Google on Tuesday made its long-awaited artificial intelligence chatbot, called Bard, available to the public – an attempt to regain its leadership in the most competitive and high-profile new technology in years.

The bot, which is based on tech that has been under development by the company for eight years, can answer questions, write text and have informative conversations on nearly any subject. But in a sign that Google is proceeding more cautiously than its rivals – something that has spurred internal tension among the company’s own employees – Bard will have a separate website and won’t immediately be prominently promoted through Google Search or the company’s other popular products.

“Bard is really here to help people boost their productivity, accelerate their ideas, and just fuel their curiosity,” said Sissie Hsiao, a vice president at Google and one of the executives working on Bard.

The company is months behind some competitors in rolling out the first version of its chatbot to the public, critical time that competitors have used to tweak their products. OpenAI, a start-up that developed ChatGPT, has allowed users to test its version since November and last week released its most cutting-edge technology, GPT-4. Microsoft rolled out a similar tool in its Bing search engine in February.

But OpenAI and Microsoft’s AI chatbots, which are based on the same technology, have also made mistakes and exhibited bizarre behavior, like Microsoft’s Bing identifying itself as “Sydney” and getting into hostile arguments with people asking it questions. That’s exposed the companies to criticism that the technology isn’t ready for widespread use. Google has tested its AI on employees for years – which one engineer at the time, Blake Lemoine, claimed had become sentient – but has erred on the side of caution with the public.

Still, Google’s slower pace has sparked frustration among some employees, who say the company has dropped the ball on generative artificial intelligence, the broad term for technology that uses powerful algorithms trained on huge portions of the internet to produce original content, from eerily humanlike text to vivid artwork.

Advertisement

Some blame Google’s slow start on concerns that the technology could hurt the company’s reputation if it’s released before it’s fully ready for public consumption, according to people familiar with internal discussions who spoke on the condition of anonymity to share information that has not been made public. Other AI leaders, like Meta’s chief artificial intelligence scientist Yann LeCun, have said tools released by the bigger tech companies have appeared to be less interesting and capable than ones released by smaller start-ups because they have more guardrails to prevent offensive or harmful outcomes.

Though Google and other companies have worked on the tech for years, and some of the computer science behind the latest products was theorized decades ago, the bots continue to surprise their creators as well as regular people with what they can do.

“Things are moving pretty fast,” said Larry Birnbaum, a computer science professor at Northwestern University who focuses on AI. “Even the people who work on these machines, many of them have been quite surprised by their capabilities.”

Last week, OpenAI released its latest version of ChatGPT, which can describe scenes and interpret what is going on in them, demonstrating “advanced reasoning capabilities,” according to its creators. Some AI boosters say human-level intelligences are just around the corner, while critics of the technology say the hype is obscuring the true risks and opportunities of the tech.

Either way, the level of activity and interest in cutting-edge AI is higher than it’s ever been, and technologists believe AI will displace jobs, create new ones, and revolutionize the way people interact with technology and the internet.

Google’s CEO has called AI “more profound than fire or electricity” and Big Tech companies like Google are scrambling to ride the wave and assure their dominance over the internet and tech development isn’t threatened by the disruptive technology.

Advertisement

OpenAI’s success with ChatGPT and the energy around it have prompted bigger companies to move faster than they otherwise might have.

“ChatGPT accelerated the forward motions around this whole ecosystem, it moved a lot of the vendors to share more openly what they have been doing for a very long time,” said Chirag Dekate, a vice president at tech research firm Gartner.

Eli Collins, a Google vice president and one of the executives contributing to Bard, said in an interview that the company has been working on generative AI tech for years and that its plans would not be influenced by recent intense interest in the space.

“We’re pretty much marching down that path that we were on, which is figuring out how we can responsibly deploy this at scale,” he said.

Google has long been king of search and AI, spending billions of dollars on developing cutting-edge technology. In breakthrough after breakthrough, the tech giant wove those developments into its existing products – from its search engine to its maps to its translation tools. Google published academic papers on scientific advancements made inside the company and improved existing products by adding AI tech to them.

For years, engineers inside the company have pointed out that large language models, the tech behind chatbots like Bard and ChatGPT, could be used in search to directly answer questions, rather than providing links or snippets from other websites, according to the people familiar with internal discussions.

