Google's upgrade of AI chatbot Bard could change search

The changes

Users can draft an email in Bard and export it to Gmail and Google Docs with new Export Actions. Bard will sport new source citation annotations starting next week. It also has a dark theme that makes it easier for developers to interact with, and developers will soon be able to export and run code with Replit, a collaborative SaaS browser.

Google also changed Bard from using Google's LamDA language model to the PaLM 2 language model. Google says this will help Bard generate and assign answers in a way that is more efficient than other LLMs as well as improve conversations between humans and the chatbot.

Separately, Google Search will see a generative AI upgrade, letting users receive conversational responses to specific questions. When the new Search Generative Experience feature recognizes a query that generative AI can answer, the first part of the results page will show AI-generated responses.

The cloud and search giant unveiled these new developments, among many others, including making Bard available in Japanese and Korean, at its Google I/O conference this week.

Opening Bard up to the world is a big step, said Daniel Newman, an analyst at Futurum Research.

"They're kind of saying, 'Yeah, we got thrown off guard early on; we had a bit of a mess of an initial debit. But we feel like we've gotten on track,'" he said, referring to Bard's early problems, particularly during its first public demo on Feb. 6, where it made a major factual error.

Your Tech Minute with Nick: ChatGPT and the Open Letter on AI

8:54

Fitting Bard into its business model

However, by not integrating Bard into Search, Google is still in the developmental stage of assessing ways Bard can fit into its larger business model, which has relied in large part on commercializing Search by selling advertising on its free-to-users search sites.

"I would certainly view Bard itself right now as more of a tech demo and a demonstration rather than a true product or commercial device service," Forrester analyst Rowan Curran said.

While it's interesting that Google has built out more integrations with third-party capabilities, similar to some of the plugins offered by OpenAI, it'll still in question whether Bard will become an actual commercial product, Curran said.

Moreover, it's likely optional for Google to turn Bard into a commercialized service. Meanwhile, Google remains strongly focused on "traditional" search and its Google Cloud Platform, the third-ranking public cloud after AWS and Microsoft.

They're kind of saying, 'Yeah, we got thrown off guard early on; we had a bit of a mess of an initial debit. But we feel like we've gotten on track.'

"What they're hoping to see is that people will see the new Google search experience as a more intuitive and interactive and natural way of getting information," Curran said.

Therefore, instead of adding a chat experience into Search like Microsoft did with Bing, Google simply enhanced its overall search experience with more options and capabilities.

"What they're doing is adding more tools for people running these searches to get the information contained within those searches," Curran added.

Google's choice not to incorporate Bard directly into Search is also a way to protect itself, he said.

"Google has a whole lot to lose other than Bard being bad," Curran said. He believes if Microsoft's efforts to incorporate ChatGPT into its Bing search engine had ended up going badly, the fallout wouldn't be nearly as bad for that tech giant as it would be for Google, whose biggest money maker is its search engine.

"It makes sense that Google took a more cautious approach to how it wants to integrate this technology," Curran said.

However, Newman said Bard and Search might stay separate for a while. "This is a part of the development phase."

Google is likely focused still on figuring out the cost associated with opening Bard to everyone before it can focus on fully integrating both Search and Bard, he said. In addition, Google likely still is evaluating costs associated with computing utilization and how Bard will run on different models.

Moreover, Newman said that the more generative AI features that Google adds to its search engine, the more significant challenges it will have monetizing Search. The more usable information the chatbot provides, presumably the less users would need to visit sites linked by Search. That could dramatically drive down revenue.

"As generative AI starts to find its way into [search], all those clicks that it possibly reduces are dollars that could be potentially be at stake if they don't figure out a new workflow for monetization," he said.

Esther Ajao is a news writer covering artificial intelligence software and systems.