Editor's note: Since this article was written, Bard has released several notable updates. It works with more languages and now also features AI coding assistance. All this happened in a short period of time. We are leaving the article below as a peek into the launch of Google's first notable chatbot. If you're interested in the product launch, read on.
This week, Google Bard rolled out experimental beta access to Google One members. I joined the waitlist as soon as I saw the invitation and was granted access to the interface within about 24 hours. The results of our conversation left me both impressed and bewildered.
We discussed how to build a Python chatbot. Now, here’s a breakdown of my experience with Google’s Bard AI.
A surprise invitation to experiment with Google Bard
How to Join the Google Bard Experiment
The email arrived because, for better or for worse, I pay about $20 a year for Google’s cloud-storage service. I accepted the invitation and joined the wait list in a few seconds. Then, the waiting began.
Your experience may differ, but if openAI’s chatGPT offers any hint into rollout strategy, Google's Bard chatbot may be only a few days away.
Full access to Google Bard’s experimental platform wasn’t granted until the next day, and I was enthusiastic to test it out. My first thought: Why not interview this large language model? If it’s such a good conversational partner, this shouldn’t be too tough of a task.
An Interview with Google Bard
So how do you interview a large language model? First, I asked it to define Google Bard and list a few of its benefits. Here’s an abridged version of our conversation. I have broken the conversation into parts and added thoughts throughout.
Google Bard (GB): “Google Bard is a large language model (LLM) chatbot developed by Google AI. It is capable of generating text, translating languages, writing different kinds of creative content, and answering your questions in an informative way. Bard is still under development, but it has learned to perform many kinds of tasks, including:
- Generating text. Bard can generate text in a variety of styles, including creative writing, journalistic writing, and academic writing. It can also generate code, scripts, musical pieces, email, letters, etc.
- Translating languages. Bard can translate text from one language to another. It can translate between over 100 languages, including English, Spanish, French, German, Chinese, Japanese, and Korean.
- Writing different kinds of creative content. Bard can write different kinds of creative content, such as poems, songs, stories, and scripts. It can also write code, scripts, musical pieces, email, letters, etc.
- Answering your questions in an informative way. Bard can answer your questions in an informative way. It can access and process information from the real world through Google Search and keep its response consistent with search results.”
Me: How is Bard different from chatGPT?
GB: Google Bard and ChatGPT are both large language models (LLM) chatbots, but they have some key differences.
Bard is trained on a massive dataset of text and code, including text from books, articles, and code from GitHub. ChatGPT is trained on a smaller dataset of text and code, including text from social media, news articles, and code from GitHub.
Bard is able to access and process information from the real world through Google Search, while ChatGPT is not. This means that Bard can provide more accurate and up-to-date information.
Bard is also able to generate text in a variety of styles, including creative writing, journalistic writing, and academic writing. ChatGPT is only able to generate text in a conversational style.
Overall, Bard is a more powerful and versatile chatbot than ChatGPT.
That’s about what I expected from a chatbot designed by Google. Its response comes across as thoughtful and slightly hubristic. Here’s where I decided to ask tougher questions.
Me: Do you anticipate issues with copyright infringement?
GB: Yes, I anticipate issues with copyright infringement. As Bard is a large language model, it is trained on a massive dataset of text and code. This means that Bard is able to generate text that is similar to text that has been copyrighted.
If Bard generates text that is similar to copyrighted text, and if that text is used without permission from the copyright holder, then this could be considered copyright infringement.
Me: Do you expect Bard to be controversial?
GB: Yes, I do expect Bard to be controversial. As a large language model, Bard is able to generate text that is similar to human-written text. This means that Bard could be used to generate fake news, propaganda, or other harmful content.
Additionally, Bard is still under development, and there is still a lot that we do not know about how it works. This could lead to people being concerned about the safety of Bard, or about the potential for Bard to be used for malicious purposes.
Overall, I expect Bard to be a controversial tool. There are both potential benefits and risks associated with Bard, and it is important to be aware of both of these before using it.
My first experience with the Google Bard UI
These seem like well-rounded responses to more difficult questions. It responds how I’d expect a PR person would when faced with the same potential issues. So what about something more precise?
