Bard is Google's answer to the AI explosion rapidly becoming the most disruptive technology of the century. Generative AI chatbots are changing how we interact with technology, and the tech giants are rushing to claim a slice of the pie.

It's fair to say that the first iteration of Google Bard was given a lukewarm reception. However, this is a rapidly evolving technology, and Google announced a whole raft of updates during Google I/O 2023.

Let's see how these improvements perform as we check out Google Bard vs. ChatGPT.

ChatGPT vs. Google Bard: Comparing the Underlying Technology

Both rely on large language models (LLMs) to generate their "human-like" responses.

A simple definition for this complex topic is that LLMs are machine learning models designed to perform natural language processing (NLP) tasks.

Google Bard and ChatGPT run on different LLMs:

  • GPT-3.5 (ChatGPT): The GPT-3.5 LLM is the catalyst that kick-started generative AI chatbots. One issue with this model is the age of the dataset. Although it has received some topical updates, the model's dataset is limited to 2021.
  • GPT-4 (ChatGPT Plus): The premium version of OpenAI's LLM includes a more recent dataset, access to live information, and image processing capabilities. As well as the OpenAI version, you can also access GPT-4 through its integration with various apps.
  • PaLM 2 (Google Bard): The first iteration of Bard used the LaMDA LLM. However, the latest version uses PaLM 2. According to Google, PaLM 2 is a substantial upgrade that offers better creative writing, coding skills, math problem-solving, and support for Japanese and Korean languages, with 40 more to follow.

Let's look at both models in action and analyze their performance.

Testing and Comparing ChatGPT and Google Bard

To test the tools, we devised an exercise to compare their performance across various metrics. But the first step in the Google Bard vs. ChatGPT head-to-head has to be getting the tools running on your machine.

  • ChatGPT: You can access ChatGPT via the OpenAI official website. You only need to create an account on OpenAI to get started. As noted, the free version is restricted to the older GPT-3.5 LLM.
  • Google Bard: Bard it is still only available in certain countries. Exclusions include EU countries and Canada, not an insubstantial omission. You can sign in with an existing Gmail account or create a new one on the Google Bard official website.

Signing up for both tools is simple, although the Google Bard country restriction is inconvenient. However, it may be possible to circumvent these restrictions using a VPN.

Comparing Google Bard and ChatGPT-3.5's Accuracy

Both chatbots are quick to point out that the accuracy of their responses should not be taken at face value. A point proved by asking Google Bard a simple question that it shouldn't have any problems with: "What large language model is Google Bard using?"

Screenshot of Google Bard answering LLM question

As you can see, this is a dated response that justifies the statement that response accuracy should always be double-checked. But a couple of other recent updates are apparent in the response too.

The first is the inclusion of dark mode, which was omitted from the first iteration. The second is the inclusion of a Google it button and suggested follow on questions.

Screenshot of Google Bard Google button and follow on prompts

These are both welcome updates to the tool.

Now let's see how it compares with the free version of ChatGPT. We asked both tools a simple question: "How many fans attended the 2022 World Cup in Qatar?"

Google Bard's response gave us precise figures and added some other attendance stats.

Screenshot of Google Bard answering world cup attendance question

ChatGPT was unable to provide up-to-date information.

Screenshot of ChatGPT responding to world cup attendance question

Next, we tried to test on non-topical data. We asked both chatbots to answer this question: "How long would it take to drive to the sun?"

Google Bard first; it answered with some comprehensive details.

Screenshot of Google Bard answering how long to drive to the sun

We were quite disappointed with the ChatGPT response. This should have been recognized as a hypothetical question, with the result only requiring some basic arithmetic.

Screenshot of ChatGPTs answer to driving to sun question

To give ChatGPT a second chance, we reworded the question and asked: "How long would it take me to travel to the sun at 65mph?"

Screenshot of ChatGPT response to reworded question

This time it did produce some meaningful figures, but it had rounded the figure up to 1.5 million hours, causing a disparity of eight years between the responses. However, this disparity could be due to the distance between the Earth and the Sun.

Testing Google Bard vs. ChatGPT-4

Dated information is a known issue with the free ChatGPT version. So let's level the playing field slightly and test Google Bard vs. ChatGPT for accuracy by testing Bard against the more advanced GPT-4 LLM.

To test them, we asked a simple question that relies on topical data for an accurate answer: "Who is Twitter's new CEO, and what is her history?"

First up was Google Bard. It correctly reported that Linda Yaccarino had been appointed CEO and accurately summed up her career to date.

