Chatbot AI are dominating the news cycle and two of the big contenders for the top chatbot for are ChatGPT, owned by Microsoft-backed OpenAI, and Google Bard.
ChatGPT was first but for now, it still reigns supreme. The OpenAI chatbot launched as a research preview back in late 2022 and started the current chatbot craze with its ability to do a wide range of tasks from writing essays to coding websites and even more things we're still learning it can do.
Since then, both Microsoft and Google have started a chatbot AI arms race. In the course of one week, we saw both sides launch competing ChatGPT competitors. First, Microsoft announced “the new Bing” powered by ChatGPT’s underlying technology — the GPT language model.
Then Google responded the next day at a Paris event with Google Bard, an AI chatbot that uses a different language model called Language Model for Dialogue Applications (LaMDA) and at first glance appears more search-focused like the new Bing with ChatGPT rather than ChatGPT's more generalist skill set despite being separate from Google Search. Google has now opened up access to the public, but you'll first need to get on the waitlist for Google Bard access.
Still, we are going to do our best to answer the big questions about these competing chatbot AI. Which one should you use? Is one of these AI better at certain tasks than the other? Is one more accurate? Or are they essentially the same but one is branded Microsoft and the other is branded Google? Below we take a look at all of these questions and more.
ChatGPT vs Google Bard: Overview
|Row 0 - Cell 0||ChatGPT||Google Bard|
|Price||Free research preview; $20 optional subscription plan called ChatGPT Plus||Free|
|Availablity||Widely available through web browser||A waitlist is now open to anyone|
|Features||Content creation, coding tool, research tool and more||Search queries, research tool|
|Major integrations||Bing, Snapchat, Slack and more||None|
|Do they cite sources by default||No||No|
|Row 6 - Cell 0||Row 6 - Cell 1||Row 6 - Cell 2|
ChatGPT vs Bard: Price and availability
Currently, there are free models of both ChatGPT and Google Bard. ChatGPT is currently in a free research preview on OpenAI (opens in new tab)’s site, though there is now a paid subscription model called ChatGPT Plus where users can pay $20 for priority access and faster speeds. ChatGPT Plus is limited to those who OpenAI approves for access, so you’ll need to sign up for the waitlist to get access.
While Bard was previously only open to "trusted testers", Google has reversed course and there is now a waitlist anyone can join for access to the search giant's AI chatbot. We have received access but are still testing features. We will update this story with anything pertinent that we discover in our testing.
ChatGPT vs Bard: Features
The features of these two chatbots are, at their core, relatively similar. Both require you to input a question or request and if you do, it will return an answer. You can then continue with follow-up questions or new requests, and the chatbot will continue to hold a conversation with you.
ChatGPT has a wider range of use cases though. The AI chatbot, when used in its interface on OpenAI’s site, can generate text to write news articles, fiction poetry, product descriptions, blog posts and more. It can also handle certain coding languages, allowing it to even provide the code required to build a basic website.
GPT-4 has opened up the gulf between these two AI even more. Unlike Bard, which can only handle text-based inputs and outputs, the latest version of the GPT model behind ChatGPT has multimodal capabilities. GPT-4 in ChatGPT Plus and the ChatGPT API can handle image inputs, allowing the chatbot to take an image and answer questions or provide context based on that image.
Bard, by comparison, appears to be largely an expanded version of Google search, though Google has made it separate from Search for now. Google executives have said since the Paris launch that Bard isn't just about search (opens in new tab) but the demonstration and our own testing have shown a search tool similar in function to the new Bing with ChatGPT that Microsoft launched. As we learn more about Bard's features we will update this article and our Google Bard hub accordingly.
But back to what Bard can do. Like Microsoft's Bing chatbot, Bard appears best used to research, as it takes traditional search queries and answers them in a conversational manner, providing context rather than raw results. For example, if you were to ask what the system requirements are for a game, it could tell not only you the requirements but also what those requirements enable, rather than just providing a link for you to click through to for more information. But tasks such as coding are beyond its reach.
It isn’t impossible that Bard or Bing won’t be able to handle the same types of requests that ChatGPT can handle in the future, but at the moment we have not seen such features demonstrated.
ChatGPT vs Google Bard: Integrations
Microsoft, OpenAI and Google all want their respective chatbots integrated into their respective ecosystems as fast as they can.
ChatGPT has already been integrated into a wide variety of products. ChatGPT is being integrated into the Opera web browser, and both Snapchat and Slack have received ChatGPT integrations. OpenAI also recently opened up the API for ChatGPT so hundreds of apps could soon get ChatGPT.
And this doesn't even include the Microsoft apps that have taken ChatGPT's underlying GPT model and used it in its own products. Microsoft Teams Premium has been out for a bit now and offers automated meeting notes, AI-generated chapters to search through meeting records more easily, and more, all powered by the same GPT-3.5 language model used in ChatGPT. It costs $10 a month per user, though currently, businesses can get it for just $7 a month.
Then, of course, there is the new version of the Bing search engine powered by GPT-4. This new version of Bing allows users to interact with Bing like it would the ChatGPT chatbot rather than being forced to use the traditional search bar. Check out our guide to using the new Bing with ChatGPT to learn more and if you want to sign up check out our guide to getting early access to Bing with ChatGPT.
This version of Bing has already started rolling out to Microsoft's Edge browser as well as a sidebar that can chat as well as compose content and provide insights. This feature could be useful for synthesizing large amounts of information on a webpage into more manageable summaries. It has also come to the Bing, Edge and Skype mobile apps. GPT-4 has also been integrated into Microsoft 365 Copilot, a new AI assistant Microsoft has begun rolling out for enterprise customers.
