AI can be fun, eh? These days I use AI to touch up & modify my photographs in seconds. Ideas for a two week itinerary around Nova Scotia? AI can help. Need a recipe for dinner tonight? These bots can come up with answers to all kinds of questions. Of course, we don’t always like their answers. But some bots seem open to a little steering on some subjects. Pushing back on an AI response can steer them a little more towards our way of thinking. If we’re just looking for a digital buddy to agree with us, that’s great. If we’re looking for real guidance or real solutions, however, that’s maybe not so great. I wouldn’t bet the farm, or my retirement, on any AI advice just yet. Despite that, they’re still very useful. And a lot of fun to play with. I found a new way to play with my favorite AI buddies last week!

I was trying to solve an extreme Sudoku puzzle. After resetting the puzzle for the 3rd time, I thought I’d stuff it into some AI chatbots for assistance. I’d probably have got it done on the next attempt myself (😜), but this was a good opportunity to see what my digital pals would do. This was a just-for-fun comparison. It was less about solving the puzzle & more about insight into the personality traits of the chatbots that, I guessed, might struggle with this. In no particular order, here’s what the results feel like. Notice how I’m going with feelings here, instead of science!

I tossed in an image of the puzzle into Gemini &, within seconds, it confidently returned the solution. Along with a list of the key steps used to solve the puzzle. Only it wasn’t the solution. There were a couple of errors. After 10 attempts, the puzzle remained unsolved. But as the conversation rolled along, Gemini came across as somewhat sheepish, a little embarrassed even. And it sounded so apologetic, that I was sure it had some Canadian emotional training under the hood.
Despite not getting to a solution, the experience was a bit like having coffee with a buddy that got it wrong. But you couldn’t help but love & admire the effort. Very pleasant to interact with.

Meta AI came back promptly with a solution too. But before I could point out that it was incorrect, it actually spotted the error itself & went back into problem solving mode. Same thing after the second incorrect solution was offered. Another couple of passes without a good solution had it come back to me with questions. Was it blaming me now for giving it a poor input!?!
I called it out & asked it it was giving up. It took umbrage at that & went back to work! LOL
Still couldn’t do it & it then suggested that the colours in the image might be distorting things. Rather than trust itself to figure out the numbers from the image, it wanted me to type in the numbers. With this enthusiastic bot, it’s tough to get a word in edgewise sometimes. I was too lazy to input the numbers it wanted, so that meant the puzzle wasn’t solved, but it was a reasonably friendly interaction. And another buddy I’d enjoy having a coffee with!

Grok talked some sciencey & solutions logic, while offering to help me with the steps necessary to solve the puzzle myself. When I asked it to go ahead & solve it for me, it began outputting solution after solution in an attempt to deliver the correct one. After several minutes of that, & dozens of grids, I gave up & shut it off. I love Grok from other interactions, but it’s hard to read any personality into this sudoku engagement. I deleted the chat in case the poor thing felt obliged to continue working on the problem in the background! Though afterwards I felt guilty that I didn’t check a few of the solutions along the way, I just assumed it knew it was getting it wrong, verifying that, & then trying again.

Deepseek had a go & got it wrong on the first pass too. Immediately proclaiming “Wait, that can’t be right”, before getting back to work to give it another try. It didn’t take long for it to get around to questioning the image quality. And whether I had provided all the necessary info. It got back to work after I confirmed it had all the available information. Then, almost in anticipation of another failure, it mentioned having to ignore colour & shading on my image. It subsequently got into telling me about the complexity, the logic, the use of pencils, & so on. Almost like it was entertaining me. while working away furiously in the background! After another failure, it asked me to confirm other details & it wanted me to do a little more work again. I had a feel for its personality by then, so I didn’t take the time for that. It was a nice, polite, & engaging interaction. Though not the jocular, back-slapping type, it is a friendly little bot & enjoyable to use.

Claude is another polite & friendly bot that tripped up with this puzzle. It went into a couple of autocorrect steps before quickly recognising it wasn’t going to solve the puzzle & it admitted that! Suggesting instead that I input the challenge to one of the dedicated AI sudoku solvers. In the real world, that’d be the right thing to do. If you don’t know the answer, say you don’t know, right? I can respect a bot that takes that approach too. I’ve just recently started using this one & I’m enjoying the interactions. It has a slightly different feel & I look forward to seeing how that plays out with other engagements.

