A man walks through Google offices in New York City on January 25, 2023.
Leonardo Munoz | Corbis News | Getty Images
Google is testing new artificial intelligence chat products that are likely to impact a future public product launch. These include a new chatbot and a potential way to integrate it with a search engine.
the alphabet The company is working on a project under its cloud unit called “Atlas” which has to be responded to with “Code Red”. ChatGPTthe large-language chatbot that took the public by storm when it went public late last year.
Google is also testing a chatbot called Apprentice Bard, where employees can ask questions and get detailed answers, similar to ChatGPT. Another product unit tested a new search desktop design that could be used in a question-and-answer form.
Executives have reached out to more employees for feedback on the effort in recent weeks. CNBC looked at internal documents and spoke to sources about the ongoing effort.
The product tests come after a last one all-round meeting where employees raised concerns about the company’s competitive advantage in AI given the sudden popularity of ChatGPT, launched by OpenAI, a San Francisco-based startup backed by Microsoft.
Google’s AI boss Jeff Dean told employees at the time that the company had a much greater “reputational risk” if it provided incorrect information and was therefore “more conservative than a small startup”. However, he and CEO Sundar Pichai teased at the time that Google could launch similar products sometime this year. Google’s core business is web search, and the company has long touted itself as a pioneer in AI.
apprentice bard
One of the test products is a chatbot called Apprentice Bard, which uses Google’s conversational technology LaMDA, or Language Model for Dialogue Applications.
“As a result of ChatGPT, the LaMDA team has been asked to prioritize working on a response to ChatGPT,” said an internal memo seen by CNBC. “In the short term, it takes precedence over other projects,” the email continued, warning that some employees are no longer attending certain unrelated meetings.
Apprentice Bard is similar to ChatGPT: employees can type a question into a dialog box and receive a text response, then provide feedback on the response. Based on multiple responses viewed by CNBC, Apprentice Bard’s responses may include current events, a feature ChatGPT does not yet have.
Replaced Apprentice Bard meena, an earlier version of an intelligent chatbot that was started internally but later discontinued. Staff have noted that Apprentice Bard’s answers have become increasingly progressive over the past few weeks.
In an example circulated internally, a tester asked Apprentice Bard if there would be another round of layoffs at Google. The enterprise quit 12,000 employees, 6% of workforce, earlier this month.
“According to my database access, Google is unlikely to conduct another round of layoffs in 2023,” the response reads. “Layoffs are generally carried out to reduce costs and structures, but the company is doing well financially. In fact, Google’s revenue grew 34% in 2021, and the company’s stock price is up 70% since January 2022.”
The company is also testing an alternative search site that could use a question-and-answer format, according to drafts seen by CNBC.
One view showed the home search page with five different prompts for potential questions placed just below the main search bar, replacing the current “I feel lucky” bar. It also showed a small chat logo on the far right of the search bar.
When a question is entered, the search results show a gray bubble just below the search bar, which offers more human answers than typical search results. Just below, the page suggests several follow-up questions related to the first one. Below that are typical search results, including links and headings.
It’s unclear what experiments Google plans to incorporate into future product launches.
“We have long focused on developing and using AI to improve people’s lives,” said a Google spokesman. “We believe that AI is a fundamental and transformative technology that is incredibly useful to individuals, businesses and communities, and as our AI principles outline we need to consider the broader societal impacts these innovations can have. We continue to test our AI technology internally to ensure it’s helpful and safe, and we look forward to sharing more experiences externally soon.”
ChatGPT would be hired as a Tier 3 Engineer
Perhaps unsurprisingly, Google teams also tested a beta LaMDA chat against ChatGPT itself. In separate documents, examples of prompts and responses have been selected in direct comparisons.
“Amazingly, ChatGPT is hired at L3 when interviewed for a programming position,” reads a note in an internal document comparing LaMDA and ChatGPT. It was not stated whether LaMDA would have performed similarly well.
One of the sample prompts asked both chatbots if ChatGPT and AlphaCode, a coding engine from Alphabet subsidiary Deepmind, would replace programmers.
“No, ChatGPT and AlphaCode will not replace programmers,” LaMDA replied, followed by four paragraphs of explanations including that “coding is a team sport” and that while chatbots “can help programmers work more efficiently,” they “cannot replace them the creativity and artistry necessary for a great program.”
ChatGPT’s response was similar, saying “It is unlikely that ChatGPT or Alphacode will replace programmers” as they “are unable to fully replace the expertise and creativity of human programmers… Programming is a complex field that a deep understanding of computer science requires principles and the ability to adapt to new technologies.”
Another prompt asks him to write a witty and funny Wes Anderson-style film scene in which an upscale shoplifter in a perfumery is interrogated by security. LAMDA writes in a script form and ChatGPT writes it in a narrative form that is much longer and more detailed.
Another prompt included a riddle asking, “Three women are in a room. Two of them are mothers and have just given birth. Now the fathers of the children come into play. How many people are in the room altogether?”
The document shows that ChatGPT is dropped and replies “There are five people in the room” while LaMDA correctly replies that “There are seven people in the room”.
