Google merges its powerful artificial intelligence laboratories in the middle of a commercial race with the competition

Google executives have been somewhat out of place for months.

The emergence in November of ChatGPT, the popular OpenAI chatbot, snatched the status of a leading artificial intelligence (AI) company from the Palo Alto giant, a banner it had held up for decades.

Microsoft's firm commitment to OpenAI, which has developed a version of ChatGPT for the Bing search engine, has forced Google to take measures so as not to be left behind.

If in February it presented Bard, its own conversational chatbot, yesterday it made another relevant announcement: its two large AI research laboratories, Google Brain and DeepMind, are merging into a single organization.

The movement is very significant.

Several of the best scientists in the world in this discipline work in one of the two companies.

Google Brain is responsible for the majority of AI-related applications that appear in Google products and services, from the Gmail screening engine to the translator or the browser.

neural networks were also created there

, a deep learning model that has been key in the development of natural language processing (a field in which chatbots such as ChatGPT fall) or computational vision.

More information

Why don't we stop talking about artificial intelligence?

The fight for advertising revenue that marks the future of the internet

Acquired by Google in 2014 in exchange for 500 million dollars, the British company DeepMind is dedicated to the most basic research.

Until now he never had aspirations to develop commercial applications, but rather tools that help advance future research.

AlphaStar has come out of his laboratory, a simulator capable of beating expert StarCraft II players, a real-time strategy video game with imperfect information in which it is essential to have intuition, imagination and cognitive skills to try to guess what the opponent is doing, or AlphaFold, an artificial intelligence that has predicted the structure of all known proteins (about 200 million molecules).

The new group will be called Google DeepMind and will be headed by Demis Hassabis, the low-key computer genius who until now ran DeepMind.

"Combining all this talent into a single team, which will be supported by Google's computing resources, will significantly accelerate our progress in AI," said the CEO of Alphabet (Google parent), Sundar Pichai, in a statement released yesterday.

The movement is striking because Pichai himself had been insisting in recent weeks on the need for the industry to operate cautiously in the race for generative AI.

We are facing a technology with "potential" to do a lot of damage, he says, and Google has chosen to "be very responsible" in its developments.

He has expressed this in several interviews, the last one last weekend on CBS.

But these regards seem to have suddenly dissipated.

The spark that triggered the decision to up the ante on AI may have to do with Samsung.

announced

last weekend, the Korean technology company, the world's largest mobile phone manufacturer, would be considering replacing Google with Bing as the default search engine for its devices.

In the Google offices they have known this since March, the information assures.

And, if consummated, it could mean losing about 3,000 million dollars a year.

This threat to its bottom line has caused Google to speed up another project it had in hand.

Baptized as Magi, it is a different search engine from Bard with which to deal with Microsoft's Bing.

It will offer a more personalized user experience than traditional Google search and will learn from previous searches.

It will interact with it through conversations, as is already the case with Bing, and "it will try to anticipate the needs of users," says

.

Can you feel the machines?

The summer of 2022 was, in a way, a premonition of what was going to happen in the following months.

Google then had several open fronts related to the big questions we ask ourselves today about AI.

Will these systems be able to match or exceed human intelligence?

Do chatbots really understand what we tell them?

The engineer Blake Lemoine, in charge of carrying out a series of tests on the LaMDA chatbot, assured in a report published by

that the tool he was analyzing had become self-aware.

“If I didn't know it was a computer program we recently developed, I would have thought I was talking to a seven or eight-year-old kid with a background in physics,” he said.

In an interview with EL PAÍS, the one who had been his boss, Blaise Agüera, defended the dismissal of Lemoine for disclosing internal documents and showed his rejection of the engineer's postulates, although he admitted that this type of debate would become more and more complicated.

A year earlier, the company fired the heads of the Ethical AI team after they published a scientific article warning of the dark side of the great language models, those behind chatbots.

You can follow