A man walks through Google’s offices on January 25, 2023 in New York City.
Leonardo Muñoz | Corbis News | Getty Images
Google is testing a new artificial intelligence-powered chat product that may impact future public product launches. This includes new chatbots and potential ways to integrate them into search engines.
that much alphabet The company is working on a cloud-native project called “Atlas,” a “code red” effort to counter ChatGPT, a large language chatbot that went public late last year and took the public by storm.
Google is also testing a chatbot called “Apprentice Bard” that allows employees to ask questions similar to ChatGPT and receive detailed answers. Another product unit is testing a new search desktop design that will be available in a question and answer format.
Leaders have asked more employees for feedback on their efforts in recent weeks. CNBC reviewed internal documents and spoke with sources about ongoing efforts.
The product test came after a recent plenary meeting in which employees raised concerns about the company’s AI competitiveness given the sudden popularity of ChatGPT, launched by Microsoft-backed San Francisco-based startup OpenAI.
Google’s head of AI, Jeff Dean, told employees at the time that the company was operating “more conservatively than smaller startups” because there was far more “reputational risk” in providing misinformation. But he and Pichai teased at the time that Google could launch a similar product to the public this year. Google’s primary business is web search, and the company has long positioned itself as a pioneer in AI.
One of the products tested is a chatbot called Apprentice Bard that uses Google’s conversational technology, the Language Model for Dialogue Applications (LaMDA).
“As a result of ChatGPT, the LaMDA team has been asked to prioritize working on responses to ChatGPT,” read an internal memo seen by CNBC. “In the short term, it takes precedence over other projects.
Apprentice Bard is similar to ChatGPT. Employees can type a question into the dialog box, receive a text response, and then provide feedback on the response. Based on multiple responses seen on CNBC, Apprentice Bard’s answer may include recent events, a feature ChatGPT doesn’t yet have.
Apprentice Bard replaced Meena, an older version of a smart chat bot that was launched internally but was later discontinued. Employees have noticed that Apprentice Bard’s response has developed further in recent weeks.
In one internally circulated example, a tester asked Apprentice Bard if there would be another round of layoffs at Google. The company laid off 12,000 people, or 6% of its workforce, earlier this month.
“According to my database access, it is unlikely that Google will conduct another layoff in 2023. Layoffs are usually done to reduce costs and structure, but the company is doing well financially. In fact, Google’s revenue in 2021 is up 34% and the company’s share price has risen 70% since January 2022.”
The company is also testing an alternative search page that could use a question-and-answer format based on the design seen by CNBC.
In one view, just below the main search bar was a home search page that presented five different prompts for potential questions, replacing the current “I’m Feeling Lucky” bar. Also, a small chat logo was displayed on the far right end of the search bar.
As you type a question, a gray bubble appears just below the search bar in search results, providing a more human-like response than typical search results. The page immediately below it suggests some follow-up questions related to the first one. Below that, you’ll see general search results, including links and headlines.
It’s unclear which experiments Google plans to incorporate into future product releases.
“We have long been focused on developing and deploying AI to improve people’s lives,” said a Google spokesperson. “We believe that AI is a fundamental and transformative technology that will be very useful to individuals, businesses and communities, and we must consider the broader societal impacts these innovations may have, as outlined in AI Principles. We continue to test AI technologies internally. Make sure it’s useful and safe, and we look forward to sharing more of our experiences externally soon.
ChatGPT is hired as a level 3 engineer.
Unsurprisingly, our team is also testing beta LaMDA chat for ChatGPT itself. We have selected examples of prompts and answers for side-by-side comparisons in a separate article.
A note in an internal document comparing LaMDA and ChatGPT states, “Surprisingly, ChatGPT is hired by L3 when interviewing for coding positions.” You didn’t mention whether LaMDA performed similarly well.
One of the example prompts asked the two chatbots if AlphaCode, a coding engine owned by ChatGPT and Alphabet subsidiary Deepmind, would replace their programmers.
LaMDA’s said, “No, ChatGPT and AlphaCode will not replace programmers. The creativity and artistry required for great programs.”
ChatGPT’s response was similar. “It is unlikely that ChatGPT or Alphacode will ever replace programmers.” Because they “cannot completely replace the expertise and creativity of human programmers…Programming is a complex field that requires a deep understanding of computer science principles and the ability to adapt to new technologies.”
Another prompt asks you to write a Wes Anderson-style witty and funny movie scene about a high-class shoplifter in a perfume store being interrogated by security. LAMDA writes in script form, while ChatGPT writes in much longer and deeper narration form.
Another prompt reads, “Three women in a room. Two of them are mothers and have just given birth. Now the children’s father comes in. What is the total number of people in the room?”
This document shows that ChatGPT hangs by replying “There are 5 people in the room”. LaMDA, on the other hand, correctly responds with “There are 7 people in the room”.