Google Brain Needs DeepMind

The Google-DeepMind partnership would be a befitting reply to Microsoft-OpenAI alliance
Listen to this story

What could be the reason behind Google’s parent company Alphabet having two AI research wings under it – Google Brain and DeepMind ? Ironically, both these companies have different philosophies and approaches towards AI. This, over the past few years, has resulted in small feuds between them. But together, they are stronger than ever, and now might be the perfect time for them to align and compete against rivals.

Guess what? It’s already happening!

Google and DeepMind are collaborating on a next-generation project called the Gemini . According to a report from The Information , both companies are “ pausing grudges ” and joining forces to go beyond what OpenAI has achieved or planning to achieve. The project is led by Jeff Dean , head of Google Brain.

AIM Daily XO

Join our editors every weekday evening as they steer you through the most significant news of the day, introduce you to fresh perspectives, and provide unexpected moments of joy
Your newsletter subscriptions are subject to AIM Privacy Policy and Terms and Conditions.

But in a surprising turn of events, Jacob Devlin , another researcher from the Google AI team left to join OpenAI. Devlin explained that Google Brain employees have been using ChatGPT’s data through ShareGPT – a website for posting conversations with ChatGPT – for training Bard . He pointed out that other companies using OpenAI’s data to train their own chatbot is a violation of its policy. On a side note, Devlin has previously worked for Microsoft as well.

Google has rubbished this claim completely . But The Information report claims the company only stopped using ChatGPT’s data after Devlin warned them not to. Talk about family drama !

Download our Mobile App

Sometimes, being different is good

The rumours about the two companies’ ideologies not sitting well together have been on since Alphabet’s acquisition of DeepMind in 2014. But even after that, it has been boding well for Google and its products. DeepMind only became profitable in 2021 and the main source of revenue for the company had always been providing technical support to Google’s products. DeepMind does not release its products and services directly to consumers. Reinforcement learning (RL), the research organisation’s biggest selling point, has been the most-utilised offering to Google.

At NVIDIA GTC 2023 , Demis Hassabis , CEO of DeepMind, also said that DeepMind’s RL technology has been enabling Google’s huge data centres to save more than 30% of energy . Moreover, recommendation systems built by DeepMind are the ones powering Google Play and YouTube.

Sometimes, it backfires

A part of the leadership behind Google’s chatbot Bard is Sissie Hsiao , the one who oversaw the development of Google Assistant. Additionally, in February, the Blueshift team of Google announced that they would be working closely with DeepMind to further the parent company’s LLM capabilities, led by Oriol Vinyals , the brain behind DeepMind and Google Research. It may be possible that all of these people have contributed to the release of Bard.

Moreover, OpenAI has hired many former Google and Meta employees that were working on AI projects in the past. These were the brains behind building ChatGPT as well. Clearly, something is missing at Google’s AI research end — why is the talent running away from it?

Interestingly, even with so many “ brains and minds ” powering the AI technology at Google, its latest chatbot Bard, is still counted as a failure , when compared to OpenAI’s ChatGPT or GPT-4. Built on LaMDA , Google’s LLM Bard has been producing hallucinatory results since the beginning.

Google Brain introduced the Transformer model in 2017, which formed the basis for all LLMs in the present day. However, the same lab is now trying hard to compete with Microsoft and OpenAI. In December, the company announced PaLM , another LLM, while DeepMind had released Gopher , their LLM.

DeepMind also has another LLM called Chinchilla , which powered Flamingo. But none of the LLMs is publicly available. In January, DeepMind announced that it will soon release Sparrow , a chatbot that does things that ChatGPT cannot. But since then, the company has been suspiciously quiet on this front. We say, the Google-DeepMind partnership would be a befitting reply to Microsoft-OpenAI.

On similar lines, the recent petition by more than a thousand AI researchers, calling to pause any further training of models beyond GPT-4 , is signed by three research scientists from the AI Safety team of DeepMind. It is hard to guess if this is to catch up with OpenAI or because of actual concerns about the dangers of this technology.

Read: Why Google Does Not Deserve DeepMind

Google’s Deep Pockets and DeepMind

Interestingly, a few research scientists from Google have also signed the same petition. This puts into perspective the same question again — is it to catch up with OpenAI or the companies are actually concerned about where AI is headed ? But now that Google has released Bard, it seems like it is all about competition.

There might be a clash between DeepMind’s and Google’s philosophy towards AI. DeepMind’s philosophy is started by Hassabis’ research-based approach. Though he has been the one who drove AI development all these years through AlphaFold, and several other innovations, he agrees with the notion that such “ powerful AI technology” poses a potential danger and can cause significant damage to humanity.

Hassabis has always believed in, “work on making machines smart, and keeping humanity at the centre”. Thus, currently, DeepMind is also applying AI techniques to solve the puzzle of nuclear fusion , to yield an abundant source of zero-carbon emitting and cheap energy source for tackling climate change.

There is no doubt that DeepMind is focused on developing AGI as well. Hassabis believes that the technology should be aligned with humanity. He advises “ not moving fast and breaking things” , which runs contrary to OpenAI’s “ fail-fast ” philosophy.

A lesser-known fact about Google’s acquisition of DeepMind is that it happened after Hassabis turned down Facebook’s offer. Hassabis explained that the reason behind accepting the $500 million deal from Google over Facebook was that the former accepted DeepMind’s ethical red lines happily.

Elon Musk, along with Peter Thiel, was the first investor of DeepMind. This took a complicated turn for DeepMind when Musk decided to invest in OpenAI, the rival. Now that Microsoft is funding OpenAI’s effort, Musk is out criticising the company’s turn towards a “for-profit” entity.

Sign up for The AI Forum for India

Analytics India Magazine is excited to announce the launch of AI Forum for India – a community, created in association with NVIDIA, aimed at fostering collaboration and growth within the artificial intelligence (AI) industry in India.

Mohit Pandey
Mohit dives deep into the AI world to bring out information in simple, explainable, and sometimes funny words. He also holds a keen interest in photography, filmmaking, and the gaming industry.

Our Upcoming Events

Regular Passes expiring on Friday
27-28th Apr, 2023 I Bangalore
Data Engineering Summit (DES) 2023

23 Jun, 2023 | Bangalore
MachineCon India 2023

21 Jul, 2023 | New York
MachineCon USA 2023

3 Ways to Join our Community

Telegram group

Discover special offers, top stories, upcoming events, and more.

Discord Server

Stay Connected with a larger ecosystem of data science and ML Professionals

Subscribe to our Daily newsletter

Get our daily awesome stories & videos in your inbox

Council Post: Retail Business through Generative AI

Today, retail technology is developing at a fast pace – whether it is business transformation or even exploring emerging tech (AR/VR and metaverse etc.) to give customers a more experiential journey. Businesses are innovating not only to remain relevant, but also, ahead. Some are really shaping the future of omni-channel retail by predicting customer expectations and market trends.

Cerebras Wants What NVIDIA Has

While OpenAI apparently utilised 10,000 NVIDIA GPUs to train ChatGPT, Cerebras claims to have trained their models to the highest accuracy for a given compute budget.