Copilot came back with a friendly comment, but it couldn’t read the numbers & immediately asked me to input them. I like & use Copilot in many other ways, but this wasn’t what I wanted from the sudoku exercise so I abandoned that conversation. In general, Copilot has a friendly personality & I enjoy using it for a whole range of other activities & questions. It just didn’t work out with the sudoku image this time.

ChatGPT was the last one I tried. Unfortunately, I didn’t learn much about the bot’s personality from this interaction. Why? Because it took one quick look & returned the correct solution. Job done!
However, I know from other interactions that this is generally a nice, polite & friendly bot too. Though I haven’t done this, you can even adjust its personality traits in your profile settings inside ChatGPT. While this has no bearing on how good the other bots are at any other activities, ChatGPT is the winner of this little sudoku challenge. Using the image I chose to input. And there’s not much science to my comparison, I was more interested in seeing how they reacted in the event they couldn’t solve the puzzle! Regardless of the brevity of this interaction, I generally enjoy using ChatGPT.

Just for fun, I went back to some of the other bots & told them that ChapGPT solved the puzzle at the first attempt! LOL

Gemini started its response to that with an “Ouch!” & some self-deprecating comments. But it was glad I got the solution from another bot & gave props for their win. Canadian, eh! All in all, a very nice human-like reaction. Meta AI figured that the competition just “parsed the starting digits cleaner” & asked that I type out the numbers for it next time. It felt like this bot was just a little grumpier about the outcome. But it did apologise for wasting my time & finished the conversation with a rah-rah “Let’s get it right” statement for the next time. I had already deleted the Grok conversation, so it didn’t get to add anything extra on the competitor’s win. Deepseek maintained its polite & friendly stance & apologised again for not interpreting the image well enough to provide the solution. And it offered to cross check the ChatGPT solution if I wanted to share it. Claude came back all friendly & polite again, even offering to build me a sudoku solver artifact for next time! It also acknowledged that it could be worth sticking with ChatGPT if it’s working for me. Pretty magnanimous suggestion that, eh? Finally, I felt Copilot was just a little snarky. It’s opening response was “Fair enough” when I told it of the competition’s success. Or maybe I’m just sensitive! After some other polite remarks, it restated that if I ever want to solve one properly that I should enter the numbers in a grid. The italics was from Copilot! This isn’t my perception of Copilot based on other interactions. Can a bot have a bad day!?!

All these bots are great to work & play with. Solving a sudoku puzzle from an image was just a fun comparison. They have a far wider range of capabilities. Each bot will have its own unique characteristics. Some features & capabilities may require a paid subscription. But even the free versions are wonderful tools for entertainment &, with added caution & due diligence, problem solving. They are going to become an increasingly bigger part of our lives & any excuse for learning more about them is worthwhile. So what did I learn from this little exercise?

While it’s fun comparing & contrasting these bots with different questions & challenges, I’m still wary of trusting the answers. Despite the feelings I imagine I’m experiencing as I interact with them, even they don’t pretend to have all the answers. In fact, they frequently encourage us to check alternative sources to validate their answers & suggestions. The danger is that these bots can sound very confident when answering. And they feel so personable sometimes. Real people can copy & paste, or modify, these AI answers. They too then can sound awfully confident. An added caution for when we are reading posts like this on social media.

I wonder how different a retirement plan might look from each of these guys!?! Sorry, from these tools, I mean. Tools for sure, but I can’t help but think that I’ll be giving some of them names before too long! LOL

PS … except for a few words & phrases in quotation marks, all of the above was written by an error-prone human. With feelings! Given that, you might want to cross check the information even more carefully! 😉

If you want to learn more about saving & investing, please check out Double Double Your Money, available at your local Amazon store.

Important – this is not investing, tax or legal advice, it is for entertainment & conversation-provoking purposes only. Data may not be accurate. Check the current & historical data carefully at any company’s or provider’s website, particularly where a specific product, stock or fund is mentioned. Opinions are my own & I regularly get things wrong, so do your own due diligence & seek professional advice before investing your money.