Starting out the week seemingly with some sort of vigour, as my brain wants me to start writing week notes again. I know it won’t be consistent, but it could be a good practice – in fact it can definitely be good practice. Writing itself is a form of a phrase I’ve recently rediscovered – “deliberate friction” (or mindful friction as I’ve also seen it.
Slowing down to spend time with one’s thoughts rather than taking them for granted or as bothersome flies.
Although I do still hate on-screen keyboards. Might have to address that.
I’m enjoying the fragmented approach I’m using over at the other 6days blog. I think a similar approach should work here, and keep things mentally consistent. I’ll aim to talk more about being freelance, rather than specific bits of client work, as well as general views and thoughts on the state of the industry.
Putting the “log” back into “blog”, perhaps.
An early coffee realisation: the hype around AI and LLMs has been highly outputting. The narrative of the paradigm has been difficult, amplified repeatedly by the lack of critical discourse, engagement with policy, or any broader social contextualising about the technology and its effects. For a scholar in these things, it’s a fairly depressing time – the distinct lack of opportunity to discuss matters without it becoming polemic instantly has been broadly discouraging.
Another realisation: Discard the hype and think of it all as merely a system for probabilistic processing. This approach feels like something I can get my head around much more usefully. By making probability an up-front part of the agenda, I instantly see how it compares and contrasts with other tools and systems.
For instance, when working with high profile data that will likely be printed out for reference, you do not have the luxury of making mistakes. Outputting figures which are “probably” correct, with the notion that you could fix them or improve them afterwards, is not an option. It’s a blocker to fully applying things like agile, iterative approaches to the data process, because you need to carefully and sensibly define a high threshold from the start.
Or in other words, Maths is very different to Language, which has a high degree of flexibility and interpretation. (Or rather, the interpretation with data stems more from the definition of the data model itself). Don’t use probability for things that Must Be Exact.
“Full stack webdev”, but also something about “Full Hierarchy team member” needs drawing out too in job roles. Obviously, generally any senior position in a small company will need to cover detailed and/or menial tasks regardless. But as a freelancer, how do you make it clearer than you can jump between levels of responsibility and interaction, just as you’d jump between server management and CSS?
The word “hyper” keeps popping into my head the more I come into contact with AI.
Hyperbad. Hypersmart. Not necessarily in a good way.
are.we normalising mansplaining, walls of text, and ignoring gaps in knowledge? Consensus seems to be so.
Hyper-doing-something. The acceleration can go both ways; progression (for some meaning) towards a point of singularity, but also fragmentation and chaos towards an infinity of points. Entropy.
AI is not the same as Understanding.
Or rather, Intelligence is not production. Production is not the same as understanding.
Wow, there are so many ways of just copying data around. Each with their own reasons and pitfalls.