

Michael sees the advantage of using ontology in AI as providing a clear understanding and logical structure that amplifies AI's intuitive capabilities, leading to more specific and nuanced work.

Why does Michael consider Paler to be the 'sleeping giant' of AI?

Michael considers Paler to be the 'sleeping giant' of AI because it serves as a data and AI orchestration tool that can create cross-capabilities across an entire enterprise when integrated with an ontology.