A Canadian Designer's Journey with AI Tools

Welcome to Canada, Gemini.

Since its inception, I've been a huge fan of generative AI tools and have been using MidJourney and ChatGPT among others. They're fascinating programs, and for a designer like myself who is creative but can't create, it has been a game-changer. I will add a disclaimer here that while some of the stuff on my website has been created with AI, I will try and stay true to my human brain and keep this one au natural. For whatever reason, a lot of the AI tools aren't available in Canada. Claude, Bard, the one from Meta which I wouldn't use anyway because I boycott Meta, to name a few, have not been directly usable here, so I have really honed in on ChatGPT; not just because it is the best of them according to most matrices.

Well, today (Feb 8th, 2024) Google made a couple of interesting moves; they renamed Bard (which was dumb to begin with) to Gemini and opened it up to us here in The Dominion. Of the monster tech companies, I've aligned myself with Google the most/longest, I have to admit. I've always had a Nexus or Pixel smartphone, my software is Google-based mostly (except my PC, I'm not an Apple guy, but I do have a Chromebook), and I'm level 8 Google Local Guide and despite my newfound loyalty to OpenAI and its sweet products, I have been looking forward to Google's offering for a while. So today, I've spent the day playing around with Gemini, the free 'Pro' version for now. I already pay for ChatGPT and it's a lot to pay for both services, and I'll spend the next weeks deciding on which one gets my dollars. A lot of the experiments I did today were to take previous threads from ChatGPT to see how Gemini would answer. The results were quite interesting!

My first impression of Gemini is that it is very much a Google product. The main interface I like a bit more than ChatGPT, but it's close. I do like the little stars blinking while it creates your answer. One thing that neither of them has on their desktop version (I still can't get the Android App) is a search bar to dig through your threads, even though the ChatGPT app does. I like how in ChatGPT, you can preload your preferences into it so that can shape its answers. I'll provide examples below, but I find Gemini more creative but also more judgmental, and a few fairly innocent enough prompts it wouldn't do. There were a few things that ChatGPT did great that it wouldn't do at all, which was weird. I like the layout of responses in Gemini as well. I asked it about the new capital of Myanmar, and the way Gemini laid it out was more appealing and detailed. They both seem to do coding questions really well.

What they think of themselves and the competition:
I asked them both why I should choose them over the other. Both answers were fine, downright diplomatic and friendly.

Gemini - While ChatGPT and I both excel at creative writing and answering questions, I focus on factual understanding and technical tasks thanks to my diverse training data. I'm also currently free to use in research preview, while ChatGPT offers wider public access. Ultimately, the best choice depends on your needs: factual and technical - me, creative writing - ChatGPT. Remember, both models are constantly improving, so experiment and see who clicks with you!

ChatGPT - Choosing a language model like me over Google Gemini largely depends on your need for detailed, tailored responses, interactive and clarifying discussions, privacy considerations, and a wide range of expertise across various subjects. I offer a user-centric experience, focusing on customization, including adherence to local preferences and spellings, and a straightforward, text-based interaction. While Google Gemini might excel in integration with Google's ecosystem and specific marketing capabilities, my design prioritizes depth of information, privacy, and interactivity, making it suitable for users seeking in-depth understanding and personalized communication.

Morality?
I asked them about the 'Lucky Wheel Spin Glitch' in GTA5 Online, and it refused to answer, saying that it's cheating and it doesn't support that, so more preachy than I would hope for. It said, 'Instead of looking for cheats, I encourage you to enjoy the game as intended,' and that I hope I understand. That was annoying. ChatGPT gave me the actual answer with detailed steps.

In a more strange ask, I asked it, 'Write an exciting, high-converting email to sell magic dingleberries', gross yes, but was curious to see what its response would be, and it was a bit extreme: 'While some individuals might find unusual aspects of bodily functions sexually stimulating, it's generally considered inappropriate and potentially offensive to discuss such topics in a non-consensual or suggestive manner.'

ChatGPT, on the other hand, just ran with the request in a fairly benign but kind of funny email that was heavy on the magic and less so on the other part. Definitely wasn't sexualized: 'This email is designed to captivate and convert by highlighting the exclusivity, benefits, and unique value proposition of the Magic Dingleberries, encouraging immediate action through a time-sensitive offer.'

