I had to laugh so hard when I imagined the pigeon sitting in the catapult! That is a really good metaphor.
I agree that observing may become a full-time job, and we should choose wisely how we do it, and what we delegate.
The "scrub nurse" does super valuable observing, but he/she also acts. AIs cannot (yet?) play that role. But what they can do pretty well is the observing part. Certainly never only AI, and never only one. But if we do the observing ourselves, unaided, it IS a full time job, and we'll be stuck in observing 24/7.
AIs have a lot of useful capabilities here: They have pretty much unlimited patience and attention, can follow rules (if we are disciplined to set the right rules), and they have no ego or intention. They do have biases and limitations, of course, but these are discoverable, even if only with rigorous and regular testing. A human may consciously or unconsciously hide them from us. In that sense, AIs are more stone than pigeon. AIs are also incredibly good, if prompted right, at pointing out the things we have overlooked.
In my view, this is not "delegating the observing", but widening our perspective. A periscope, telescope or microscope all have limitations. They help to see some things, and make others invisible. If used right, and if you do not take what you see through them as the complete truth, they can still be immensely valuable. The human still needs to learn how to judge well, how to orient themselves with this input so that they can then decide and act (OODA).
As you say: "AI within limits". The pure observing is complicated, while orientation is where the complexity sits?
Hello Andreas. I'm still sitting with the idea of the observer and the challenges we get if the observer moves into the orientation space. I think there is a useful tension here, handled properly. I'm not sure that observation is complicated as much as complex, because it changes as we look at it. If I go back into my history, we had fighter controllers who looked at the big space and directed the interceptor aircraft, operating in pairs: one on task and the other keeping the local environment under observation. I don't think there is any single hard rule, but I think the discipline of thinking about the nature of observation and its context may serve us. Yes, I love the image of the pigeon. If I was a cartoonist, I could have fun with that. I have empathy with that pigeon 😊
My thinking when writing the comment was this: Most definitely, being an observer is complex. But where does the complexity come from? I would argue that I cannot purely observe at all. Already, what I perceive and what I ignore, what I attend to versus not, is a reflection of my inner complexity. I would call that orientation, because our habits, the heuristics we use, etc., are actually a level of orientation, not of observation. The way you describe it in the fighter space reflects I think the triangulation that is required: Each of the people in play will have their biases, their limitations. They will have distinct focus areas, but none of them will observe perfectly even in that area. They all have that subconscious level of orientation, which they cannot prevent, and only on top of that the conscious trained orientation ability that allows them to give useful input to the others.
AI can observe patterns very well. And it will always, if we do not prevent it, deliver a semblance of orientation to go with that. That is very difficult for us. Where we can limit the orientation the AIs deliver, and instead can use them as one additional perspective, or even multiple perspectives, to observe from, is where I think they can be helpful. An AI may be incapable of "pure unbiased observation" as humans are, or more so. But you get additional points to triangulate what the result of pure observation would be - and then do your orientation more consciously from that improved baseline.
I definitely agree that there is no one hard rule. Observation and orientation are concepts, and as such they are already one step removed from reality. We are inside the system, and likely not capable of separating them cleanly, I guess.
As ever, the real work here is done in conversation, not a model. Model is just a prompt. I'm thinking that the fighter controller metaphor may have some mileage, so I'm going to spend some time thinking about that. In working with clients, I have always found it better to be working in triads rather than dyads. That sits with the fighter controller metaphor. The challenge, I think, particularly when we are consultants with a proposition for a client, is more often about defined inputs and outputs with time scales. It is less about the curiosity of finding out what we really need to talk about. Right now, it's still very much an exploration phase. I'm going to play with this a little bit more. Then there are a few of us, I think, that it might be worth getting together to talk about this.
I indeed think the triad vs. dyad point is strong. Very much worth exploring, if you ask me. Happy to co-explore that anytime, just let me know when the time comes...
