
AI Industry Insider: We Ignored The Most Dangerous Part Of AI For YEARS
Notas do episódio
What happens when AI becomes capable enough to solve extraordinary problems, but still unreliable enough to confidently get simple ones wrong?In this conversation, Fernando Lucini explains why the real AI revolution is not just about the models themselves, but the systems, tools, and “harnesses” being built around them. Drawing on decades across Autonomy, HP, Accenture, and Faculty, he breaks down what AI can genuinely do today, where it still fails, and why companies need to understand both the hype and the truth at the same time.But this isn’t just a story about better models.It’s about a deeper shift:🔹 What happens when AI becomes more capable faster than it becomes reliable?🔹 Could open-source models and the software built around them reshape where the real value in AI gets created?🔹 And if powerful systems still hallucinate, forget context, and behave unpredictably, how much autonomy should we actually give them? At the centre of the discussion is a harder question:If AI can solve problems humans cannot while still failing at tasks we consider basic, how do we build businesses and systems that capture its upside without trusting it more than we should?Because the future of AI may depend less on creating one perfect model, and more on learning how to harness imperfect ones.