Walking? Why don’t you use an exoskeleton for that?!
Michael Coutlakis
Contact pageHas anyone noticed how some people have only recently reached the infatuation stage with respect to AI/LLMs? Many of us have been interacting with LLMs for over three and a half years to engage in discussions at any level or topic. We already know what AI is capable of. So we find it somewhat bemusing when these latecomers, with their newfound knowledge, naively suggest “have you tried using AI for that”? Or “why don’t you ask AI”?
This suggestion to defer to AI might seem harmless and well intended at first. But I don’t think it actually is – and as AI becomes more prevalent, the reasons will become increasingly important. So I want to unpack it very clearly.
Firstly, the suggestion is extremely patronising / condescending at multiple levels. This is especially the case when suggested early in a conversation — for example, after hearing about the existence of the problem, but without interest in any details, plans or actions.
- It suggests I wasn’t smart enough to think to use AI – even though consulting AI has become a common first step in problem solving. Ask AI? What a good idea, why didn’t I think of that myself?!
- It suggests lack of trust in my capabilities and conveys the implicit assumption that the AI output would be better, faster or preferable in some way.
- It conveys discredit for my own agency and decision making — these, if trusted, would involve delegation to AI as deemed appropriate.
Imagine if someone asked me why don’t I use an exoskeleton to walk. (Referring to the augmentative / enhancement type exoskeletons – for example, the recreational types now being produced, or something in Iron Man or Avatar).
Assuming I’m able bodied (or even, perhaps not), I’d probably think why? What’s wrong with my walking? Ok, sure, maybe an exoskeleton would help me walk faster, but do I really need that? Won’t my muscles atrophy if I start using an exoskeleton? Also, I like walking — why would I want to use an exoskeleton? Do I really need the extra dependency?
In any case, I think I will start responding to the question of “why don’t you use AI” with “why don’t you use an exoskeleton to walk?”. Fair warning.
Secondly, there is the question of abdication to AI. At what point, or to what extent, do you hand over the reigns to AI? I don’t think I’ve understood the intricacies enough to give a good analysis. But I think there are psychological and identity implications. If all I do is palm everything off to the AI, what value am I adding? Humans are meant to be creative. If we let that part of ourselves die because AI handles everything for us then that’s just sad.
Thirdly, I think it is important to maintain a healthy balance between getting things done / being productive and learning / staying competitive. Suppose I have a debugging task. I could ask AI to just fix it. Even if the fix seemingly worked, I’d still want to understand why it was broken and how it got fixed to be sure. I prefer using AI as a tool rather than a solution. It can provide potential debugging techniques and relevant information, but I still want to “do” the process myself. I think that gives the best balance of learning, understanding the process and keeping the AI on track so it doesn’t go sloppy. I’m not here to merely act as an AI intermediary.
Agentic AI is where things went downhill a bit for me. I think it works ok in some cases where you don’t really care about the implementation and just need some specs to be met with passing tests.
But I feel like I’ve got my best results from AI picking it’s brain for options, preferably after thinking it through first and starting with a few proposals, then thinking through the options and working from there, rather than hitting the “I’m feeling lucky do everything” button.
There’s also the problem that a certain level of knowledge and expertise are required to set AI the right tasks, ask the good questions and interpret the output. In the early days of AI I wondered if there was still any point learning software, languages, maths and so on if the AI could do it all. Then I started talking to the AI about signal processing and it spoke maths back to me. If I didn’t know enough, I simply wouldn’t be able to interact with the AI at an appropriate level.
Finally, I think AI technical debt, option overload and AI burnout are realities. Hopefully we will recognise this sooner since we’ve already been burned by the social media era. Sometimes less is better.
So back to the original question – have I tried using AI for that? Most likely it probably already was part of the process (thanks for checking I hadn’t forgotten about it) but maybe just not in the way you’d expect.
AI was not used to either write or proof read this document