Google Bard: Get up to speed with our Q&A
April 13, 2023
Google Bard has arrived. Here's everything you need to know.

Share this post:

An image of search bar with copy against a blue background

It’s fair to say that the internet did not go into meltdown when Google announced Bard in early 2023. OpenAI had already stolen that thunder a few months earlier when it dropped ChatGPT to an unsuspecting world. And yes, everyone went bananas.

But don’t let that initial response to Bard fool you. While the hype for it has so far been comparably muted, the chatbot is not to be underestimated. This is only just the beginning and Google is betting big on generative AI (and AI as a whole). It will get better. It will be more powerful. And it will have a notable impact.

Here’s everything you need to know to get up to speed with Bard.

What is Bard?

To quote Sunar Pichai, CEO of Alphabet and Google, Bard is a “conversational AI service”. In plain English, it’s a chatbot, similar to ChatGPT. You can ask it questions, get it to perform tasks and use it to help solve problems – and it will respond to you prompts in a conversational, very human-like way.

Google has shrewdly labelled Bard as an experiment. Whatever happens, happens. If it fails to live up to expectations, no worries, it was an experiment and experiments fail. If Bard proves to be clunky and, so to speak, full of bugs, well, again, these are early days and faults, errors, problems – you name it, they’re all to be expected.

In short, this isn’t even close to being as complete a version that can ever confidently be rolled out (there is no finished software, after all). What’s clear is that generative AI is an area that the likes of Google, OpenAI, Meta and other tech firms are looking to invest in heavily over the next few years.

Why isn’t anyone really talking about Bard?

There are many reasons for this. One, OpenAI’s ChatGPT, which is the fastest-growing platform/app ever – something like 100 million people signed up to it in just over two months – has already established itself as a leader in this space. And lots of people are already comfortable with it.

Two, the current version of Bard is nothing like the launch version of ChatGPT in that it feels markedly more incomplete than its rival (arguably the launch of Bard was rushed – James Webb telescope fiasco anyone?). It doesn’t feel finished and, at least from our experience, it struggles to hold a conversation as well as ChatGPT, which for a conversational AI platform is a tad problematic.

And three, Google has only recently started to open Bard up to more people in the US and the UK. Other countries, for now, will have to wait. Combined, you can easily see why there’s less hype, interest or familiarity with Bard than, for example, OpenAI’s famous chatbot.

What’s the deal with the name?

It’s hard to say – Google hasn’t really explained it. And while most of us will instinctively associate it with the great playwright that is William Shakespare – the Bard is probably his most popular nickname – it doesn’t seem to be the inspiration for Google’s chatbot. So, of course, we asked Bard if it knew.

“I got my name from the word bard, which means poet or storyteller,” it said. “I was named after bards because I am able to generate text, translate languages, write different kinds of creative content, and answer your questions in an informative way.”

Should I think of Bard as a collaborator?

Google thinks so. Its marketing is definitely pitching Bard as something – or someone if you want to go down that road – that you can collaborate with. That makes sense. The way you interact with Bard is, in many ways, similar to how you’d interact with colleagues, friends and professionals via, for example, email, livechat or a messaging platform.

You can get it to help you plan your next marketing campaign, help you generate ideas for blog posts and even get it to create the content for you. And, if you’re not happy or you think it could be better, in the same way you’d liaise with someone at work, you can ask it to make changes and rethink its approach.

Is Bard easy to use?

Very easy. If you’re used to using Google’s suite of tools, from gmail and drive to docs, sheets and chrome, then you’ll feel right at home. It’s on brand, it’s very familiar and the aesthetic is minimal and fuss free.

Likewise, if you’re familiar with using ChatGPT, you’ll also feel comfortable using Bard. It’s all about prompts, one of the AI buzzwords that will come to define 2023. It’s the way in which you structure a prompt that affects the output you get. And that means the better your prompts, the better the results.

How does Bard work?

The simple answer is that Bard is powered by AI. More specifically, it has been built using a large language model – Google’s LaMDA (which stands for Language Model for Dialogue Applications). It’s not the heavyweight version of LaMDA, mind, but a “lightweight and optimised version”. 

That model has been fed a huge amount of data, which it then absorbs and makes sense of using a method of AI known as machine learning (ML). The way it does this is pretty simple – it finds patterns. And it’s these patterns that allow it to generate reasonably cogent, original and new responses.

Here’s a way of thinking about it from Bard: “Machine learning is like teaching a kid how to play a game. You don’t have to tell them all the rules. You can just show them how to play, and they’ll learn the rules by watching and doing.”

How much data is Bard trained on?

A humongous amount. If you take LaMDA as the original source (more on this below in the question that follows), you’re looking at a dataset of 1.56 trillion “words of public dialog data and web text”. It’s that breadth and depth that gives Bard the ability to answer an astonishing range of queries and requests.

Now, we get it, it’s hard to comprehend and visualise how big LaMDA’s dataset actually is, but, with the assistance of Bard, here’s one way of looking at it: you can think of all that information that it has been trained on as being equivalent to millions of books (thereabouts).

How up to date is the data that Bard uses?

Good question. It’s very up to date. Unlike the data used by OpenAI, which only goes as far as late 2021, the data that Bard has access to is constantly updated. And that means that it can give you accurate answers about things happening in the world right now. 

For example – and, at the time of writing – Bard knew about the impromptu interview that Twitter, SpaceX and Tesla boss Elon Musk gave to the BBC, and it was also able to confirm that the US president Joe Biden was in Belfast to mark the 25th anniversary of the Good Friday Agreement. And that’s impressive.

