AI as Rosetta Stone or You Never Know Until You Try
AI as Rosetta Stone or You Never Know Until You Try
Forget what you’ve heard about AI’s limitations. The real limitations are your imagination and willingness to experiment. If you’re going to fully leverage AI, you need to be willing to try new uses and accept that some experiments will fail. If you aren’t, you might be missing out on some fantastic uses of AI. The little story below illustrates the value of trying. I hit a roadblock and decided to see if AI could help me.
I was giving feedback on an article that include extensive screenshots of AI chat sessions in German. Although I know a few German words and phrases (like ein bier bitte), I was clueless about what was going on in these chat sessions. The entire paper was centered on these chats, so not being able to read them was a problem.
AI tools are pretty good at interpreting images, so I thought I’d see if Gemini could translate the chat sessions based on screenshots from the paper. I took a screenshot of the chat session (which was an image, not text in the paper), pasted it into Gemini and then gave it this prompt:
Please translate this into English.
That’s it. Nothing special, no complicated technique. I just asked for a translation. It worked flawlessly, as far as I could tell. It may have gotten a word or two wrong, but in a minute or two, I had what I needed and was able to provide feedback on the paper, which turned out to be pretty good. (Unfortunately, since it’s unpublished I can’t provide any screenshots from my Gemini session.) Here’s the thing: Doing this was dead solid simple. Anyone with even a slight bit of knowledge about using AI can do the same thing. By the way, I tried this in ChatGPT and Claude and got the same results with some minor wording differences.
There were a few risks here. The biggest one is AI getting the translation wrong. I had to assess the reasonableness of the translation. Fortunately, the context of the paper helped me assess the translation. It fit with the related text in the paper, so I was confident that Gemini was close enough. Getting the same translations from ChatGPT and Claude boosted my confidence. However, if the exact wording had been critical (like in a contract), I would have taken further steps to ensure accuracy. Overall, I couldn’t have been more pleased with the results.
Great, so here’s one more thing AI can do. But, how often do you need to translate screenshots? For most of us, that’s probably a pretty rare use. Why did I write an article about my experience then? There’s a bigger message here, one that you’d be well served to keep in mind. Generative AI is immensely powerful and flexible. It is a general purpose technology with almost endless applications. So, if you encounter a problem, it might be worth seeing if AI can help you.
Think of it this way. I hit a snag with the paper; I couldn’t understand the chat sessions. So, I decided to try Gemini and it worked flawlessly. But what if it hadn’t? I would have wasted a couple of minutes (literally), but would be no worse off otherwise. Even if it didn’t work, I would have learned a little more about the limitations of AI. That’s a win for me.
Here’s the big message: Experimentation with AI often pays dividends. AI isn’t like a word processor. Amazing as they are, word processors are limited as are most technologies. Generative AI isn’t like that. It’s capabilities are limited only by your skill and imagination. OK, I’m exaggerating a bit here, but the idea is sound. Nobody knows the full range of what AI can do … nobody. Every day people are finding new, interesting uses for AI. You can do the same, you just have to be willing to try.
I’ve done lots of experiments with AI. A fun one that’s making its way around the Interwebs is asking ChatGPT to create an image of me based on our chats. Here’s the result:
I’m not sure why ChatGPT thinks I have a pocket watch or wear vests, but this was a fun little experiment. I also asked Gemini to trouble shoot a golf cart problem and to find a specific university policy. The last one is kind of interesting. Finding policies is a nightmare at many universities, so I wasn’t confident that Gemini could help, but it took me directly to the relevant policy. I also asked Gemini where I could buy high quality at bargain prices. That experiment failed, by the way.
That’s the thing about experiments, some succeed and some will fail, but you’ll learn, even from the failures. Yes, you need to be smart about your experiments. My decision to try AI for translation was easy. The downside was two minutes lost. Other experiments will cost much more in time and effort. So, you need to weigh the potential benefits against the potential costs. For high-consequence experiments, you need to verify AI’s results carefully.
Here’s the big takeaway. Adopt an experimentation mindset that is imaginative and willing to accept and learn from occasional failures. Do this, and you’ll soon find some great new uses for AI and, as a bonus, will naturally learn about the capabilities and limitations of AI. Remember, you won’t know until you try.
Over the next week, I want you to try something crazy with AI every day. It could be something work-related or something fun. Just try something new. Use AI to suggest a fun foreign-language nickname for a pet (Sasha is also known as Wilde Katzen, Wild Cat, around our house). Ask AI to look for subtext in a long email thread. Use your imagination. If your experiment works, great. If it doesn’t, what can you learn from the experiment? Devote 15 or 20 minutes each day to literally playing around with AI and you’ll develop an experimentation mindset. You’ll also be amazed at what you can learn. I promise.
Footnote:
Google Lens makes this even easier. I’ll explore Google Lens in a future article.
If you’d like help in using AI to save time and effort, I’m available for speaking and consulting engagements. I’m especially good at helping reluctant faculty and staff more comfortable with exploring AI. If you’re interested, I’m happy to talk with you to see how I can assist. Just email me at craig@AIGoesToCollege.com.