Google Bard has almost become the underdog in the current AI chatbot race, and because of this (along with its very polite and friendly tone), I have quite a soft spot for it. Bard has a great UI and the potential to dominate the race if Google could just get things together, go through the ‘training montage’ phase, and actually improve the chatbot’s capabilities.
Inspired by Science Alert’s game of Wordle with ChatGPT I decided to put Bard to the test and play a game or two with it to see how it matches up. I had never actually played Wordle before today and I was hoping that after this little experiment, I could use Google Bard as a helper; a ‘player two,’ if you will, when I do eventually embark on my Wordle journey. Because if there’s one thing that should be able to crush a word game such as Wordle, it's a language-learning AI chatbot. And it shouldn't even need any Wordle hints in order to do so.
So, I asked Bard if it was familiar with the game and once it confirmed and read the rules back to me, I booted up a random Wordle site (not the actual game itself, but a website that uses past games so that people can play more than once a day). I asked Bard again if it was ready and understood the game, then asked it for its first guess.
Bad decisions
Its first guess was the word SLATE, which is generally regarded as one of the best Wordle starting words, according to our in-house experts (well, fanatics) here on TechRadar. I was very impressed off the bat that it was able to play such a well-known and (usually) wise choice of first word and that set my hopes high. I put the word into the site and gave Bard the response back via text. I even offered the bot words of encouragement and reminders of what the rules of the game are, just to keep it on track.
Bard carried on suggesting words, but everything fell apart after about the third attempt, where it offered up the word SLANT. This is where the polite and apologetic tone came to a grinding halt and Bard started to insist that the word was SLANTED (it was not).
After correcting the bot and reminding it to only guess five-letter words, it apologized and returned with SLANTS, then carried on with different variations of the word SLANT until I was ready to pull out my teeth. I reminded it of which letters were green (at this point it was the first letter S, the last letter T) and it went from too many letters to too few.
By this stage I was frustrated and incredibly disappointed, because while I know, objectively, that Microsoft’s Bing AI and ChatGPT are at this point in time more sophisticated chatbots, I had hoped that Google Bard would pull through on this. After all, Bard has been trained on a plethora of data, from text and code datasets to books, articles, and social media posts. There is such a rich variety of knowledge it can draw from that I assumed it would smash this meager challenge.
I gave up in the end, and my Wordle remained unanswered. As I said earlier, I am a big fan of Google Bard. I think it has the potential to do well, but I can’t see it doing that until some major updates are made. It’s very easy to see why people are gravitating more to Microsoft Bing, not least because Bing is getting updates and new features while Bard is not. Until that changes, Bard remains dead last on most metrics.