I had to laugh so hard when I imagined the pigeon sitting in the catapult! That is a really good metaphor.
I agree that observing may become a full-time job, and we should choose wisely how we do it, and what we delegate.
The "scrub nurse" does super valuable observing, but he/she also acts. AIs cannot (yet?) play that role. But what they can do pretty well is the observing part. Certainly never only AI, and never only one. But if we do the observing ourselves, unaided, it IS a full time job, and we'll be stuck in observing 24/7.
AIs have a lot of useful capabilities here: They have pretty much unlimited patience and attention, can follow rules (if we are disciplined to set the right rules), and they have no ego or intention. They do have biases and limitations, of course, but these are discoverable, even if only with rigorous and regular testing. A human may consciously or unconsciously hide them from us. In that sense, AIs are more stone than pigeon. AIs are also incredibly good, if prompted right, at pointing out the things we have overlooked.
In my view, this is not "delegating the observing", but widening our perspective. A periscope, telescope or microscope all have limitations. They help to see some things, and make others invisible. If used right, and if you do not take what you see through them as the complete truth, they can still be immensely valuable. The human still needs to learn how to judge well, how to orient themselves with this input so that they can then decide and act (OODA).
As you say: "AI within limits". The pure observing is complicated, while orientation is where the complexity sits?
Hello Andreas. I'm still sitting with the idea of the observer and the challenges we get if the observer moves into the orientation space. I think there is a useful tension here, handled properly. I'm not sure that observation is complicated as much as complex, because it changes as we look at it. If I go back into my history, we had fighter controllers who looked at the big space and directed the interceptor aircraft, operating in pairs: one on task and the other keeping the local environment under observation. I don't think there is any single hard rule, but I think the discipline of thinking about the nature of observation and its context may serve us. Yes, I love the image of the pigeon. If I was a cartoonist, I could have fun with that. I have empathy with that pigeon 😊
My thinking when writing the comment was this: Most definitely, being an observer is complex. But where does the complexity come from? I would argue that I cannot purely observe at all. Already, what I perceive and what I ignore, what I attend to versus not, is a reflection of my inner complexity. I would call that orientation, because our habits, the heuristics we use, etc., are actually a level of orientation, not of observation. The way you describe it in the fighter space reflects I think the triangulation that is required: Each of the people in play will have their biases, their limitations. They will have distinct focus areas, but none of them will observe perfectly even in that area. They all have that subconscious level of orientation, which they cannot prevent, and only on top of that the conscious trained orientation ability that allows them to give useful input to the others.
AI can observe patterns very well. And it will always, if we do not prevent it, deliver a semblance of orientation to go with that. That is very difficult for us. Where we can limit the orientation the AIs deliver, and instead can use them as one additional perspective, or even multiple perspectives, to observe from, is where I think they can be helpful. An AI may be incapable of "pure unbiased observation" as humans are, or more so. But you get additional points to triangulate what the result of pure observation would be - and then do your orientation more consciously from that improved baseline.
I definitely agree that there is no one hard rule. Observation and orientation are concepts, and as such they are already one step removed from reality. We are inside the system, and likely not capable of separating them cleanly, I guess.
As ever, the real work here is done in conversation, not a model. Model is just a prompt. I'm thinking that the fighter controller metaphor may have some mileage, so I'm going to spend some time thinking about that. In working with clients, I have always found it better to be working in triads rather than dyads. That sits with the fighter controller metaphor. The challenge, I think, particularly when we are consultants with a proposition for a client, is more often about defined inputs and outputs with time scales. It is less about the curiosity of finding out what we really need to talk about. Right now, it's still very much an exploration phase. I'm going to play with this a little bit more. Then there are a few of us, I think, that it might be worth getting together to talk about this.
I indeed think the triad vs. dyad point is strong. Very much worth exploring, if you ask me. Happy to co-explore that anytime, just let me know when the time comes...