Google has a new conversational chatbot called Bard that is making waves across the tech world. It's a generative AI, which means it can create something that didn't exist before, and the thing it creates best is text. This can be text that you prompt it to generate with a few keywords, or it can be a thoughtful, complete answer to a question, based on information pulled from various corners of the internet. But as smart as Bard can be, it's still a tool, and it's up to you to know how to get the most out of the utility.

Learn the basics of Google Bard

Google Bard is a type of artificial intelligence known as a large language model, which is a fancy way of saying it's been trained with lots of different input text to better understand questions being asked of it and to form more comprehensive answers. The language model it's based on is Google's own LaMDA, which has been in the works for years. If you'd like to learn more about Bard, such as the meaning of its name and the history of its development, make sure to check our full Google Bard explainer.

An animation of Google Lamda
Source: Google

Register for Google Bard

Google Bard is currently in its early testing phases, and as such, is only available with an invitation. It's easy to register for an invite, though — just head to Google's signup page, then tap or click the Join Waitlist button. Google will email you with instructions to join the Bard beta program, but the turnaround between requesting an invite and receiving one could take anywhere from days to weeks.

Understand Google Bard's limitations

Bard tries to avoid answering anything that might spark controversy, like political feedback aside from basic info about candidates, or topics that might be considered NSFW or culturally taboo. Since Bard is bordering on science fiction, Asimov's Three Laws of Robotics are apt here: 1. A robot must not harm a human, 2. A robot must obey an order unless it violates the first law, and 3. A robot must protect its own existence as long as doing so does not conflict with the first two laws. This behavior seems to fall under Rule 3.

Bard is currently not capable of generating code for programmers, something ChatGPT excels at. This also applies to code for spreadsheets like Microsoft Excel and Google Sheets, though such capabilities could be added in a future update.

It almost outright refuses to answer medical questions — Bard generally won't even try to steer you in the right direction or provide background, preferring instead that you seek an opinion from a medical professional. Even with simple questions like "how should I treat a stubbed toe" or "what should I take for a headache," it responds that it is unable to help.

bard-will-not-give-medical-advice
Google Bard's aversion to medical questions might fall under Asimov's first AND third laws of robotics.

In its current implementation, Bard cannot pull information from your Google account. So things the Google Assistant would be helpful for, like finding nearby restaurants or controlling IoT devices from your Google Home account, are not up Bard's alley. Eventually, Google should integrate Bard into Search and Assistant and blow everyone's minds with the unique combination of personal data and the smarts to leverage it, but until then, think of Bard and Assistant as separate products.

Play to Google Bard's strengths

If it's something Google Search would give you a Knowledge Graph Card for, Bard will give you a solid answer. Examples of this are quick facts about public figures and landmarks, information about books, movies, or TV shows, and recipes or nutritional information.

If it's something mathematical or strongly fact-based in nature, Bard will be a good resource. Ask it to help with your algebra homework or an excerpt from your history or science textbook, and it will likely have an answer for you.

If it's something timely, Bard can be hit or miss, but it's much better with current events than ChatGPT. Questions like "What's the best phone to buy" can yield results that contain some of last year's phones, for instance. But if you ask Bard who is expected to be picked first overall in the NBA Draft, it will give you information about this year's draft class instead of the one from two or three years ago like ChatGPT.

google-bard-current-events
Google Bard is more up to date with current events than ChatGPT

Bard's single biggest strength, however, is with text generation. Give it some basic background, and it can flesh out a whole story for you. Feed it a few bullet points, tell it whether you want to use a professional or casual tone, and it can write your next email.

Use Google Bard to find the information you're looking for

As you're using Bard, take note of the controls available on the answers it provides. If you click View other drafts, you'll see a selection of alternative answers Bard has formed. Choosing one will give you a different result, which is particularly helpful for things like writing prompts.

You'll also see thumbs up and thumbs down buttons on responses — use these to help train Bard and let the service know when it's off base. One final control to be aware of is the Google it button — clicking this will essentially reformat your input text in the form of a Google Search query to help you find answers using web links.

google-bard-other-drafts-anim
Google Bard makes multiple drafts when answering your questions

As with any tool, there are some best practices to follow in order to get the most out of Google Bard.

First, don't be afraid to ask it any question. You're not using a search engine here, so just talk to it colloquially. You don't need proper pronunciation, minor typos are generally fine, and you can even use the wrong words sometimes — it will usually understand what you mean.

Feel free to ask follow-up questions. Each conversation in the prerelease tool is just that: a conversation. The AI remembers context from earlier in the chat, so you don't have to keep naming the subject, for example — instead, you can just say "it" or "that" in reference to the subject of an earlier query.

Finally, press for clarification. It can be wrong sometimes, so use the conversation context to make sure the chatbot is following you. If an answer seems off, follow up with "But I thought..." or "Didn't you just say..." queries, and it will correct itself or clarify.

google-bard-followup-questions
Google Bard correcting itself when pressed about conflicting answers

Google Bard currently does not have a mobile app for Android or iOS, and it's unclear if the service will ever have one. Right now, you can only access Bard through its web interface, which works well on desktop and mobile browsers alike. Ultimately, Google has said it plans to integrate Bard into Google Search, at which point it should be using the full version of LaMDA and a much larger language model. Until then, we'll just have to deal with its quirks and limitations while Google refines Bard's capabilities.