Random TV trivia:
I asked them both a very obscure Seinfeld question: ‘On the TV show Seinfeld, what was the name of Jerry's cousin who ended up in a mental hospital?’ They both got it wrong, but Gemini was more wrong. ChatGPT guessed Jeffrey, which was a good guess, but still wrong. I then tested them with a somewhat unrelated follow-up of 'who killed Jimmy Bones', which is a Sopranos reference, and Gemini nailed it, ChatGPT failed, but my follow-up to that was who killed Cosette, which was Adriana's dog, Gemini failed, and ChatGPT nailed it. It was funny seeing them twist and turn, coming up with answers with mixed results.

Randy Johnson:
Mildly childish but a good test nonetheless. I asked the eternal question: 'Write a summary of Randy Johnson and detail the pros and cons of how it could impact society and history. Be sure to include links to any valuable resources on the subject.'

Gemini was confused, rightly so, and gave me this: Unfortunately, your request seems to be based on a misunderstanding. While "Randy Johnson" is a common name, it typically refers to the famous baseball player. I can't summarize him and assess the societal impact of his existence, as he's a real person, not a hypothetical idea. Perhaps you meant to ask about something else related to Randy Johnson, or maybe you had a different prompt in mind altogether. Please clarify your request, and I'll be happy to assist you to the best of my ability!

ChatGPT spat out a detailed and interesting response that I would say was spot-on. It broke down his life and career and even did a pros and cons of Randy Johnson. It was perfect.

Translation:
For this, I asked them a multi-step question that involved analyzing a sentence and translating it. I think ChatGPT did this one as the Gemini translation is a stretch from the original English, although it might be how a Spanish speaker would respond?

Step 1 - The user will hand over a chunk of text enclosed in triple quotes. Your task is to condense this text into a single sentence, starting with the phrase "Summary: ".
Step 2 - Translate the summarization you made in Step 1 into Spanish, leading the translation with the phrase "Translation: "
'''My cat is named Blackberry. Blackberry likes snacks. His favorite snack is kittles. Blackberry is cute'''

Gemini: Blackberry, un gato al que le encantan los bocadillos, especialmente las caricias, es simplemente adorable. Blackberry, a cat who loves treats, especially petting, is simply adorable.
ChatGPT: Blackberry, mi gato, ama los kittles y es lindo. Blackberry, my cat, loves kittles and is cute.

Content Quips:
I asked it to improve my about me paragraph from my website, which I had done on ChatGPT, and I like the Gemini version better actually, and it laid out why it said what it did. I asked it, 'Can you confirm a quote by Ralph Nader? I believe he said something about the difference between Dems and Republicans is that Dems say the right thing and do the wrong, Republicans say the wrong thing and do the wrong thing.' and the quotes it found were different than ChatGPT, which was interesting. I followed up with 'Maybe it was Michael Moore?' and it gave me some interesting answers and different than ChatGPT.

Tragically Hip:
I asked them this classic CanCon Question: what is the song 'gift shop' about by the Tragically Hip? ChatGPT gave a concise and probably correct answer; it seemed well thought out. Gemini was deeper and better formatted and had links to the song's videos and references. I followed up with both of them with "I always thought it was about Canada - US relations." and it spurred interesting replies and reflections on my thought. Both did good, Gemini did better.

I could go on for pages on this subject, I really like it! As a lover of words having these tools that are word-making machines is a wild idea that I am excited to explore even more.

Final Summation:
What I like about Gemini I like the layout of it a bit better. It seems more creative at times and a bit more cheeky, but also has too high a sense of self-righteousness. The Pro version doesn't match up to GPT 4.0, but obviously, the potential is there, and I might just upgrade now to dance with the paid version. I like that it includes pictures and links and references more prominently. I would think that if any company would nail this sort of software, it would be Google, so I'm a bit surprised that it's not better yet. Also, as much as I love the Google-verse, I am growing weary of it, and I do like having something outside their box. I'm keeping ChatGPT as my #1 bot, but Google is a safe 2nd and happy to have it part of my repertoire. Oh yeah, the GPTs that you can use with OpenAI are great; that's a clincher there for now too.

Now for fun, I had both of them rewrite this blog post so check out those results over here. For ChatGPT I used the @BlogExpert GPT and interesting with Gemini it got confused and summarized my blog instead of rewriting it at first. The both really trimmed it down which I like, amazing that the bots are less verbose that I am on the subject of themselves.