Advertisement

The company makes the majority of its revenue from search ads, and some employees have argued that chatbots are the next evolution of Google’s core product, the people said. Bard is based on LaMDA, a large language model the company first showed off in May 2021.

In fall 2021, Google still wasn’t sure what product LaMDA would be included in first, according to Lemoine, the former Google engineer who was fired after he told The Washington Post that he believed the bot was sentient. Search and Google’s voice assistant were the main candidates, Lemoine said.

Google is well-positioned to take advantage of the explosion of interest in AI, said Dekate, the tech analyst. The company has invested for years in computer servers that can run the massive calculations needed for advanced AI, and providing the tech to other companies that want to use it in their products will be a big business, he said.

“The market is actually heading in a direction where Google is incredibly comfortable in,” Dekate said. “The real impact of this will likely be in the partner ecosystem.”

Still, confusion remains among employees about what exactly Bard is supposed to be. During a Bard demonstration at a Feb. 8 search event, Prabhakar Raghavan, one of Google’s most senior executives, used Bard to answer questions similar to those routinely typed in search. At the time, he said generative AI tech would be used to provide longer answers to questions the company decides have “no one right answer.” The company showed search results on the screen.

But at an employee town hall, an executive said Bard was not meant to be used as a search tool – something executives reiterated this week.

Advertisement

“Bard is a creative collaborator and it’s complementary to search,” said Hsiao.

Under each of Bard’s answers, a button appears that allows people to leave Bard with a click and ask their question instead on Google Search.

Chatbots from Google’s competitors are already being used by millions of people to draft essays and reports, screen potential romantic partners and write computer code. Images made by AI tools like OpenAI’s DALL-E and Stability AI’s Stable Diffusion have flooded social media for months, drawing wonder and consternation, while triggering debates on what it means to be an artist and how copyright laws should be applied to AI-generated content. Venture capitalists are pouring billions of dollars into AI start-ups, even as the rest of the tech funding ecosystem retrenches.

Google’s Bard is very similar to the bots released by Microsoft and OpenAI. Users type in a question in a text box, and the bot responds with answers. The number of questions and responses are capped to prevent the bot from being prodded into developing a combative personality, as happened with Microsoft’s Bing chatbot after users had multi-hour conversations with it. The company also has turned off Bard’s ability to produce computer code, a key limitation compared to ChatGPT.

The tech giant will begin rolling Bard out on Tuesday to people who sign up in the United States and United Kingdom. Eventually, it will be available in more countries and in languages other than English, the company said in a blog post.

Collins and Hsiao said the company is committed to putting out its products responsibly. They said multiple times that Bard is still an “experiment.” A demo given to reporters purposefully included an example of Bard making a mistake by inventing a new scientific name for a plant when answering questions about houseplants that are easy to care for.

Advertisement

Google’s response to the new wave of AI has been incoherent at times and appeared to be rushed. In early February, Microsoft held a major event where it launched the Bing chatbot, letting journalists use it themselves and then quickly giving access to regular people who use its web browser and search engine.

A day after the Microsoft event, Google showed its Bard chatbot in action during an event that also encompassed other Google products. A phone meant to demonstrate some of the tech went missing, and the company said Bard would only be available in the coming weeks. Its stock fell 7 percent as Wall Street analysts questioned whether the company was losing ground to competitors. A blog post that first announced Bard and Google’s plans for generative AI in search results contained an example of the AI making a mistake, underscoring concerns that the company’s tech wasn’t ready for prime time and that its rollout was rushed.

Last week, Google said it would begin putting generative AI tools in its workplace products, such as Google Docs and Sheets, in the coming weeks. Two days later, Microsoft said it had launched similar tools to 20 customers already, although a broader rollout would take longer.

When Microsoft unveiled that its own chatbot was named Bing – after its search engine – and would become a key part of the company’s search tool, concerns from investors that Google was falling behind in its core business began growing.

But replacing search with chatbots would upend the massive economy of content creators and advertisers that Google has helped create with its search engine.

Countless news organizations, bloggers and other web publishers rely on traffic from Google search, and the entire search engine optimization industry has sprung up to help publishers understand Google’s opaque and ever-changing search algorithm.

The tech giant also relies on a thriving internet to give people things to search for in the first place, bringing in consumers for it to track and target with advertising.

Comments are no longer available on this story