Google’s AI model Bard prefers iPhones to Pixels, but it’s not why you may think

Here’s a quick list of things that I remember Google talking about during its I/O presentation of 2023:

• AI
• The Pixel 7a
• Google being responsible with AI
• The Pixel Fold
• Being responsible with AI
• Bard
• AI

Now, I’m not sure how obvious it is, so let me spell it out: AI is so hot right now (yes, please read it in that voice). What was dubbed as only a mere “experiment” up until recently is now one of the internet’s superstars, or at least in the regions where Bard is available for testing. Because despite Google announcing that Bard is to become more “widely available”, the page still tells me to go away, but oh well.

And I’m excited to play with Bard, because — unlike with ChatGPT, which a dev can take and shape as they see fit — Google actually pre-prepped me mentally with a showcase of some real-world applications for how to use Bard. And I can’t wait to try and break all of them!

So I find it incredible that someone out there, who has access to Bard, decided to ask it to join the Android vs. iOS debate. Because of course they did. The funny part? Bard initially made the impression that it prefers iOS.

AndroidPolice then decided to join in on the fun and gave the bot a chance to explain itself. Apparently, Bard finds iOS to be user-friendly and with a polished UI — and hey, no arguments there — but it is also an updates connoisseur, as it appreciates the frequency and consistency of iOS updates.

You see, it’s funny, because there’s this majority of Android phone manufacturers who don’t get why that last bit is important, but I digress.

But then it got even crazier, as when Bard was asked the same question again, it did a total 180 and picked Android. This time, Bard cited UI customization and a wider selection of apps as reinforcement, but it also said that Android phones are more affordable.

If Bard keeps this up, it may end up being a Monopoly pro in a year or two.

So what’s going on here, why is Bard unable to make a final decision? Well, because it’s not its job to end internet feuds. And because we haven’t reached the point where AI can actually reach its own conclusions.

Now, get your thinking caps on. We’re about to make some deductions! And based on common sense, Bard is likely doing this because:

• It is trying to evaluate something like “What most people liked in the last 24 hours”
• It runs a quick background check and then spurts out the response more likely to appeal to you in particular

The second one, however, seems unlikely. After all of Google’s talk about being responsible with AI, designing Bard in a way which helps it encapsulate people in their own bubbles sounds to be the opposite of “responsible”.

Right now, talking to AI feels a bit like trying to prompt a toddler to say something funny. Until the internet people get adjusted to AI being a thing — and said AI (Bard) gets an actual wider release — were likely to get more memes instead of tests about the platforms’ actual capabilities.

But what do I know, I don’t have access to Bard.