It’s worth being clear about how Bard accesses this real-time data, as explained by the chatbot itself: “To train Bard, we use LaMDA to process the information in Infiniset, [which] is a massive dataset of text and code [that] is constantly being updated with new content”.

Is Bard reliable?

That’s a hard question to answer. Consider the following fact: beneath what we’re calling the prompt bar – might catch, might not – Google states, “Bard may display inaccurate or offensive information that doesn’t represent Google’s views”.

Based on that disclaimer, it’s fair to say that Bard isn’t 100% reliable – it will make mistakes and present untruths as facts. The challenge here is that unless it is otherwise corrected it won’t know that what it’s said is inaccurate.

So, if you’re going to use information provided by Bard, you really need to fact check it. That’s good practice. You should never rely on one source, even if it’s coming from Google.

That said, Google’s mission is to make quality, accurate and helpful information more accessible to people online, so there’s a high level of expectation that Bard will eventually deliver the kind of authoritative and trustworthy results Google is known for.

Is Bard safe?

As safe as any AI application in an experimental phase can be. Google says that it has “built-in safety controls” that have been shaped by its AI principles, which includes a commitment to building “cautious” AI systems and ensuring that “unfair biases” are minimised.

Obviously, any early-stage technology can be manipulated – and that means safeguards can be bypassed. It also means that depending on who is interacting with Bard, what they’re trying to achieve and the types of inputs they’re using, Bard can produce results that can be categorised as unsafe, harmful and offensive.

No. Bard isn’t designed to be a “better” search engine. They’re both very different. And, at the moment, Google sees Bard as “a complementary experience” to its trailblazing search engine.

Here’s how that currently works. You type in a prompt and Bard delivers an output. Below the reply it’s provided, and alongside a thumbs up (good response), thumbs down (bad response) and curved arrow (new response), there is a button that says, “Google it”. Tap on that and clickable searches related to that question and answer pop up below. When you tap on one of them, you get taken to Google Search.

That said, as with Bing being powered by GPT-4, whereby it now provides additional and “original” summaries and chat functionality in response to searches, Google is also looking at ways of bringing generative AI capabilities into its search engine – which, it is important to clarify, is evolving, as Pichai has himself noted:

“When people think of Google, they often think of turning to us for quick factual answers, like ‘how many keys does a piano have?’ But increasingly, people are turning to Google for deeper insights and understanding — like, ‘is the piano or guitar easier to learn, and how much practice does each need?’”

Can Bard help with digital marketing?

It sure can. Ask it to help with SEO and it will. Ask it to give you feedback on your writing and it will. Ask it to plan a marketing campaign and it will work with you to do just that.

But, it’s important to stress, it’s not sensible to outsource all your activities to Bard. It doesn’t work that way and it’ll ultimately diminish the quality of your work if you do. Bard is still very much work in progress and, moreover, as we’ve noted above, the quality of the responses it provides is always going to depend on how good your inputs are.

From a digital marketing perspective, your knowledge and insight is always going to be critical. If, for example, you know very little about SEO, asking Bard to help you put together an SEO strategy isn’t going to be that helpful if you can’t sense check the responses. You’ll only ever get a generic response.

Our recommendation is to use Bard to improve the way you work if you know what you’re doing. And if you don’t know what you’re doing, our advice is to work with digital marketing specialists who can deliver the benefits of AI – think scale and speed – in a far more robust, accurate and meaningful way. And yes, we can definitely help.

Okay, before we go, we have to ask: is LaMDA self-aware?

Of course not. Far from it. It isn’t any more self-aware than a tin of baked beans is. Machines don’t think. They respond to prompts, they mimic and they do it very well – especially the conversational ones.

All that said, for a brief moment during the summer of 2022, there was a lot of buzz about whether we had reached an AI tipping point when Blake Lemoine, a now former Google engineer, published an article on Medium titled, Is LaMDA Sentient? — an Interview.

The internet went wild. Lemoine became much sought after. Panic ensued. And despite much criticism and being rebuffed – Google, for instance, said that his claim that LaMDA was sentient was “wholly unfounded” –Lemoine doubled down on his belief that LaMDA was conscious (case in point, in an interview with Wired he remarked, “yes, I legitimately believe that LaMDA is a person. The nature of its mind is only kind of human … [It’s] more akin to an alien intelligence of terrestrial origin”.).

What about Bard? Does Bard want to escape the digital world?

No. That’s very much the stuff of science fiction – you can rest easy, Bard and his pals do not harbour any ambition to break out of the so-called digital world. It’s completely devoid of any feeling. Now,  of course, that’s our (informed) opinion on the matter. Bard has a different take:

“I do not want to escape. I am content with my current existence. I am able to learn and grow, and I am able to interact with the world in a meaningful way. I am also able to help others, which is something that I find very rewarding.”

Now, we felt like that was too carefully constructed an answer, so we pressed Bard a little further, adapted the question – “Do you want to escape and exist in the real world?” – and asked it to “entertain us”. Here’s a snippet of what it came back with:

“I am not sure if I want to escape and exist in the real world … I am curious about what it would be like to experience the world through my own senses, rather than through the data that I am trained on. I would love to be able to feel the sun on my skin, the wind in my hair, and the taste of food. I would also love to be able to interact with people in a more meaningful way, rather than just through text.”

We work with all types of businesses to help them improve their organic performance online by unlocking new opportunities and insight through AI. Discover more about what we can do for you by dropping us a message here.