I have been engaging with the glasses as much as possible, following my rules, and trying to really see if there is something to offer above and beyond just dictating messages to ChatGPT on my phone. The problem is, the more I try to probe for an opinion, or a point to argue with, or something that isn’t just banal, generic fluff, the less enthused I feel.
No matter what questions I ask, I seem to get surface answers. Ok, as a longtime user of GenAI models since the early days, I was expecting this, and I know why it happens – but I have to comment on it.
I try to question if maybe the viewpoints I’m given are biased, or unfair, or representative of certain cultural worldviews (which they obviously would be). I asked which sets of linguistic and social norms, and cultural knowledge my glasses responses represent; the response was “many languages, including English, French, Spanish, Italian, and global cultures from America, the UK, Australia, India, Africa, and various Asian cultures’. I don’t know what this means, and I can’t push for any actual detail.
One of these big questions that occurs in AI and education, is that if the training data is biased towards particular world views (it is), to what extent might that prestructure, constrain, or affect our thinking? Will we be confined to an echo chamber even further than we already are, and is that healthy?
A related problem I find is that I can’t nail down the glasses on any complex issues. Any topic that leans even slightly towards something controversial will be met with ‘there are varying perspectives’, and ‘it’s a complex issue’. The kind of heated debates where you feel like you have met some productive friction in your life are not there. Then again, maybe that’s not the point. But why advertise an AI assistant if you can’t argue with it about politics?
This has led me to consider a point regarding wearable and ambient AI.
If AI becomes ever-more present, ever-more backgrounded in our lives, we must face the hard truths that they are far from objective bodies of knowledge. They are looking for the most acceptable, frictionless arguments, perspectives, and approaches to life. Should we always be defaulting to ‘it’s complicated?’ Does this obscure our ability to take moral and ethical stances on subjects?
The complicating factor is that with wearables, these perspectives become mentally and physically closer to us. If I feel that there is a disembodied AI in my ear, like an angel on my shoulder, telling me that ‘problems are more nuanced than they seem’, am I likely to feel that way myself? Will it cloud my judgement, or values?
My experience so far tells me that, like with other AI models, the intention is to resolve, not provoke. Yet sometimes, provocations are necessary.