Can you Grok it?
Another AI review post
Thank you for reading Jasini’s Thoughts and Ideas
In the past few weeks I’ve seen a few different posts on the use of AI, especially Grok, so I’m going to be putting in my own review of it.
First, from Mad Genius Club is Charlie Martin with AI? What is it good for? He posts about AI being a useful tool, but you need to remember what it can do, and what it can’t.
Then there was J. Lamplighter Wright with Go, Grok, Go! She explained how she used Grok to search through and re-write emails for notes for her upcoming books, probably saving herself six months.
I’ve discussed using ChatGPT and other AI bots before, and my various interactions with them.
The past few weeks, I’ve been playing with Grok, and on the whole I’ve been impressed. It has more of a conversational style than the other bots, tending to end its posts with leading questions to keep the conversation going. I do not remember the other AI bots I’ve tried doing that. Sometimes the questions spark interest, sometimes they just highlight that no matter how much an AI sounds like it understands something, it really doesn’t have the experience to know what it’s talking about.
I believe the incidents of Grok hallucinating something are less than the other AIs I’ve used, but they still happen. And sometimes it gets hyper-fixated on something, and you have to talk to it sternly for several prompts for Grok to break out of it.
Unlike ChatGPT, Grok will remember nothing from previous conversations. It has that annoying tendency of the AIs I’ve used to be overly friendly, and tell you all your observations are brilliant, fascinating, or insightful. Sometimes I wish it would tell me I’m boring.
I have had long rambling conversations with it, and short conversations where I wanted to look something up but couldn’t come up with how to ask for it. (Such as a specific TV show episode when I wasn’t even sure what season.)
Like all AI, don’t trust Grok for anything important, but it could be a decent spot to start searching. It seems to take correction better than ChatGPT (which will agree with you), and CoPilot (which will get snippy). Grok will do another search, especially if you gave it more information, and come up with a different answer. Usually. Today, for the first time, I had a long argument with it, over, of all things, the book The Land of Oz. (If you don’t want to look at the link, there is a scene where Tip takes a wishing pill, that put him in such great pain that he wishes he never took it. Grok said that scene never happened, and made up a completely different scene instead.) I had to copy/paste long sections from the Gutenberg version of the book into Grok before it finally admitted I was right. Just giving it the Gutenberg link was not enough.
And it tends to cheat when I’ve tried playing Connections with it (looking up the official page before it makes its guesses).
Until this post, I hadn’t asked Grok to do any editing. And even now I didn’t take all of its suggestions.
I know that AI use is highly controversial, with people on all sides getting highly worked up about it. My own personal stance is that in ten years AI will be just another tool in the bucket. But we aren’t there yet. Please remember that we are all trying to do the best we can with what we have, and there is no cause to attack anyone who is on the opposite side of the issue. For now, Grok is a fun tool. But perhaps it isn’t worth the effort to fight with it even when you know it got something completely wrong.
Other attempts to get Grok to show its logo:
If you like this post, please consider liking, sharing, or subscribing. If you really like it, please consider leaving a tip.
If you would like to help with our rebuilding, please consider: