Readers like you help support Android Police. When you make a purchase using links on our site, we may earn an affiliate commission. Read More.

There’s been a lot of chatter around artificial intelligence lately, particularly after OpenAI debuted its revolutionary ChatGPT service that Microsoft is now looking to integrate into Office 365 and Bing search. Meanwhile, Google has been a self-professed “AI-first” company since announcing a shift in focus at I/O 2017, and more recently, it unveiled a new AI-powered chatbot called Bard. But in a sea of buzzwords and initialisms, it can be hard to wrap your head around what these new tools actually do.

Google Search already uses AI to understand colloquial language and power tools like Google Lens and Google Assistant, so you might be wondering how Bard is any different. The key is in Bard’s conversationality and ability to answer questions — but there’s much more to it, so let’s dive in.

What is Bard and where did it come from?

In the simplest terms, Bard is a generative AI — this is the generic name for AI models like ChatGPT and DALL-E that can create new content. Generative AIs can make video, audio, and imagery, but as an AI chatbot, Bard is focused on creating text — specifically, text that answers your questions in a natural, conversational way.

Another umbrella term that describes Bard is large language model (LLM). This is a type of neural network that has been fed huge amounts of text in order to teach it how to process natural language. The training data used here is one limiting factor, and a big reason why ChatGPT thinks it's still 2021 — that's when the AI learned to talk, so to speak. Bard might suffer from similar issues under the hood, but it makes up for this with its Google Search integration, which gives it data on current events in addition to its base LLM training.

Considering the timing, Bard may seem like a product that was rushed out the door to compete with the release of ChatGPT 4. But interestingly, Google actually laid the groundwork for ChatGPT and other LLMs by making its Transformer deep learning model available to the public in 2017. Bard is powered by Google's Language Model for Dialogue Applications (LaMDA), which was released in 2021. So OpenAI’s new tool shares a common lineage with Google’s, but Bard itself has been years in the making.

An animation of Google Lamda
Source: Google

Bard is based on LaMDA, a conversational AI model Google debuted in 2021

What does Google Bard's name mean?

Bard gets its name from the word meaning “poet” — as in the Bard of Avalon, William Shakespeare — in a reference to its linguistic capabilities.

The name also has geeky roots, so it fits perfectly with Google's naming conventions. Bards are a type of playable character in Dungeons and Dragons, and they served as inspiration for the old Apple II game called The Bard's Tale, which itself spawned a series of games that includes a remastered 2018 edition of the original.

How does Bard work?

Google wants Bard to supplement the Knowledge Graph Cards you see in Search when making queries that have a simple answer. While a Knowledge Graph Card can supply you with a word’s definition or an overview about a person or place, Bard responses are meant to address NORA questions, as Google calls them – searches with No One Right Answer.

To do this, Bard first uses LaMDA language models to understand your question and its context. Since LaMDA uses datasets that contain dialogue, it understands nuance and colloquialisms that search engines struggle with. After that, Bard draws on information it finds across the web to form an answer, which is then made into the type of conversational reply you might expect from a real person (again, thanks to LaMDA). Bard's goal, and all AI chatbots for that matter, is to provide high quality responses.

A demonstration of Google Bard

Google wants you to use this tool to further your understanding of topics and help make decisions. During a demonstration in Paris, the company asked the AI chatbot to help decide which car to buy, then asked follow-up questions about the advantages of electric vehicles. Such prowess could negate the need to click search results, but Google is being careful to maintain its relationships with websites and content creators. Senior Vice President Prabhakar Raghavan had this to say on the matter:

As we scale these new, generative AI features in our search results, we continue to prioritize the process that will allow us to send valuable traffic to a wide range of creators and support a healthy, open web.

In its current form, Bard is a standalone utility rather than a function of Google Search. You can open a chat with the AI, and the things you say during a single session will be remembered, allowing you to use the context of earlier questions to get more refined answers from Bard. You can even press the chatbot for clarity when the information it gives you seems off, and it will correct itself and apologize.

Bard, ChatGPT-4, and other AI chatbots may provide quick answers, but they may not be correct. These platforms provide answers that are influenced by biases in training datasets, conflicting or outdated information, or — in the case of CNET's Wordsmith — facts the AI makes up. Google has a lot of experience with sorting out fact from fiction and determining when to surface safe opinions — that is, opinions and theories that are mostly accepted as fact, and surfacing the most accurate information available at the time. Google's Featured Snippets, for example, answers queries by collecting data from multiple online sources, determine a midrange by using machine learning models to quantify sentiment and other variables, filter out answers that fall outside the midrange, and present a response the precedes its search responses.

When can I use Bard?

Google Bard is still in its infancy, so it's only being made available to a limited amount of users as an experimental test version at this time. If you want to try it, you can sign up for early access on Google's website — just know there is a waitlist.

During testing, Bard is utilizing a lightweight model version of LaMDA, which has allowed Google to make the prerelease version of the tool available to more users. The company aims to use this testing phase to fine-tune Bard’s accuracy, quality, and speed.

An example of how Bard answers questions

The experimental version of Bard is a standalone utility, but the tool will eventually be integrated into Google Search

Currently, there are some quirks and limitations to be aware of, so once you get your invite to use the new tool, make sure to take some time and learn how to use Google Bard to find the information you're looking for.

Once Bard has finished with its testing phases, it should ultimately be integrated into Google Search. At that point, using the feature then should be as simple as typing any query into the search bar — you’ll notice things are different when Google gives you a thorough answer in plain English instead of a card and a list of links.