I believe knowledge is what you know based on facts and experience; wind sensors could gather data and store it in a database without a human touching it beyond initial setup. With enough data, and basic information about where the sensors are located, the computer becomes very knowledgeable about wind in a region without human intervention.
I believe intelligence goes beyond that: knowing that such a system is a solution to an observed problem, architecting said system, using the output to solve a problem, analyzing the results, and deciding where to deploy additional systems.
I think both examples above can be done by AI (if not now, then soon)—but only after being prompted carefully by a human. However, a generalized AI that can do all of the above for any problem in the known universe is likely very far off.
if knowledge is a justified true belief, i’m down for saying LLMs have beliefs. to the extent that they are incorrigible, their faith may actually be superhuman.
I believe intelligence goes beyond that: knowing that such a system is a solution to an observed problem, architecting said system, using the output to solve a problem, analyzing the results, and deciding where to deploy additional systems.
I think both examples above can be done by AI (if not now, then soon)—but only after being prompted carefully by a human. However, a generalized AI that can do all of the above for any problem in the known universe is likely very far off.