Google's AI Now Plays Catch-Up To Newbies

"Alphabet Inc.'s Google, the pioneer of some of the technology that paved the way for a recent string of eye-catching developments in artificial intelligence, is now trying to play catch-up.

In recent months, Google's competitors have publicly released AI-based programs that can generate images and text passages from simple prompts, capabilities that the tech giant has tested internally for years.

Those releases -- and the public attention they have generated -- are prompting Google to redouble efforts to be an "AI-first company," a phrase Chief Executive Sundar Pichai introduced as far back as 2016.

Google executives have recently sped up work to review and release artificial-intelligence programs to the general public, while also assigning teams of engineers to work on new ways to integrate new developments into areas such as the core search experience, said people familiar with the efforts.

Unlike OpenAI and other startups such as Stability AI, Google has released its most powerful image- and text-generation models only to a limited group of testers. Google executives in recent years have stressed the need to test new artificial-intelligence tools for signs of bias while guarding against potential misuse, concerns shared by many academics.

Such caution has at times frustrated researchers at groups such as the AI unit Google Brain, some of whom have left to raise money for their own startups where they can more easily release new products, said people familiar with the matter.

Last week, the head of Google's research division, Jeff Dean, published a more-than-7,000-word blog post summarizing the company's recent work in artificial intelligence, writing that the developments are "making their way into real user experiences that will dramatically change how we interact with computers."

The pressures add to a difficult business environment for Google, whose search and ad-tech operations have both been targeted by Justice Department lawsuits. Google also recently said it is cutting about 12,000 employees.

"We have long been focused on developing and deploying AI to improve people's lives," a Google spokeswoman said. "We believe that AI is foundational and transformative technology that is incredibly useful for individuals, businesses and communities, and as our AI Principles outline, we need to consider the broader societal impacts these innovations can have."

Microsoft Corp. said this past week that it would make a multiyear, multibillion-dollar investment in OpenAI, the company behind the viral ChatGPT chatbot and image-generation program Dall-E 2. Microsoft declined to comment on financial terms, but people familiar with the deal said the two parties discussed an investment of as much as $10 billion.

Microsoft Chief Executive Satya Nadella said last week that the company plans to infuse all of its products with artificial-intelligence tools such as those developed by OpenAI. Google's closest competition in online search, Microsoft's Bing, would be one likely target area, analysts said.

Google has been researching and testing the possibilities of artificial intelligence for years. In 2017, a group of Google researchers published a paper laying out a new AI model called the Transformer that ushered in a new generation of large, powerful programs for processing text, images and other forms of data.

At a Google conference in 2021, Mr. Pichai demonstrated two conversations with an experimental artificial-intelligence program called LaMDA, which stands for Language Model for Dialogue Applications. The model responded to questions with complete thoughts from the perspectives of the dwarf planet Pluto and a paper airplane, drawing applause from the live audience.

OpenAI drew on a $1 billion investment from Microsoft in 2019 to develop a powerful new model, GPT-3, based on the Transformer developed by Google, leading to new applications such as the first version of Dall-E.

In November, OpenAI publicly released a demo of a chatbot called ChatGPT. The application quickly drew more than one million users, generating creative answers to prompts such as, "Write a movie script of a taco fighting a hot dog on the beach."

Soon after, Google employees began asking whether the company had missed a chance to attract users. During a companywide meeting in December, Mr. Dean said Google had to move slower than startups because people place a high degree of trust in the company's products, and current chatbots had issues with accuracy, said people who heard the remarks.

At times, Google has also struggled to unite overlapping efforts by different artificial-intelligence teams within the company, including London-based DeepMind, which it acquired in 2014, said people familiar with the matter.

In 2021, Google ended yearslong efforts by DeepMind to establish a more independent corporate structure, such as potentially moving to a nonprofit structure or spinning off entirely, The Wall Street Journal reported.

Like OpenAI, DeepMind has worked on building computer systems that can closely mimic or even replicate human thought.

In December, Google and DeepMind researchers introduced a language model that could produce reliable answers to a limited set of medical questions, while still overall falling short of those typically provided by clinicians." [1]

1. EXCHANGE --- Google's AI Now Plays Catch-Up To Newbies
Kruppa, Miles.  Wall Street Journal, Eastern edition; New York, N.Y. [New York, N.Y]. 28 Jan 2023: B.1.