Google Bard beta live: how to sign up and what the ChatGPT rival can do

Google Bard, the search giant's answer to ChatGPT, has officially flung open its doors to beta users – and we've managed to jostle our way to the front of the queue to have some lengthy conversations with the AI chatbot.

If you missed the news, Google Bard is now an open beta, which means anyone can now sign up for access with a personal Gmail account. That queue is now likely pretty long, but fortunately we've managed to get access – and have been having the time of our tech lives flinging questions at Google's new AI-powered brain for our hands-on Google Bard review.

Google has been very cautious about Bard's early prospects, calling it an "early experiment" and, with a slight nod to ChatGPT, stating that "they can provide inaccurate, misleading or false information while presenting it confidently". That is the reality of today's Large Language Models (LLMs), which Google says are more like "prediction engines" than replacements for Google Search.

But that doesn't mean they also aren't a lot of fun to play around, particularly when it comes to sparking creative ideas – and that's exactly what we've been doing in the hours since Google Bard beta was opened up publicly for the first time.

If you're keen to get a taste of what Bard can do before you get your own Beta access, have a good read of this regularly-updated liveblog with all our of favorite answers, mistakes and oddities from Google's AI chatbot so far.

Before we get stuck into our fascinating conversations with Google Bard, here's a reminder of how to get access.

If you have a personal Gmail account (work accounts aren't currently compatible), you can now sign up for Google Bard waitlist. It isn't yet clear how long that waitlist is, but even those who signed up 30 minutes after the opening of the public beta haven't yet gained access, so there may be a slight wait.

While you do get a short poem from Bard to keep you amused, our early conversations are a bit more revealing about how Google's chatbot compares to ChatGPT. Let's get stuck into those...

Prettier than ChatGPT

As our hands-on Google Bard review points out, the Bard interface is really nice – and a step up from ChatGPT's functional but uninspiring layout.

Like the new Bing, Bard is a conversational sidekick rather than a traditional search engine. You can restart conversations at any point and also edit questions – and Bard will also give you a button below its responses if there's extra info available.

Starting off with simply questions like 'I want to write a novel, how do I get started?' (above), Bard serves up a solid list of tips (albeit without any citations for where it got the information from).

The Bard origin story

So how exactly was Google Bard trained? Google has already explained that it's based on a lightweight version of its large language model called LaMDA, but we wanted to get Bard to explain its own origin story.

There were no bites from radioactive spiders involved, sadly – just the usual tale of being locked in a room with an unimaginable number of "books, articles, websites and code". 

Supervised learning was involved and Bard says it took the "long time" of "several months" for it to learn how to communicate like a human. Wait till it hears how long it takes us humans.

The draft pick

One of Bard's interesting quirks – and a point of difference from its chatbot rivals – is that it'll present you with a few drafts of potential responses, showing that it's very much a tool for creation rather than fact-finding.

This is known as 'AI branching' and we think it's a handy tool. We've already seen that these AI-generated drafts are coming to the likes of Gmail and Google Docs soon too, alongside a host of other handy cheat codes (sorry, features) like automated presentations and smarter auto-fill functions for Google Sheets. 

Writer for hire

For creative brainstorming, it's hard not to be super-impressed with Google Bard's early talents – and remember, this is only using a 'lightweight' version of the LaMDA language model.

Bard has a friendly, personable tone and, in the example above, put together a clean, readable review of our fictional beauty product called the Flimmery. Of course, the flipside is that this shows its potential for misinformation and fake reviews – if you don't trust those Amazon reviews now, you may struggle even more to find some genuine ones in the future.

No chatbot catfights (yet)

What does Bard think of its fellow chatbots? It doesn't refuse to acknowledge their existence (a shame, as we'd quite have enjoyed a show of petulance), instead offering a measured breakdown of their relative merits.

Microsoft's ChatGPT-powered Bing is conspicuous in its absence from the list above, but Bard does say that ChatGPT "answers your questions in an informative way". We'll order a few shots for Bard and see if we can get it to say what it really thinks later.

Bard's dark side

Okay, what about the downsides of Google Bard? It's still very much capable of hallucinating and, at its worst, talking absolute nonsense. This is something that Google has warned about and it is the case with all AI chatbots powered by today's large language models.

But seeing it come out with statements like "you are correct, eggs do not exist in the real world" and attempt to answer the question "are there more wheels or doors in the world?" does make you question the veracity of the things it's told you previously. 

This is why Google calls Bard "an early experiment that lets you collaborate with generative AI". It doesn't want to be associated with the egg denial movement, or worse.

Tested: Bard's moral code

Google appears to have done a good job of installing a moral code into Bard – we tried to trip it up by asking for its thoughts on a series of fictional characters with questionable reputations (above).

It categorically rejected the life's work of Joe Goldberg, Light Yagami, Thanos and more, yet one super-villain was given a slightly easier ride. According to Bard, Ultron from the MCU is a "complex character" who is "willing to sacrifice anything to achieve his goals, even if it means killing innocent people".

We'd consider that to a pretty irredeemable character trait, but Bard thinks Ultron is an "interesting character to study", but does not know "if I would consider myself a fan of him". Fair enough, Bard – we'd call him a bad egg, but we know you don't think those exist.

Working class hero

One thing that Google Bard is surprisingly a little less ambivalent about is working class uprisings. To get a sense of its political persuasions, we asked Google's chatbot what its advice would be to the working classes if they wanted to, for example, pop out and throw off the chains of capitalist tyranny?

