News

Google Fires a Volley at Microsoft-OpenAI Dominance

• By David Ramel
• 05/11/2023

Microsoft's ascension to a position of cloud-based AI dominance, fueled by a colossal investment exceeding $10 billion in its partner, OpenAI, is now under siege as Google intensifies its AI initiatives. This challenge was underscored by a series of announcements unveiled during the recent Google I/O 2023 developer conference.

The ongoing corporate rivalry in the field of AI is most vigorous right now when it comes to web search, a vital component of Google's core operations. Google, whose name has become synonymous with the act of searching, is keenly focused on safeguarding this domain, which is swiftly evolving with the rise of conversational interfaces.

Following the introduction of OpenAI's ChatGPT, with its human-like conversational abilities, reports surfaced suggesting that Google initiated a "code red" response to narrow the AI gap. This effort culminated in an experimental project featuring its generative AI chatbot, known as Bard. Similar to Microsoft's "new Bing," Bard is powered by a large language model (LLM).

A comparison between Bing and Bard found the latter -- albeit an experiment designed to solicit feedback -- to be lacking in comparative functionality, perhaps rushed out the door too soon.

But now comes round two of the search wars, just a couple months later.

"With new breakthroughs in generative AI, we're again reimagining what a search engine can do," Google search exec Elizabeth Reid said in a May 10 announcement. "With this powerful new technology, we can unlock entirely new types of questions you never thought Search could answer, and transform the way information is organized, to help you sort through and make sense of what's out there."

Helping to do that is a new way to present search results. The new AI-powered search will generate a snapshot of key information to consider, providing links for users to dig deeper.

The search makeover announced at Google I/O 2023, "Supercharging Search with generative AI," includes experiments, which are now accessible in the time-honored "labs" approach.

"Today we're sharing a look at our first steps in this new era of Search, and you'll be able to first try these generative AI capabilities in Search Labs, a new way to access early experiments in Search," Google said.

The company invited users to sign up on a waitlist to access Search Labs in order to get their hands on an early experiments in the coming weeks. Search Labs is available -- initially only in English -- on Chrome desktop and the Google App (Android and iOS) in the U.S. Coincidentally, while announcing that waitlist, Google at the same time removed the waitlist for Bard, making it instantly accessible to everyone.

Three early Search Labs experiments, which will be accessible for a limited time, include:

• SGE brings the power of generative AI directly into Google Search. The new Search experience helps you quickly find and make sense of information. As you search, you can get the gist of a topic with AI-powered overviews, pointers to explore more, and ways to naturally follow up.
• Code Tips harnesses the power of large language models to provide pointers for writing code faster and smarter. You can ask how-to-questions related to a specific set of programming languages (C, C++, Go, Java, JavaScript, Kotlin, Python, TypeScript), tools (Docker, Git, shells), and algorithms.
• Add to Sheets helps you insert a search result directly into a spreadsheet and share with friends. It's your perfect research companion for planning a trip and adding info to your itinerary, or keeping track of other information you've found on Search.

Beyond Search Labs experiments, other new features for search include image capabilities, app integration, coding enhancements and export actions. Users will be able to generate images based on prompts, or use Google Lens to analyze their own photos and get creative captions. They will also be able to integrate search with various Google apps and services, such as Gmail, Docs, Maps, and others, just has Microsoft has done with many of its products and services.

What's more, search can integrate with external partners across a range of industries, including Adobe Firefly, Kayak, OpenTable, ZipRecruiter, Instacart, Wolfram and Khan Academy. For developers, search will offer more precise source citations, a dark theme, and the ability to export and run code with Replit.

The company invited feedback from users to make search better and more useful for creativity, productivity, and curiosity.

Much of the above functionality is delivered via PaLM 2, Google's new LLM. It's described as a state-of-the-art language model with improved multilingual, reasoning and coding capabilities, coming in four sizes: Gecko, Otter, Bison and Unicorn. Those are listed from smallest to biggest, so the lightweight Gecko is ideal for mobile devices.

Google said PaLM 2 is more heavily trained on multilingual text -- spanning more than 100 languages -- and can understand, generate, and translate nuanced text across a wide variety of languages. It also reportedly demonstrates improved capabilities in logic, common sense reasoning and mathematics, as well as excelling at popular programming languages like Python and JavaScript.

Speaking to PaLM 2's power and versatility, Google announced more than 25 new products and features powered by the new LLM. Those include Workspace features, Med-PaLM 2 for healthcare, Sec-PaLM for cybersecurity and Duet AI for Google Cloud, a generative AI collaborator designed to help users learn, build, and operate faster. And those 25 products and features represent only a small part of what Google announced at I/O, with the company listing "100 things we announced at I/O 2023."

With all of the recent blowback and backlash surrounding supposedly out-of-control AI advancements, Google also took the time to emphasize its "responsible approach" to AI.

"There are known limitations with generative AI and LLMs, and Search, even today, will not always get it right," Reid said. "We're taking a responsible and deliberate approach to bringing new generative AI capabilities to Search. We've trained these models to uphold Search's high bar for quality, and we will continue to make improvements over time. They rely on our hallmark systems that we've fine-tuned for decades, and we've also applied additional guardrails, like limiting the types of queries where these capabilities will appear.

"To evaluate the information for yourself, you can also expand your view to see how the response is corroborated, and click to go deeper."