I think that a large reason why the discourse sucks is because the public is seeing hardly any discussion using language that is specific about types and features.
Let me explain. With older technologies, the language to talk about it is well developed. I like to use nautical terminology as an example that most people are somewhat familiar with. There are a lot of terms like jib, sheets, mainsail, schooner, keel, galley etc, and a lot of people don't know what those mean. But it is pretty easy to recognize that there is a whole terminology to describe ships and the features thereof which is used by experts and which can be very specific. If one guy says the boat won't make the trip, and it's because the keel goes more than two fathoms deep, and the other guy says that of course it can because the galleon can go bireme past five hundred knots, even landlubbers will be able to figure out who knows their stuff.
But in the current AI discourse, AI is AI is AI. Agentic LLMs? AI. Non-agentic LLMs? Also AI. Diffusion Models? Al. Search engine? AI, kinda. R2D2? AI. Autocomplete? Sure, why not, AI. It's like if sailors used language that barely distinguished between specific nautical technologies. Boat is Boat is Boat.
Now, nautical terms have developed over thousands of years, and AI (whichever type you mean) is a new technology that is not fully developed. But imagine a discourse where Boat is Boat is Boat. The bosses learn that we just got Boat working, and are making plans to conquer the New World. Other people are concerned about the moral implications of conquering the New World. Meanwhile, one of the sailors tasked with this tried using Boat over the weekend fishing with a friend, and he's not sure he can paddle it that far. Another guy tell him not to worry, that there's a new model of Boat and it doesn't even need a paddle. He says at the current rate of advancement, we should have a new Boat capable of going underwater in less than ten years! We thought of coming up with a name for this new type, but it's basically just another kind of Boat.
If we want better talk about AI, we should use better language about AI. We are living in the time where the language used to talk about these things will be developed. We might find there is more agreement when we are talking about the same things.
Let me explain. With older technologies, the language to talk about it is well developed. I like to use nautical terminology as an example that most people are somewhat familiar with. There are a lot of terms like jib, sheets, mainsail, schooner, keel, galley etc, and a lot of people don't know what those mean. But it is pretty easy to recognize that there is a whole terminology to describe ships and the features thereof which is used by experts and which can be very specific. If one guy says the boat won't make the trip, and it's because the keel goes more than two fathoms deep, and the other guy says that of course it can because the galleon can go bireme past five hundred knots, even landlubbers will be able to figure out who knows their stuff.
But in the current AI discourse, AI is AI is AI. Agentic LLMs? AI. Non-agentic LLMs? Also AI. Diffusion Models? Al. Search engine? AI, kinda. R2D2? AI. Autocomplete? Sure, why not, AI. It's like if sailors used language that barely distinguished between specific nautical technologies. Boat is Boat is Boat.
Now, nautical terms have developed over thousands of years, and AI (whichever type you mean) is a new technology that is not fully developed. But imagine a discourse where Boat is Boat is Boat. The bosses learn that we just got Boat working, and are making plans to conquer the New World. Other people are concerned about the moral implications of conquering the New World. Meanwhile, one of the sailors tasked with this tried using Boat over the weekend fishing with a friend, and he's not sure he can paddle it that far. Another guy tell him not to worry, that there's a new model of Boat and it doesn't even need a paddle. He says at the current rate of advancement, we should have a new Boat capable of going underwater in less than ten years! We thought of coming up with a name for this new type, but it's basically just another kind of Boat.
If we want better talk about AI, we should use better language about AI. We are living in the time where the language used to talk about these things will be developed. We might find there is more agreement when we are talking about the same things.