Studies have found that children can ask up to 107 questions per hour.
As a parent, I have first hand-experience of this.
While researching this post, I found a monograph on children’s questions and cognitive development. The below quote stuck out:
“Together, the results of these four studies support the existence of the IRM (Information Requesting Mechanism) as a way for children to learn about the world. Children ask information-seeking questions that are related in topic and structure to their cognitive development. Parents give answers to these questions, but when they do not, the children persist in asking for the information, suggesting that the goal of this behavior is to recruit needed information. The content of these questions shifts within exchanges and over the course of development in ways that reflect concept building. Finally, children generate questions efficiently in order to gather needed information, and then are able to use this information productively; they tap into their existing conceptual knowledge in order to do this.”
In the past, I have attempted to answer these questions as best as I can, while also acknowledging that some (like, why is water wet?) were far beyond my capabilities. In those cases, I’d have to think about it, give an explanation of why I don’t know, and then do my best to find out as soon as possible.
Sometime, somewhere, that information might have been taught to me in school. But as I grew older, it’s probably been replaced by knowledge about heat pump versus condenser tumble dryers, or which variety of coffee bean has the highest caffiene content.
So, I’ve decided that in order to stress-test and explore my wearable AI, while sticking to my rules (Rule #3, ask as many questions as you can), I will try to be childlike in my inquisitiveness, and ask some difficult childlike questions about the world, things that I often take for granted or treat as uninteresting or unnecessary. I also decided to ask these questions at inopportune, out-and-about moments.
Interim results are positive; the glasses have been enlightening on where pencils come from, why chickens lay eggs, and whether four-leaf clovers really exist (they do).
That is great – useful, and interesting.
If we are asking, “can wearable AI help us learn in a wider range of spaces and contexts” then for factual information or explanations, I think the answer is probably “yes”.
On the other hand, if we are asking “can wearable AI help us to think” then the answer is less clear.
Factual recall is something that AI models are somewhat good at (although they might occasionally create total fiction and there’s no way of knowing when). In the cases above the AI gave me pretty good, factually correct answers.
So I received the facts that I needed, in a pleasant, patient, if somewhat slightly robotic way, and I accessed these answers hands-free, while out in the environment and not sitting at a screen. This is promising.
On the other hand, if I had tried to find the answers to these questions myself, then the journey would have been slightly different. I would have had to at the very least, open my phone, search for the information, filter different sources, perhaps stumble on some incidentally relevant information, and potentially evaluate several different answers. Who knows what incidental learning I deprived myself of by deferring straight to the AI answer?
In short, the wearable AI gave me the right answers, but shortcutted the journey of getting them.
I would argue that this shortcutting is even more pronounced when using a wearable model that is ironically right next to your brain, because it is there, ready to answer, with minimal friction, at all times. It requires no physical activation, and this is a profoundly different experience to typing into a user interface, or even dictating to a handheld object.
Thinking back to kids’ inquistitiveness, when children ask questions, I also don’t think it’s just about seeking a totally factual, sensible and rational answer. Yes, it is important to know these things but it’s also important to think around what a subject is, and why it matters.
More than that, it’s a chance to stretch an idea until it breaks, instead of resolving the answer matter-of-factly and calling it a day. This tells me that there are big questions about whether AI supports ‘thinking’ or erodes it.
Accessing the AI for immediate answers also collapses the space and time between identifying a knowledge gap and closing that gap. In the past, that space was larger, giving more time to think, speculate, or hypothesize. When that space is instantly accessible via speech, at any time, without the physical need to access a device, the gap narrows even further.
When I ask these questions, I am out – in the physical world. I am not multi-tabbing. This is an important point, because it means I am more likely to accept the information rather than triangulate it with other sources. If the information is uncertain, I might be more likely (at a laptop) to open another tab and conduct a separate search. If I am walking, or have my hands full, or I’m chopping a watermelon and suddenly want to know how long a watermelon takes to grow, then I am probably going to take it at its word.
Taking this further, if I find the answer on the internet, I can return to that source and evaluate it, perhaps understand why the information has been presented in a certain way, or consider where the source came from itself. With the AI, the answer is generated instantly on the fly, and we have no information on where it came from really, other than ‘training data’. On the computer interface, newer models can point to links and source material – but wearable AI cannot.
This is one way that I can already tell that wearing AI might change the way that we access information.
We need to consider not just the answers that wearable AI provides us with, but the temporal and contextual nature of when we ask questions and what effects that might have.