Google flaunts AI prowess with ‘Search Labs.’ Should Microsoft be worried?

For the longest time, Google has held an undefeated position as a search engine behemoth, all while recently proposing that AI is just another evolutionary point in its steady search domination. Microsoft’s rapid progress with AI is giving the software giant a leg up in chasing Google, the incumbent titan.

The search engine behemoth is not taking it easy. Just yesterday, Google concluded its annual I/O developers conference, and as expected, its latest obsession, generative AI, took center stage. After months of being under pressure to reinvent its core search business to take advantage of the rise of artificial intelligence programs that can generate content, Google finally unveiled ‘Search Labs.’

Simply put, Search Labs is a version of Google’s world-conquering search engine that uses large language models and AI tools trained on enormous volumes of text to answer users’ queries conversationally. During his presentation on Wednesday, Sundar Pichai, Google’s chief executive, said the new experimental space allows users who sign up for a waitlist to access it. 

For now, the search feature will initially only be available in the US, though; Google said it would look to roll out the feature more widely in the coming months. Pichai also emphasized that certain features may eventually graduate to the leading search engine, while others will be scrapped entirely.

For context, Google has been racing to catch up with Microsoft on consumer AI products since December last year, shortly after Microsoft’s investment into OpenAI. Threats to Google’s long-held search engine dominance deepened when Microsoft partnered with OpenAI to relaunch Bing in February.

A month later, OpenAI revealed its language model, GPT-4, which users can access through a premium version of ChatGPT and via Bing. Amid all that, Alphabet merged its DeepMind and Google Brain AI research units to accelerate AI development in April.

The large language model by Google

The AI-driven search feature, as well as updates to its Bard chatbot and products such as Gmail and Google Docs, are powered by the company’s large language AI model, PaLM 2, which was also launched on Wednesday. 

PaLM 2 is also a competitor to rival systems like OpenAI’s GPT-4. “PaLM 2 models are stronger in logic and reasoning, thanks to broad training in logic and reasoning,” Pichai said onstage. “It’s also trained on multilingual texts spanning over 100 languages.” 

Google senior research director Slav Petrov also told journalists in a roundtable before the model’s announcement that PaLM 2 is much better at a range of text-based tasks, including reasoning, coding, and translation. 

“It is significantly improved compared to PaLM 1 [which was announced in April 2022],” said Petrov. In a research paper describing PaLM 2’s capabilities, Google’s engineers claimed the system’s language proficiency is “sufficient to teach that language.” They noted this is partly due to a greater prevalence of non-English texts in its training data. 

So far, within Google’s domain, PaLM 2 is already being used to power 25 features and products, including Bard, the company’s experimental chatbot. Updates available through Bard include improved coding capabilities and better language support. It also powers features in Google Workspace apps like Docs, Slides, and Sheets.

Can Google catch up with Microsoft?

Frankly, Google was an AI juggernaut until its recent stumbles, especially its botched unveiling of its Bard chatbot and slow rollout of generative AI capabilities across its various products. That has left the tech giant playing catchup to arch-rival Microsoft.

However, with an estimated 93% of the worldwide search market, Google has more to lose. A report by Bloomberg quoted Prabhakar Raghavan, a senior vice president at Google, who shared in an interview ahead of the conference that the company is trying to be thoughtful about how to set expectations with users about the capabilities of generative AI.

“One of the defining questions for me is, how do you delineate what these things can and cannot do in a way that’s understandable to users, to businesses? Or to the ecosystem at large?” Raghavan said, according to Bloomberg.