AI blog question challenge
Rishabh P. Sharma posed the AI blog question challenge. Rishabh writes about being an “AI Hater”. It seems strange to me to have such intense feelings about AI. I guess that makes me “neutral” by Rishabh’s measure. I think pragmatist is perhaps more accurate.
How was your first experience with AI models?
I was feeling bored during Christmas 2019 and started learning to code (Python). By December 2022 I was working on an app to automate a staff allocation task. This was an optimisation problem. A problem where you’re trying to find the best combination from a large set of options. To help solve it I used the PuLP module to work with mixed integer linear equations (MILP). Sounds confusing? It did to me. To expedite the development process I used ChatGPT to help with writing the equations. It did very well. Not perfect of course but a massive help. From that point on I have continued to explore the utility of Large Language Models (LLMs) at home and at work.
Do you use AI or are you completely against using it?
I use LLMs for a wide range of tasks - research, coding, learning, initial drafts, search, maths problems, formatting, summarising… LLMs can be a very good assistant to have so long as you give them clear instructions, rein them in, “supervise” and fact check their work. There’s a sweet spot to be found knowing where my judgement ends and the LLMs output can be trusted. This shifts as models improve, outcomes matter, and where my domain expertise lies.
There is a fair bit written about the atrophying of skills through cognitive offloading to LLMs. My personal experience suggests to me that’s true. It’s a trade off. There are some things I need to do that are mundane or which I am not especially interested in. I am willing to outsource my thinking and the effort on those things for the time back and head space for other stuff I am more interested in doing. Conversely I have learned and found out how to do things much quicker by using LLMs that I would have by searching up answers or finding and scanning through the right text books.
Do you have any preference among different models, for example Claude vs ChatGPT? If yes, how do you choose?
I use Perplexity so can choose between half a dozen different models plus a few other features to get things done. I find Claude Sonnet 4.6 handles most things well with or without “thinking” enabled and depending on the task in hand.
What aspect of AI models do you like and what do you not like?
I like that they are not human. By that I mean I can ask as many questions and follow up questions as I need and will always get a response I can use or build on. I read a blog post that made the point, “…when you have an infinite answer machine, your ability to ask good questions becomes infinitely more valuable”. Give it a read. I could not agree more with the proposition. A rinse and repeat cycle of iteration to get output I can use.
However, LLMs are not all equal. At work we are restricted to using MS Co-Pilot. MS have a remarkable ability to degrade pretty much everything they are involved with. Co-Pilot is a good example of that. Sycophantic confident incompetence. You end up second guessing all its output and having to verify everything. Potential time-savings quickly vanish.
On that note the influence big tech corporations have on the advancement of AI is probably the most regrettable. I guess that’s the paradox which is most difficult to reconcile. Without it the technology may never have seen the light of day. With it the need to make returns on investment colours the implementation of it.
How do you feel about AI generated images? Does it annoy you if someone uses them in a blog post?
AI generated images seem as formulaic as AI generated text when it comes straight off the bat. If that’s your bag though and serves your purpose all power to your elbow. Annoy me? Err… no. If I got annoyed about that I’d have a word with myself.
The internet is flooded with AI slop now, full of generated text, images, audio and videos. How do you filter it from authentic human creation? Do you have a strategy?
Attention is all you need. I choose what I pay attention to. I know how to use a remote on my TV, tune in a radio, and use the internet. If I come across something that interests or entertains me I will stick with it. If it does not I move on. The question whether something is AI generated or “…from authentic human creation” is not the deciding factor.
Are you hopeful for a better future with A.I. or a dystopian one?
I vacillate between the two. A.I. will contribute to rather than cause either of the two. I can be as a prone to badly predicting the future as the next person. Blinkered by what I know. Unaware of that which I do not. Let’s keep an eye on the baby as we pour the bathwater out.
