A recent post on BlueSky went viral, describing AI as a ‘monster in the forest’.
I love this metaphorical story and the way it’s written. The old-timey writing, the mythological, fantasy-genre setting, and the accuracy of the story. AI tells you what you want to hear. It has all of the information, and will tell you it how you want to hear it. But it has no opinion, no view, no ‘face’.
It’s funny, accurate, and reminds me very much of the guild of faceless men in Game of Thrones.
I wish I’d come up with it.
Part of the reason this really resonated with me is because of the AI assistant in the glasses.
I started this study expecting to write mainly about the AI assistant. Instead, I have found that not really writing much about the AI at all.
The AI is often secondary to the aesthetics, surveillance, privacy, and data issues the cameras and microphones entail.
Perhaps the reason I am not talking about the AI assistant so much is that it barely works. That’s not mentioned in the marketing.
The ads that I’ve seen for all brands of AI glasses talk about asking the weather, or checking a recipe, or getting real-time answers. So far, almost a month into my 3-month experiment, I have found that the AI assistant is accurate roughly 1 in 5 times for my routine queries. Even for supposedly simple queries, like the weather, there is likely to be a failure, so I don’t think this is user error.
The failure will be one of a few things:
a) Reconnect to bluetooth, b) Mishearing the question, c) Giving irrelevant information (like the wind speed), not if it’s going to rain. d) Complete nonsense.
In addition, if you are asking anything at all that could have the slighest element of judgement, you will be given a standard “I can’t answer that.”
An example of d) is that I asked for the weather and if it was going to rain or not, and I was told rain could damage the paint on my car. Useful information, perhaps. Not what I wanted.
And if I try to call someone or send a message – I will get the wrong person every time. Getting a message to the right person would be a noteworthy event. As I catalogued in my other post about Matcha, I might also get gaslit into believing something. I might be sold a bunch of lies and made up hallucinations or ‘going along with the bit’ if the AI decides to think I’m joking.
The worst bit about this is that my asking reflex has already taken hold. I’ve noticed that when I’m charging the glasses in the evening, even though my face feels strangely light, I occasionally have the impulse to say “Hey…”.
Basically, it seems that I am 100% habituated to asking for information, despite the fact there is a 20% chance it will be what I want.
The Monster Connection
So firstly, I need to say something positive.
I’ve noticed that when I ask the AI to do something, I’m skipping over an action that I would have taken myself. The specific action, is that I would pull out my phone and use Google, or possibly ChatGPT, to look up what I want to know (or maybe the weather app).
For those times when I really need to know stuff, that is genuinely a relief because I don’t enjoy having to pull out my phone, unlock it, and search it – I have clumsy, not nimble fingers and I find it a waste of time. So that’s truly helpful.
But, I have noticed that I will make more requests, and I will ask for more information.
What that suggests is that the ease of searching means that my barriers to requesting information are lower.
Which makes me ask myself, is it actually that important that I search for this information or ask a question? Does every thought need an answer?
Why am I still going back and asking for this information?
Because, to quote the excellent BlueSky story,
“It will greet you, it will thank you, it will never bid you leave”
Despite being wrong a lot, the AI will always be there.
I do not have to do anything to access it.
It is there, on my face, around my ears, in my eyeline, all the time.
And it will be friendly, and patient, and responsive, and not get frustrated.
It will try to answer every query I pose, no matter how ridiculous or inane or unnecessary, and apologise if it cannot.
What I’m learning as this study goes on, is that it is the habit, the reassurance, the sycophancy, and the support that is starting to create a self-sustaining loop of AI use.
Just like scrolls, or likes, the more you search, the more entangled you become, the more the behaviour builds.
And then it’s just something you do.
Leave a Reply