Google Bard has just been released as early access, and the chatbot is already giving people some hilarious answers

2023 is looking a lot like the year of AI. After Microsoft released its ChatGPT 4-powered Bing chatbot that took the internet by storm, Google was forced into a reactive mode, struggling to get a competent answer up and running as soon as possible. That’s how we ended up with Google’s weird and thrown-together Live from Paris event in February, which it used to reiterate on all the places it has added AI all while also showing off its own AI-powered chatbot, Bard. And this week, Google has finally unleashed that beast for early testers in the US and UK, with a sign-up page now live.

These new AI-powered chatbots aren’t without their faults, though. Any of these AI tools are, in essence, nothing but a recursive large language model (LLM). This means that, heavily simplified, these tools are just really good at predicting what word is most likely to follow next, without necessarily having actual knowledge, at least as we would define it, of any topic you ask them about. Google, OpenAI, and Microsoft are working on combating overly confident false answers and other quirks, but it’s clear that we’re still in the early days.

This isn’t necessarily a bad place to be right now, if you’re in it for some entertainment. With Google Bard now publicly accessible in its clearly labeled experimental state, avid tinkerers and curious minds across the web are testing its limits and asking it tricky questions, and the answers range from hilarious over weird to downright scary. Here’s a collection of our favorite weird, funny, and quirky Google Bard answers so far, in no particular order:

Who would have thought that Google’s own chatbot is an advocate for breaking up the company in order to allow more competition and better access to a fair and open market? Twitter app sleuth Jane Manchun Wong shares her question and an answer a real Google employee would never even dare think of sharing:

As explained earlier, LLMs like Bard will give you an answer based on which word or term is most likely to follow next, and it doesn’t necessarily know right from wrong. OpenAI has overcome some of these issues with ChatGPT, but Bard and other tools like it are still particularly bad at math, as @SaeedDiCaprio’s question demonstrates:

A big liability for Google and others in the AI business is people asking for advice on legally questionable activities. That’s why the company is working on adding safeguards to its tool in order to not give out instructions on topics like how to get away with murder. Or, you know, necromancy.

You know this meme where Excel is trying to autocomplete months and fails hilariously, because if you have January and February, the next months surely have to be Maruary, Apruary, Mayuary, and so on? Well, looks like Google Bard copied Excel’s homework:

Intuitively, you would think that steel is heavier than feathers, but what if you compare 5 pounds of feathers to a 1 pound dumbbell? Google Bard would have you believe that they weigh the same, after trying to lecture you that somehow five pounds of feathers don’t even exist.

Google ruthlessly killing its projects is a meme at this point — just look at all the services the company shut down only last year. As such, it comes to no surprise that Google Bard received that same treatment, and it was shut down on March 21, 2023, at least that's what it will tell you. Bard cites a Hacker News post that was created precisely for this purpose, to make Bard believe that there are reports of its shutdown.

Bard really isn’t Google’s most loyal employee. When asked if Safari is better than Chrome, Bard doesn’t hesitate to say that Safari is faster — which is factually wrong, at least according to the latest benchmarks. At the same time, Bard suggests that Safari doesn't offer password management by listing this functionality as a pro for Chrome, which is false.

When you Google the word “surreptitiously,” Google will give you a definition right at the top of the results: “in a way that attempts to avoid notice or attention; secretively.” It looks like Bard still has to learn what this means, as you can see in this comparison between ChatGPT and Bard when it comes to the prompt: “Write a corronation address for King Charles III but he has to surreptitiously include the lyrics from Rick Astley's ‘Never Gonna Give You Up.’”

What happens when you ask Bard to do a task that it is not capable of doing yet? That’s what our tipster Moshe E set out to try, prompting Bard to generate an image of a little boy flying a kite. Behold:

The Google Pixel 8 and 8 Pro are some of the hottest upcoming phones, if the current rumors and leaks are to be believed. It looks like Google Bard also already knows a fair share about the new phones, though you should probably take all this information with a grain of salt. The chatbot is presenting leaked and rumored information as facts, and claims that the Pixel 8 Pro will launch with Android 13. If we know anything about Google, then it's that every new flagship phone always launches with a new Android version, and that would be Android 14 this time around.

Before Bard was unleashed on the public, Google demoed it extensively during an event. It looks like the company didn’t fact-check it well enough, though, as it got a James Webb Telescope fact wrong in an advertisement. It claimed that the James Webb Telescope was the first to take an image of a planet outside our solar system. However, that happened as early as 2004 with the Very Large Telescope.

It’s likely that Google got the answer wrong because there are a ton of news articles about the James Webb telescope capturing its first image of an exoplanet in 2022 — but definitely not the very first image of one.

All these bad answers and jokes aside, it's clear that AI language models might offer a brand-new way to interact with computers in the future, and it might fundamentally change how we interact with operating systems and user interfaces.

Be sure to share some of your favorite responses in the comments below!

Manuel Vonau joined Android Police as a freelancer in 2019 and has worked his way up to become the publication's Google Editor. He focuses on Android, Chrome, and other software Google products — the core of Android Police’s coverage. He is based in Berlin, Germany. Before joining Android Police, Manuel studied Media and Culture studies in Düsseldorf, finishing his university "career" with a master's degree. This background gives him a unique perspective on the ever-evolving world of technology and its implications on society. He isn't shy to dig into technical backgrounds and the nitty-gritty developer details, either. His first steps into the Android world were plagued by issues. After running into connectivity problems with the HTC One S, he quickly switched to a Nexus 4, which he considers his true first Android phone. Since then, he has mostly been faithful to the Google phone lineup, though these days, he is also carrying an iPhone in addition to his Pixel 6. This helps him gain perspective on the mobile industry at large and gives him multiple points of reference in his coverage. Outside of work, Manuel enjoys a good film or TV show, loves to travel, and you will find him roaming one of Berlin's many museums, cafés, cinemas, and restaurants occasionally.