Google’s Bard AI chatbot currently has no announced integrations. The current expectation is that it will eventually be integrated into Google Search, but beyond that, we are unsure how it will be used. Google has also rolled out AI-based tools into Lens and Maps, though these are not specifically integrations of Google Bard.
However, Google did announce that it will allow for third-party developers to use Bard, so it will be interesting to see what others come up with.
ChatGPT vs Google Bard: How they cite sources
Plagiarism is a big point of contention when it comes to the use of chatbot AI like ChatGPT or Google Bard. The AI language models that power the chatbots need to be trained on existing knowledge sets in order to function, and that means feeding them tons of content created by third parties. This isn’t uncommon — humans similarly synthesize research of primary and secondary sources into original content — but when a human does this, it has to cite where it took the information from.
ChatGPT, however, does not inherently provide sources for the content it generates. It can provide sources if properly prompted, but it’s not a default setting. So you have to be incredibly careful when using the chatbot or risk stealing intellectual property unknowingly. To learn how to get ChatGPT to appropriately cite sources, check out our best tips for using ChatGPT.
Bard also doesn’t inherently provide citations for its responses. In a demo, Google asked Bard to answer “what new discoveries from the James Webb Space Telescope can I tell my 9-year-old about.” Bard provided an answer (an incorrect one, further investigation showed) but did not cite where that information came from. Likely, it was some website, but if your 9-year-old took that information and failed to cite it in a school report, they’d be in trouble. Public testing has found that Bard does occasionally cite sources, but it is rare.
Interestingly, Bing’s new GPT-powered chatbot does cite its sources. It utilizes footnotes to cite the websites where the information it pulled comes from, though you’ll need to click through for further details. While there are very real ethical concerns about this still essentially being intellectual property theft, as the sites it pulls from require traffic to make money and stay up on the internet. But at least Bing is not outright plagiarizing. That being said, you should still do your due diligence, as it is no guarantee that the chatbot managed to cite everything necessary.
ChatGPT vs Bard: Accuracy
Chatbot AI make mistakes. Regularly. It’s not all the time, but even in Google’s demo of Bard AI it incorrectly answered a question. Both Google and OpenAI admit — relatively openly — that ChatGPT and Bard can provide inaccurate or inappropriate information.
This largely comes down to how these chatbots work. They utilize language models — GPT-3.5 or GPT-4 for ChatGPT and LaMDA for Bard — and these language models require vast amounts of information to function. In the cases of GPT and LaMDA, a lot of this information can come from the internet, though it needs to then be fed into the AI via training the language model.
In GPT-3.5’s case, this training was only done till 2021 when Open AI stopped training the language model. Bing’s version of GPT is more up-to-date, as it pulls current information, theoretically similar to Bard AI. Bing also technically is still a search function so it actively searches the internet for answers. Bard seems to do this at times as well.
There are issues that come with this training, namely that the information it pulls can be incorrect or biased and the chatbot is not inherently trained to know. The chatbots are only trained to provide outputs associated with inputs, they cannot determine whether or not that information is providing a correct answer or if the answer it is providing is absent of inherent biases.
Let’s go back to the example from Google’s demo: “what new discoveries from the James Webb Space Telescope can I tell my 9-year-old about.” One of the responses was “JWST took the very first picture of a planet outside of our own solar system.” However, research through Google Search showed that, according to NASA, in 2004 the Very Large Telescope (VLT) took the first picture of a planet outside our solar system. While the JWST did, in fact, recently take its first photo of an exoplanet, that is still an incorrect answer.
Yes, ChatGPT is amazing and impressive. No, @OpenAI has not come close to addressing the problem of bias. Filters appear to be bypassed with simple tricks, and superficially masked. And what is lurking inside is egregious. @Abebab @samatw racism, sexism. pic.twitter.com/V4fw1fY9dYDecember 4, 2022
Additionally, inherent bias is incredibly important, and often something that these machine learning-trained AI are oblivious to. An example of this is when UC Berkeley professor Steven Piantadosi asked ChatGPT to write a sequence of code that filtered out good scientists from bad scientists. The AI ended up filtering out the scientists based on their race and gender even though it had not been asked to do so specifically. While the AI itself isn’t racist or sexist inherently, because of the inherent biases in the data it learned, it picked up those same biases without knowing that it was doing so and produced biased outputs. So be careful and verify what ChatGPT is providing you.
ChatGPT vs Google Bard: Outlook
At the moment, these AI chatbots are still incredibly new. That means as impressive as they are, they still make mistakes, produce content that crosses ethical and potentially even legal lines and are still figuring out their best use cases. Still, the technology is incredibly impressive and could revolutionize how we do a wide variety of tasks.
If you want to test out the capabilities of these chatbot AI, ChatGPT Plus is the best version available to the public at the moment, but it also costs money. To sign up for Plus, create a ChatGPT account and then once you are logged in click the option on the sidebar menu that says "Upgrade to Plus."
Meanwhile, ChatGPT's free research preview still allows you to do a wide variety of tasks for free, but it runs on an older GPT-3.5 model that uses potentially out-of-date information. Bing's chatbot offers more up-to-date information but offers less functionality than ChatGPT — you can check out our guide to getting early access to Bing with ChatGPT if you want to sign up.
Finally, there is Google Bard. Bard is still limited to “trusted testers” and members of the public who get approved from the waitlist. For now, it appears to be the most limited in function while still having some concerns regarding accuracy. That doesn't mean you shouldn't still check it out, just temper expectations for now. Luckily, the world of AI is changing daily, so these AI could gain functionality at any point, and many already have.