"Catch ChatGPT" Google Conducts AI Chatbot Testing

Google booth at CES 2023, the world's largest electronics and IT exhibition [Image source=Yonhap News]

[Asia Economy Reporter Chunhee Lee] As OpenAI's 'ChatGPT' is gaining worldwide attention, American financial media CNBC reported on the 31st of last month (local time), citing sources, that Google is testing an artificial intelligence (AI) chatbot to compete with it.

According to the report, Google is conducting the 'Atlas' project to test the chatbot 'Apprentice Bard' using its language program 'LaMDA'. It is also reported that methods and designs for integrating the Apprentice Bard chatbot into Google's search engine are being tested.

This test by Google is analyzed to be a response to the sense of crisis felt as OpenAI's ChatGPT, supported by Microsoft, rapidly rises.

The Apprentice Bard chatbot is expected to have functions similar to ChatGPT. When a question is entered into the chat box, it provides answers in text, and employees give feedback on these answers as part of the testing process.

Unlike ChatGPT, Apprentice Bard was able to answer questions about recent events. For example, when asked "Will there be more layoffs at Google?" Apprentice Bard replied, "According to my database, the likelihood of Google conducting another layoff this year is low." This response reflects Google's layoff of 12,000 employees, which is 6% of its total workforce, earlier last month. The chatbot added, "Layoffs are generally conducted for cost reduction and restructuring, but Google's financial condition is stable," and "In fact, Google's revenue increased by 34% in 2021, and its stock price has risen 70% since January last year."

Methods to apply the Apprentice Bard chatbot to Google's search page for use in a question-and-answer format are also being tested. Instead of the existing 'I'm feeling lucky' button directly below Google's main search bar, a method showing five different potential questions is being tried. When a question is entered, a gray speech bubble providing an answer similar to a human conversation appears just below the search results, and below that, related questions to the first question are recommended.