Technology Social Media and Updates

Custom AI Needed Until Generalization Boosts Corporate Efficiency

Let's talk AGI and GDP - and a question I have... I have no idea when AGI is coming or even what the precise definition is. But there's one thing I have strong priors around: large companies have incredibly bespoke processes - even within the same firm across regions and business units. The way they do support, R&D, billing, etc. are quite unique. This is why there is so much software customization out there and why users still largely hate using the software. If we don't have the leverage of generalization in AI, how far can we get in automating business processes and impacting corporate performance (revenue and costs)? Right now, it's not obvious that the public companies who aren't directly selling AI are more profitable or growing faster because of AI usage. When and how will that change? One approach is of course heavy customization (Forward Deployed Engineers) or internally-built agents. Is this enough? Can businesses define their processes in sufficient detail for agents to be able to automate them comprehensively? If agents have narrow intelligence, can earnings impact still be broad? It's obvious that AI will create a massive boon for domains with universally verifiable rewards (math, coding, science, etc.) But for boring business processes, how long will it take for us to see the impact?