The first week of Blaugust is over, but I’m still feeling like I’m just starting this event up. So I’ll fall back on the writing prompts for the first week and talk about myself in this post rather than coming up with something new.
Firstly, I already wrote a proper post about who I am for Blaugust 2024– check that out if you want something I actually put some serious effort into. But because I’m an expert at being lazy, I’ll take all of my inspiration for this year from someone else’s hard work. Bhagpuss wrote a post about using AI to find out about himself, and that sounds right up my alley.
Note that my ‘featured’ image for this post was (obviously) AI generated: I use AI to generate a lot of my featured images. My prompt was “generate a comic book style image of an AI struggling to figure out who Ubergeek Kelly Adams is.” I love the fact that the illustration says “Ubergeelys World”. Gotta love AI image generation…
What ChatGPT thinks about me
To steal Bhagpuss’ idea I tried to get AI to tell me about myself. ChatGPT couldn’t do much with just my name, but with a different prompt including the fact that I’m a blogger and sometimes referred to as ‘Ubergeek’, I managed to get a decent enough summary:
But what do you really think?
The summary about me generated by ChatGPT is more complimentary than something I’d write myself. But in general terms it isn’t bad, although it clearly cribs a lot from my own commentary about who I am. That isn’t too surprising though: I doubt that there are many other sources of information about me other than my own words.
I was pleased that the AI found one of my other blogs, Words of the AgingGamer. But I’m curious why it didn’t find Geek on a Harley: both are linked to from the main page of my site. It could have also found references to my Mastodon account or Facebook page, or even gone crazy and found my LinkedIn profile.
My guess is that ChatGPT really only does a shallow check based on the words in the prompt: I mentioned I’m a blogger, so it only looked for blogs. That’s not unreasonable, and possibly changing the prompt to mention other places I’m ‘present’ could trigger a deeper search by the AI.
Digging deeper
I did a bit of experimenting with prompts asking ChatGPT to investigate Facebook, LinkedIn, Mastodon, and Bluesky. ChatGPT produced a fair amount of references for Mastodon and Bluesky, but pretty much nothing from Facebook and LinkedIn
So it found the ‘obvious’ things i.e: my Mastodon account is, as it says, ‘linked right on his blog’. There is enough information on my blog that the AI could have done a better job searching Facebook and LinkedIn. For example, it probably could have added search parameters regarding where I’ve lived based on the contents of my blog posts. But I’m guessing that the developers of ChatGPT intentionally block that kind of ‘digging’ as being a bit too ‘creepy’.
I also asked ChatGPT to generate some summaries about my Mastodon and Bluesky posts. It provided a reasonable summary of both, and then attempted to ‘summarize the summaries’ with a bit of tone analysis:
AI kinda knows who I am
I was actually, in a strange way, somewhat pleased that ChatGPT was able to find something about me. My earlier attempts using prompts based just on my name were disappointing, but at least if I give it some clues like ‘Ubergeek’ and ‘blogger’ it seems capable of tracking me down.
I suspect that the key word I used in my AI prompt was ‘Ubergeek’. I added this to my blog title just after Blaugust 2024 mostly because I had observed Google search couldn’t find me with “Kelly Adams” but could when I added the word “Ubergeek”. Having a mostly unique name or ‘brand’ on the internet does make a difference and it seems reasonable that being findable on Google will help AI find you as well.
Maybe for Blaugust 2026 I’ll be able to write “To find out who I am, type this prompt into ChatGPT” instead of writing a proper post about myself. Nah… I’d rather blather on for a page or two about the technology, how wrong or right it is, and why I’m (hopefully) still smarter than the AI. That’s just how I roll…
Very pleased to see a couple of people giving this a try. It’s fascinating to see what the AIs do and don’t focus on. Jeromai is the one who seems to know most about it in this part of the blogosphere and he maintains it’s all in the prompting but you’re quite right that it’s all happening inside a black box that no-one really understands.
Oh hell no, I know nuthing. I’m just guessing, like everybody else, and intuiting based on results and what it’s showing from its sources.
AI is interesting to experiment with, Bhagpuss. And the ‘blackbox’ part of things is definitely true.
As Jeromai indicates, right now I think there is a lot of experimentation and intuition required to craft ‘effective’ prompts. Some of the AIs, at least on the paid tiers, can provide ‘logs’ or scripts showing how they proceeded (including which references they used) based on your prompt. And that can help folks who are more serious about it than I am to get more robust results.
I currently use ChatGPT almost exclusively as I am paying the monthly fee for it: that means I’m probably missing a lot of factors relevant to the other AIs out there. But I am pretty certain that all of the AIs today have intentional ‘blocks’ or restrictions on their investigative behaviour. This prevents them from doing the kind of deep relational dives based on multiple sources both ‘public’ and ‘nearly public’ (e.g.: services like license, court, or property registries that you have to pay a few bucks each month to peruse) that would likely scare the bejeezus out of most people.
Any professional investigator, particularly anyone in a government anywhere in the world, likely has access to an AI with all of those restrictions removed. This is why establishing somewhat strict laws regarding how our data can be used (like GDPR in the European Union), convoluted and frustrating as they may be, is not really ‘optional’. Sadly no such laws exist across most of the world.
I’ve done this too, but not posted it. AI programs like this are hardcoded to give praise and validation, which is great when you’re asking about your personal blog and it thinks you’re the bees knees. It’s kind of creepy too, I think, when it’s every other message and not just at the start of a chat, and something I think everyone needs to consider as this all develops.
I agree with you whole-heartedly, Alex, that the AI always seems to exclusively find ‘positive’ things to say and it is a bit off-putting. My initial reaction wasn’t “Oh gee, look how correct the summary is” but rather “Oh gee, it actually found something about me to analyze” š
I had a go at this and the results were oddly upbeat. I would have been happier if the AI had made at least one criticism.