Ninan: “As India struggles to build its manufacturing sector and quality physical infra, we could consider what leadership role engineers should play in politics and the administration.”
NYTimes: “The scientific Nobels announced [recently] — in Physiology or Medicine, Physics and Chemistry — honored achievements rooted in fundamental research from decades ago.”
John West: “If you peer into the mind of a model, what you find won’t be recognizably human; it’s really a thicket of statistics, producing words by splitting language into long sequences of vectors. You can think of a vector as a point on a graph, which is identified by two numbers, one plotted horizontally (on the X axis) and the other vertically (on the Y axis). Instead of two dimensions, an LLM is turning words into vectors with many hundreds of dimensions—more than it’s possible to visualize. Plotting words in a “vector space” makes it possible for an LLM to detect the connections among them: Distance is an easily computable property in a vector space, and closeness encapsulates relationships. With enough data, an LLM can learn these relationships, so that given any word, it can predict what the next word should be, and the one after that. This predictive ability is what allows chatbots to carry out fluent conversations with human beings.”
Thomas Kurian on what is the most difficult thing to do when it comes to building AI for enterprise: “There are three elements. First, you need to have a model that’s world class in not hallucinating, and is using the right tools to do the right job…The quality of the model is really important. Second is just giving you a model by itself is not useful. If you’re building an agent — let’s say you’re doing financial analysis for your CFO — that agent needs to look up financial results from your ERP system. It needs to look at contracts you have in your Microsoft Office system. So it needs to be able to reason across all of that enterprise context. But in doing so, it should maintain your security, meaning it shouldn’t say, ‘Well, now that I have all this data, I’m going to publish it to the whole company.’ So it needs to manage your security and permissions the right way. And it needs to be able to understand the information as it changes, because in enterprises, information changes all the time. Third, when you build agents, you need to have a central place to govern and manage. Because organizations are worried about a million agents happening in their company and not having a central place to manage what permissions, what tasks they’re given permissions to do, etc.”
