Generative AI has enormous promise, but there are serious issues that need resolving before AI can be widely trusted and adopted by enterprises and the general public. Let’s explore some these issues and how to solve them.
Outline:
- AI Is Inherently Probabilistic: Developers are accustomed to software systems that are (mostly…) deterministic. It’s a new world where the output you get changes for the same input! How can developers ensure quality and correctness in this situation?
- Alignment: Do AI models and applications behave correctly for our purposes? What about hallucinations, hate speech, and bias? This is the single most significant area of concern and the hardest to address. How do we use AI safely, effectively, today?
- Regulations, Policy, and Total Cost of Ownership: I’ll discuss these issues briefly. Finally, I’ll offer some thoughts about where generative AI will be in five years.