Screenshot of Google Bard answering Twitter CEO question

Despite apologizing for the lack of detail, the premium version of ChatGPT was equally impressive. However, it is worth reiterating that if you are using the free version of ChatGPT, this question would be met with a blank stare!

Screenshot of GPT-4 answering Twitter question

When comparing Google Bard vs. ChatGPT for accuracy, the Bard was the better performer compared to the free version of ChatGPT. If you are using the premium version of ChatGPT, there isn't much separating them. However, as with any chatbot, a pinch of salt is required when determining the accuracy of results.

Comparing Google Bard's and ChatGPT's Creativity

Testing these tools for the quality of their creative output is trickier. There is no sliding scale for creativity that allows direct comparisons to be drawn. So, we will set an identical creative task and make a personal judgment on how the answers compare.

We asked both tools to compose a short poem about chatbots in the style of William McGonagall (a Scottish poet renowned for lousy poetry).

The original version of Google Bard gave us this:

Screenshot of Google Bard Poetry

When compared to the ChatGPT-3.5 answer below, this is flat and uninspiring.

Screenshot of ChatGPT Poetry answer

Although this is not an extensive test, this example shows what we found to be the case overall—ChatGPT is better at creative tasks than Google Bard.

At least, this is the case for the early version of Bard, which was what we used for the original test. The new Google Bard is trained on a larger text dataset, which Google says will give it more creativity. To put this to the test, we asked it to create the same poem:

Screenshot of latest Google Bard poetry attempt

As previously noted, measuring creativity is a judgment call, but for us, the new poem still comes up short compared to the ChatGPT effort. ChatGPT still appears to have the edge in the creativity stakes. But what about ChatGPT-4? Let's give it a bash at the same poem and see how the premium version compares.

Screenshot of GPT-4 composing poetry

Once again,

It seems to me,

That ChatGPT is the place to be,

When it comes to creativity!

ChatGPT vs. Google Bard: Which Is the Best AI Chatbot?

The different LLMs that power the tools are fundamental to how the tools perform. They are the key reason for the differences in the test responses, each having its own strengths and weaknesses. Choosing between them will be determined more by user intent than user preference.

Some of the deciding factors include:

  • Accuracy: If you are looking for factual and real-time information, the outright winner must be Google Bard. It is also easier to fact-check information with Google Bard, thanks to its direct links to Google Search. The inclusion of links to cited articles is one of the welcome new features to Bard and is also available when using GPT-4 based chatbots.
  • Creativity: Despite the upgrade to Bard, we still found ChatGPT was able to produce more creative answers. If you are looking for writing suggestions and assistance, ChatGPT is the better performer.
  • Security and Safety: There have been concerns about the online safety of ChatGPT, which uses a reactive approach to security. In other words, safety concerns are dealt with as they arise. Google Bard does have a more proactive approach but still uses feedback to improve processes. If you have safety concerns, there are plenty of ways your kids can use ChatGPT safely.
  • Ease of Use: Both tools are easy to use, and how you interact with them is similar. The inclusion of follow-on suggestions and links to cited sources are definite improvements to Google Bard. These make checking the integrity of results easier with Bard and make ongoing research easier.
  • Integrations: One of the key benefits of Google Bard is its integration with tools like Gmail and Google Workspace. This means that you can export chats directly into platforms like Gmail and Google Docs.
    Screenshot of Google Bart export options

Summing up, if you are looking for accuracy and simplified integration with other Google services, then Google Bard is the better option. It also outperforms ChatGPT in terms of interface, with the addition of a Google it button and source-citing, making it the better all-round package.

The integration Bard offers is crucial here, Gmail has 1.8 billion users, and this functionality will be a major draw for many when choosing between chatbots.

These tools excel in different areas, for business and practical uses, and there is no doubt that Google Bard is the best option. However, we still found that ChatGPT edged it in creative tasks. And, although clumsier, there are apps and extensions to export ChatGPT chats. This does slightly negate the advantage that Bard has with integrations.

Google Bard is free, and there is a free version of ChatGPT, so testing both tools is easy to do.

Tools With Great Features, Yet Still Imperfect

There are plenty of rough edges to both platforms, which is to be expected. This is a new technology, and both are examples of live-testing programs that make mistakes in public. The other side of the coin is that such widespread usage and the feedback it generates will hasten the development of both models. A point amply demonstrated by the host of new features added to Bard.

We are at the start of a technological revolution that is so disruptive that many of its impacts are yet to be determined despite much speculation. But there is no doubt that tools like Bard and ChatGPT are going to transform much of how we work, rest, and play.