Unveiling Google Bard: The AI Chatbot Race And Unintended Leak

Nowadays a lot of big companies are trying to bite a piece of cake in the AI field. That’s why they’re developing new features, tools and chatbots, and Google isn’t an exception. However, as we delve into this AI-driven narrative, we uncover unintended consequences, privacy challenges, and a frenzied competition for AI supremacy among tech giants. Join us on this journey to unravel the potential and perils of AI advancement through the lens of Google's Bard.

Everything You Need to Know About Google Bard

Google unveiled Bard, their own AI chatbot, in response to the popularity of OpenAI's ChatGPT. The main distinction between it and ChatGPT is that the information used by Google's service will come from the Internet. Bard can code, solve math problems, and assist you with your writing needs, just like the majority of AI chatbots. Even though Bard was a completely novel idea, Google's Language Model for Dialogue Applications (LaMDA), which had been released two years earlier, was the technology behind the AI chat service when it was first introduced. On March 21, 2023, a little more than a month later, Google Bard was made available.

‍

‍

To assist you in discovering the information you need, Google's Bard AI chatbot can now search through your Gmail, Docs, and Drive in addition to the web. With the new integration, you can ask Bard to perform tasks like locating and summarising an email's contents or even highlighting the key passages of a document you have saved in Drive. These integrations, which Google refers to as extensions, have a wide variety of uses, but they should spare you from having to dig through a ton of emails or documents to get a certain piece of information. Bard can then utilize the data in other ways, for as by adding it to a chart or developing a summary with bullet points.

According to Jack Krawczyk, product lead at Bard, to use the feature, you can either have Bard search directly within your Gmail by, for instance, prefixing your query with @mail. Or, you may just say, "Check my email for information related to my upcoming flight." Additionally, Bard's extensions go beyond Gmail, Docs, and Drive. Additionally, Google revealed that the chatbot will link to Maps, YouTube, and Google Flights. As a result, you can now ask Bard to identify nearby attractions, pull up real-time travel information, surface YouTube videos on a particular subject, and much more. Since this is the first instance of a language model product fully connecting with your data, Krawczyk explains, "We want to be certain that we do that correctly." According to Krawczyk, Google intends to increase Bard's integrations with "products across Google as well as partners outside of Google."  Although granting Bard access to your private emails and documents may cause you to worry about data usage and privacy, Google assures you that neither human reviewers nor the public model's training process will see this material. For sure, what can go wrong?

Google’s Bard AI Talks Were Inadvertently Appearing in Public Search Results

After presenting the new Bard functions, it appeared, Google leaked chats into its search results. It all started when Gagan Ghotra, an SEO expert, initially brought up the concern that URLs pointing to user discussions with Bard were appearing in Google's search index, the repository of websites that the search engine scrapes to respond to users' searches. 

‍

Ghotra published a screenshot of users' private exchanges with Bard in which they asked for advice on where to look for otters in Singapore and how to utilize Bard to sharpen their writing. Users who clicked on such links were directed to the archived transcript of a discussion another user had with Bard on the subject. Peter Liu, a research scientist at Google DeepMind, soon emphasized that chat discussions that were previously shared with another user via the Share feature common to AI chatbots—which enables users to offer connections to individuals of their choosing—were shown in search results.

The problem is made worse by the fact that chats are now included in search results. This is in contrast to previous practice. According to Margaret Mitchell of the AI company Hugging Face, "this is very different from what Google does with products like Google Docs and Google Drive," who notes that on those other products, for Enterprise users, share options will warn you about the risks of sharing content with people outside your organization's domain name. She believes that users will find the apparent contradiction between what Google has done in the past and what it appears to be doing currently puzzling.

‍

‍

Conclusion

In the end, the problem is so serious because users have made AI chatbots into the confessional of the twenty-first century. A recent study demonstrates that users are prepared to provide ChatGPT with their personal information, and this week Lilian Weng, a member of OpenAI's AI safety team, said that users might want to consider using her chatbot as an alternative to a therapist. Willison praises Google for taking swift action to address the security hole, but she also points to a larger problem with the competition among tech giants like Google, Microsoft, OpenAI, and several other firms for AI dominance. "It's a good illustration of how all three companies are moving at breakneck speed to compete with each other," he says, "which makes it more likely that mistakes like this one will slip through to production."