There's been a recent resurgence in interest in artificial intelligence (AI). The revival began late last year with the release of ChatGPT, a remarkably accomplished chatbot developed by start-up OpenAI. The program can answer questions, fill out job applications, complete essays, and even create haikus. Further fanning the flames was a courtship by Microsoft that included a $10 billion investment -- after the company had already ponied up $3 billion. Microsoft also announced plans to integrate ChatGPT into its search engine, Bing.
Alphabet (GOOGL 10.22%) (GOOG 9.96%) then made the mistake of hurriedly launching its next-generation chatbot, Bard, which flubbed its launch with a high-profile error during its showcase. This left many investors with the impression that the Google parent had ceded leadership in the AI race.
An announcement today by CEO Sundar Pichai should be a reminder that Alphabet has a long and distinguished history in AI and that the game is still in the early innings.
Image source: Getty Images.
An AI powerhouse
In a missive to employees that was later shared as a blog post, Pichai announced that the company would combine the Google Brain team and DeepMind -- two of the biggest names in AI -- into a single AI powerhouse. The resulting group, dubbed Google DeepMind, will seek to pool resources and "significantly accelerate our progress in AI," according to Pichai. The CEO of the newly combined entity is Demis Hassabis, former CEO of DeepMind. Jeff Dean, head of Google Brain, will assume the role of Google's chief scientist and report directly to Pichai.
A brief look at the history of these two groundbreaking teams suggests this move could harness Alphabet's existing expertise in AI and set the stage for its future success.
Google Brain
In 2012, when the company was still called Google, Alphabet drafted the talents of noted AI expert and Stanford University professor Andrew Ng and those of Google senior fellow Jeff Dean to launch the Google Brain project. The first breakthrough came after researchers culled 10 million images from YouTube and fed them into an AI system, which was then able to identify pictures of cats. This accomplishment, groundbreaking at the time, kicked off the modern-day race to harness AI.
Since then, the Google Brain team has made a number of significant breakthroughs in AI and machine learning. These include understanding and replicating human speech, translating foreign languages, assembling a complete human genome, detecting breast cancer in x-rays, and even automating the creation of other AI models.
The Google Brain team has also been extensively integrated into many of Alphabet's day-to-day operations, improving everything from Google Maps to its all-important Search and underpinning Google Cloud, the company's cloud computing business.
These accomplishments -- and many more -- should dispel the notion that Alphabet has fallen behind in the field it helped pioneer.
DeepMind
In 2014, Google acquired DeepMind, reportedly paying $500 million, but since then it has run the company as an independent entity.
While it may not be a household name, DeepMind has a laundry list of accomplishments. In 2016, DeepMind's AlphaGo system mastered the 2,500-year-old game of Go, one of the most complicated and difficult board games in the world, which had stymied AI researchers for years. Soon after, the system defeated the reigning Go world champion.
DeepMind also analyzed Google's massive data centers, assessing 120 variables to work out the most efficient cooling process. DeepMind achieved a 40% reduction in the amount of energy needed for cooling, ultimately slicing 15% of its overall energy spending. This resulted in hundreds of millions of dollars in energy savings, recouping the cost of buying DeepMind in a single stroke.
Most recently, researchers at DeepMind developed Gato, a general AI system that -- rather than being really good at just one task -- has mastered more than 600. The system is able to "play Atari, caption images, chat, stack blocks with a real robot arm and much more," according to a blog the DeepMind team authored to tout its latest accomplishment. While Gato doesn't yet have specific applications, its multifunctionality is a huge step forward for a single AI model.
What does all this mean for Alphabet and the future of AI?
Alphabet has a long and distinguished history in the field of AI, a fact some seem to be forgetting. While the excitement about ChatGPT is understandable, it's too early to declare a winner in the space.
Furthermore, Alphabet already has a large language model (LLM) capable of outperforming GPT-3.5 -- the system that currently underpins ChatGPT. Alphabet's Pathways Language Model (PaLM) has been scaled to 540 billion parameters, nearly three times larger than GPT-3.5. Alphabet is reportedly launching a new application programming interface (API) for PaLM and making it available to developers -- the first step in a direct challenge to its high-profile rival.
Yet for all this AI experience and the vast opportunity ahead, Alphabet shares are trading for about 4 times next year's sales. The stock hasn't been this cheap in nearly a decade.
The integration of Google Brain and DeepMind will bring together some of the best and brightest minds in the field of AI -- so don't be too quick to count out Alphabet in the race to harness AI. It could be a costly investment mistake.