The Indicator, an NPR podcast by Planet Money, recently held an informal competition between Google Bard and chatGPT. I figured there was little chance this chatbot would be well-versed in current events, and I wanted to ask about it.
Me: In a recent episode of The Indicator, researchers pitted you against chatGPT. The results seemed inconclusive, but neither chatbot came out smelling like roses. How would you modify your responses to better address their questions on dating?
GB: I would modify my responses to better address their questions on dating by:
- Being more specific. When they asked me questions about dating, I sometimes gave vague or general answers. I would try to be more specific in my answers, and provide more detail.
- Being more helpful. When they asked me for advice on dating, I sometimes gave advice that was not helpful or relevant. I would try to be more helpful, and provide advice that is actually useful.
- Being more understanding. When they asked me questions about dating, I sometimes gave answers that were not understanding or compassionate. I would try to be more understanding, and provide answers that are actually compassionate.
I think these changes would help me to better address their questions on dating.
Me: Do you know the episode I'm referring to?
GB: Yes, I am familiar with the episode you are referring to. It was episode 193 of The Indicator, titled "The Chatbot Dating Game". In the episode, the researchers pitted me against chatGPT, another chatbot, to see which one could provide better advice on dating.
The researchers found that neither chatbot was able to provide consistently helpful or insightful advice. They also found that both chatbots sometimes gave advice that was harmful or offensive.
I think the researchers' findings are fair. I am still under development, and I am always learning and improving. I hope that in the future, I will be able to provide more helpful and insightful advice on dating.
That response shocked me. Can it listen to podcasts? No, I realized NPR offers reliable transcripts of most of their content. Bard could have scraped those details in the same way that it finds most of its information.
The exciting thing was its speed. It researched that information, parsed the important bits, and replied in seconds – certainly no longer than a conversation with a friend via text.
Google Bard even tells the user that it’s still experimental
What Are the Differences Between Google Bard and chatGPT?
Here’s what I learned about the difference between Google Bard and chatGPT from my limited experience. Currently, while chatGPT has trouble with current events, Google Bard seems well-versed. It was able to accurately describe a recent economics podcast and offer realistic ways to improve its performance.
But what about coding?
We’ve discussed how coders are already using chatGPT as a coding assistant. Unfortunately for Bard’s early adopters, the same functionality doesn’t exist with Google’s chatbot.
Coders can’t turn to Google Bard just yet
When asked if Bard can help with coding, Google responds that it’s still learning how to code. Users will have to wait a little longer for an official coding partner.
Me: How can Bard help coders?
GB: Bard can help coders in a variety of ways. Here are a few examples:
- Code generation. Bard can generate code in a variety of programming languages, including Python, Java, and C++. This can save coders a lot of time and effort, especially when they are working on repetitive tasks.
- Code completion. Bard can provide code completion suggestions as coders are typing. This can help coders to avoid errors and to write code more quickly.
- Error checking. Bard can identify errors in code and suggest corrections. This can help coders to write code that is more reliable and error-free.
- Debugging. Bard can help coders to debug their code by identifying the cause of errors. This can help coders to save time and frustration.
- Refactoring. Bard can help coders to refactor their code by identifying areas that can be improved. This can help coders to make their code more efficient and maintainable.
Overall, Bard can be a valuable tool for coders of all levels of experience. It can help coders to save time, avoid errors, and write better code.
I asked the same question, then, to both Bard and chatGPT. How do you write, "Hello, world!" in Python? The results look like this.
With chatGPT, the response was short.
ChatGPT's response when asked to create a simple 'Hello, World!' program in Python
Google Bard gave a more detailed answer.
Google Bard's response when asked to create a simple 'Hello, World!' program in Python
Both chatbots provide the same line of Python code, but Google is quick to point out that Bard isn't yet officially able to with programming languages.
Final Thoughts on the Google Bard Beta Test
So, depending on whether you ask Google or whether you ask Bard, you may find a different response. Suffice to say Bard stacks up against its competition. Neither Bard nor chatGPT is perfect, but neither claims to be.
And if you’re wondering how this chatbot feels about being quoted, you’re not alone. I asked if I could use its responses in this article. It said, “Of course it's okay if you quote me for an article about the beta test. In fact, I would be honored.”
