I’m a firm believer that a good strategy shouldn’t just help you know what to do. It should also help you know what NOT to do. This is because it needs to be able to answer WHY we will or won’t do something during its time in existence, particularly if it’s to be used by people who are more operationally focused.
It sort of reminds me of what Do Not Pay set out to do in terms of having something work on your behalf to wade through the red tape. I believe it started with automating the process of challenging parking tickets. I mean, that’s an interesting one because if you were a cynic, the org issuing the tickets might not be hugely invested in making it easy to challenge parking fines.
Is this automated navigation of beauracracy a legit tactic? If it does become a thing, I wonder if there’ll be any push back to avoid AI from making claims on behalf of people? We already see the use of CAPTCHA to thwart malicious use of automation, but could we see it from organisations that really don’t want to be automatically bombarded with complaints, refund requests etc. etc.?
Hmm, what about small language models that are focused on navigating specific services? An automated system navigator of your own? I do wonder how that would work in practice given the amount of tacit knowledge required to navigate some services/organisations that’s earned through trial and error. A testiment to how much stuff resides inside of people’s heads and sometimes nowhere else.
Massive sympathies with anyone battling with the Crowdstrike issues today. This is absolutely the sort of issue you don’t want appearing on a Friday morning just in time to derail all your personal plans for the weekend (and possibly the days and weeks ahead).
A couple of weeks ago I had a really fascinating experience of conversing with a lady at the train station via her phone, and not in the conventional sense. She spoke no English, so she would scribble kanji (or similar) characters on her phone which would then translate them into English. I then wrote an english response which then converted it back into kanji. I know this isn’t anything that new or exciting, I’ve been using Google Lense for a few years to help me translate the kids Welsh homework. But pretty interesting to use it out in the real world. And it didn’t work perfectly, there was a lot of rephrasing thing because the translation wasn’t quite working as intended. I presume that on the fly voice translation is probably a thing now, with I presume the same unexpected/hysterical inaccuracies. We’re probably not too far off from a (bad) version of the Star Trek universal translator.
I finished reading Hello World by Hannah Fry. Might set a new personal record from point of starting a book and finishing it (5 years? Not continuous reading time I stress). Whilst I’m sure a lot of the examples of algorhithms and AI have moved on massively in recent years, there’s some great ethical musings that I think will remain evergreen regardless of the sophistication of the tech. It’s also pretty balanced in terms of both the risk and rewards of automation. Worth a read.
