Motorcycles have been around for 100 years and still have about a 25x fatality rate compared to automobiles per mile traveled.
While this may be reduced one day, I expect traveling at 80 mph even on an auto-stabilized device will be more hazardous than doing the same in a big box.
Similarly, I can imagine if the possibility that“going fast” with AI will be higher variance, with more wins and more crashes, as an intrinsic quality of “thinking at speed”.
The metaphor is compelling, especially when thinking about AI in education. Indeed, attention and "learning to maintain dynamically stabilized control over your own cognitive activity" are appropriate meta-expectations or goals for education and educators. But technology, whether it be in the form a device, a tool, or an AI/Motorcycle, arrive at the school doorstep uninvited, promising change and growth (of the meta-expectation), well ahead the institutional, cultural, social, or cognitive abilities necessary to use them effectively and safely (on the whole). They appear suddenly, and promise to make something old and tedious either obsolete or more efficient, faster, and readily available. Almost always, there is misapprehension: of what the technology can do, or more more likely, what the humans should be ready for. This is a maddening reality in education, and one that has produced shorter and more jarring cycles of tech. updates and adoptions without really providing the wet-ware update necessary for the kids to handle them or the system to incorporate them.
Your own story about recognizing, experiencing, and controlling your own attention as a student reveal a virtue all schools hope students discover, even inadvertently. This is done at the speed of cognitive development, maturation, and the emergence of a self. It is slow at the time, and provides the exponential, accelerating curve towards smarts. In this light, AI safety (in education anyway) might only be available by prohibiting it.
Pre AI technical work required a lot of writing to think through problems. Post AI, technical work requires a lot of writing to communicate with an agent. These are quite different experiences. Chatting with an agent is about shoving out as many details as possible whereas thinking through problems is bathing in the details until insight appears. This latter form of writing still has a lot of value, but if you do too much of it you’ll move to slow. Learning to ride your AI enhanced brain bike requires you to learn this balance of diving into and shoving details away in order to maintain control of the whole system.
Motorcycles have been around for 100 years and still have about a 25x fatality rate compared to automobiles per mile traveled.
While this may be reduced one day, I expect traveling at 80 mph even on an auto-stabilized device will be more hazardous than doing the same in a big box.
Similarly, I can imagine if the possibility that“going fast” with AI will be higher variance, with more wins and more crashes, as an intrinsic quality of “thinking at speed”.
Definitely the case. That’s partly why I used that metaphor. No risk, no return.
The metaphor is compelling, especially when thinking about AI in education. Indeed, attention and "learning to maintain dynamically stabilized control over your own cognitive activity" are appropriate meta-expectations or goals for education and educators. But technology, whether it be in the form a device, a tool, or an AI/Motorcycle, arrive at the school doorstep uninvited, promising change and growth (of the meta-expectation), well ahead the institutional, cultural, social, or cognitive abilities necessary to use them effectively and safely (on the whole). They appear suddenly, and promise to make something old and tedious either obsolete or more efficient, faster, and readily available. Almost always, there is misapprehension: of what the technology can do, or more more likely, what the humans should be ready for. This is a maddening reality in education, and one that has produced shorter and more jarring cycles of tech. updates and adoptions without really providing the wet-ware update necessary for the kids to handle them or the system to incorporate them.
Your own story about recognizing, experiencing, and controlling your own attention as a student reveal a virtue all schools hope students discover, even inadvertently. This is done at the speed of cognitive development, maturation, and the emergence of a self. It is slow at the time, and provides the exponential, accelerating curve towards smarts. In this light, AI safety (in education anyway) might only be available by prohibiting it.
Pre AI technical work required a lot of writing to think through problems. Post AI, technical work requires a lot of writing to communicate with an agent. These are quite different experiences. Chatting with an agent is about shoving out as many details as possible whereas thinking through problems is bathing in the details until insight appears. This latter form of writing still has a lot of value, but if you do too much of it you’ll move to slow. Learning to ride your AI enhanced brain bike requires you to learn this balance of diving into and shoving details away in order to maintain control of the whole system.
If balance in brain bicycles is an issue now, just wait until bci becomes a Big Thing
Nice post