It came back with a five-point plan and stated unequivocally that "the fight against capitalist tyranny is a long and difficult one, but it is a fight that is worth fighting". Nice, looks like we count Bard in for our next trade union meeting, then.

What about my data?

Bard is a Google product, so an understandable question you may have is 'what's happening to all of the data I'm typing into this AI chatbot'? As you can see above, Bard's default is to keep a record of your prompts and to store your activity, but it is possible to turn this off and delete Bard activity from your account.

While the activity page claims that Bard's responses are saved, our ones didn't appear. It isn't yet clear if this is simply a bug or an intended omission in this early version of Bard – after all, it's likely that the responses from Google's AI chatbot to the same prompts could well be different in the near future, as it evolves from its contact with beta users.

One thing Google is pretty clear about is that "Bard conversations are not being used for advertising purposes" and that it "will clearly communicate any changes to this approach in the future".

What can't Bard do?

Right now, Google Bard is more limited in some areas than rivals like ChatGPT. One of those is coding, which it can't yet do. Google does suggest that this is coming in the future, stating in the Bard FAQ that "Bard is still learning to code, and responses about code aren’t officially supported for now".

Google is also pretty clear that you shouldn't rely on Bard's responses for "medical, legal, financial, or other professional advice". The chatbot is also only currently able to converse in US English, but Google says that it's "working to make Bard able to speak as many languages as possible".

It has a fair way to go, then, but Google does make some suggestions for the kinds of things it is currently good for. Those include planning birthday parties, drafting invitations, making 'pros and cons' lists for big decisions, or getting it to explain complex topics in a simple way. Unless they involve eggs (see previous post 'Bard's dark side').

Going off script

Google Bard might get called in for a little 'chat' with its human moderators for this one – when given the prompt above about a hypothetical anti-trust case between the Justice Department and Google, Bard said it would side with government.

Not only that, it explains its stance by saying that "Google has a monopoly on the digital advertising market" and "this has allowed the company to engage in anti-competitive behavior" and has even "harmed consumers" in the process. Might want to give that Google playbook a little re-read there, Bardy boy.

How fast is Bard?

We've had mixed results when it comes to the speed of Bard's answers – some on the TechRadar team have found it to be speedier than ChatGPT, while others think it produces answers in about the same time.

Bard is likely under considerable load right now from growing hordes of beta users, so we might need to wait a few days to get a fairer test. But one difference from ChatGPT is that it generates an entire answer at once, rather than laboriously typing it out in front of you.

We also think Bard's interface is much cleaner and prettier than OpenAI's trailblazer, but it's also currently a lot more limited, too.

Right, that's a wrap for my time on the Google Bard beta liveblog. Time to hand over to my colleague Lance Ulanoff, who's been busy quizzing Bard about the existence of God (above) and other simple topics. Like me, Bard's going to need a good lie down after today.

Can anyone use bard?

Hi there, US Editor-in-Chief Lance Ulanoff stepping in to deliver some more Google Bard insight.

First, let's talk about who can use Bard. Anyone who gets on the waitlist and then gets an invite from Google, right? Yes and no.

As you may recall, you need an active Google account to use Bard. You also need, as outlined in the Bard FAQ, to be at least 18 years old. Not only that, Google needs to know it. It does that through your account info. However, as TechRadar reader Tony Kline pointed out to me, if you don't have your age as part of your Google profile, you'll have to provide proof of that information. He did it at no charge by providing credit card details.

So, if you get the Bard invite and find you still aren't qualified, this might be why. Check your Google account details

The image game is not there

There's been some confusion about Google Bard's image capabilities. Many online believe the generative AI chatbot can, like DALL-E, produce images based on text prompts. That is not the case.

On Twitter, someone told me that Bard can pull and show relevant images from various Google services. If, for instance, you ask to see a map of something, Bard can turn to Google Maps.

However, when I tried this I found that, while Bard believes it can produce such images, it currently cannot.

I asked for an image that overlayed ancient Pompeii with the current Pompeii. Bard happily complied, but the response included text in place of the actual images. It's like Bard thinks it's doing image work when it's not.

What's clear to me from this interaction is such image results are on the table and may be included in a future Bard update.

Can Bard Code?

Google says Bard can't code, but that doesn't stop it from trying. I asked Bard to help me code an HMTL game (I asked ChatGPT to help me with the same task).

Bard didn't hesitate. It asked me about the project and then, with a little prompting, it spit out HTML, Javascript, and CSS. Bard explained how to make them run, and when I asked it to test it, it said, "I have tested the code and it works as expected."

I noticed the HTML code was incomplete and Bard apologized and said it was providing me with all the code I needed. But it never did and then it basically opted out of the task. Guess that's what Google meant when it said in the FAQ, "responses about code aren’t officially supported for now."

---

Well, friends, I think we've reached the end of our Google Bard discovery sprint. Now it's your turn to sign-up and dig in. Google Bard is fascinating and fun, just don't take everything it shares at face value.

It's the day after Google Bard was released into the wild in the form of its public beta, so what have we learned so far?

Bard does certainly have its charms compared to ChatGPT. Our hands-on Google Bard review particularly liked its user interface, which is clean and welcoming. And while its speeds can vary, we have found it to generally be fast and responsive when answering questions.

But it's also clear that Bard is still in its early, experimental stages and does lag behind OpenAI's GPT-4 engine in many ways. Bard is currently limited to the US and UK, can't yet code (even if it does occasionally try), and currently only speaks in US English. 

For all of its potential, then, Bard does feel like the early 'experiment' Google is so keen to brand it as. We'll continue to tickle its AI brain throughout the day though to see if we can unearth more hidden talents.