The DeepMind AI laboratory shifts the main target from local weather change
Demis Hassabis, co-founder of Google’s startup DeepMind for artificial intelligence (AI).
Jeon Heon-Kyun | Getty Images
LONDON – Artificial Intelligence (AI) Laboratory DeepMind has shifted its focus from climate change to other areas of science, pursuing its original mission of creating Artificial General Intelligence (AGI) widely regarded as the holy grail of emerging technology to several people who are familiar with the matter.
While DeepMind, which was acquired by Google in 2014 for $ 600 million, denies having shifted its focus, several key climate change researchers who were part of the company’s energy unit have left the company in the past two years and few have Applied some changes. related announcements.
The unit of energy, which has drawn a fair amount of attention over the years, has gone and none of the company’s employees mention it on their LinkedIn profiles based on CNBC analysis. When asked, a DeepMind spokesperson said, “Over time we’ve moved away from a narrower focus on domains and cross-functional teams in DeepMind are now contributing to our growing climate and sustainability projects.”
They added, “In addition to ongoing partnerships with Google to scale the benefits of our energy-saving technology, new projects are ongoing in several areas, including more efficient approaches to machine learning.”
One of DeepMind’s early, and perhaps most successful, projects was to cut Google’s massive electricity bill and instantly reduce the company’s carbon footprint. The search giant, technically a sister company of DeepMind as both are operated by Alphabet, announced in July 2016 that it had succeeded in reducing the energy consumption of its data center cooling devices, which are designed to protect Google’s servers from overheating 40 % with the help of a DeepMind AI system.
DeepMind didn’t stop there. It has been working with the Google Cloud department on a new platform that enables AI control of cooling systems in commercial and industrial facilities.
Sundar Pichai, CEO of Google, said in a blog post in September that DeepMind and Google Cloud are making the platform available to airports, shopping malls, hospitals, data centers and other commercial buildings and industrial facilities worldwide. However, DeepMind and Google Cloud have yet to provide specific examples of where and how the platform is being used.
The DeepMind Energy unit
In 2017, DeepMind began recruiting more experts to Google’s new campus in King’s Cross, London, to investigate how AI can be used to slow the effects of global warming. A new team called “DeepMind Energy” has been formed and will be led by Jim Gao, a former Google technical director who co-led the data center project with DeepMind. Gao declined to comment on this story.
DeepMind Energy grew to around 14 people and was commissioned to come up with new AI technologies to combat climate change.
In 2019, DeepMind Energy announced its first big win. It had increased Google’s revenue from its wind farms in the US by around 20%. The wind farms are part of the Google network for projects in the renewable energy sector.
DeepMind’s AI was used to predict the energy output of the wind farms up to 36 hours in advance of actual generation – useful, since energy sources that can be scheduled to deliver a certain amount of energy at a specified time are often larger for the grid Are worth.
While the company’s successes matter, it has yet to be publicly confirmed where and how the energy-efficient AI has been applied outside of Google’s data centers and wind farms.
National Grid Nightmare?
At one point, DeepMind wanted to use its AI technology to optimize National Grid, which owns and operates the infrastructure that powers homes and businesses across the UK.
“We’re at the early stages of talking to National Grid and other major vendors about how we can investigate the kind of problems they’re having,” said Demis Hassabis, chief executive of DeepMind, in an interview with the Financial Times in March 2017. “It would be amazing if you could save 10% of the country’s energy consumption without new infrastructure, just by optimizing it. That’s pretty exciting.”
In March last year, it emerged that talks between DeepMind and National Grid had collapsed. The organizations spent much of their time working together, sometimes at a National Grid facility near Reading, Berkshire, England. However, there were many hurdles to overcome if anything was ever to be implemented.
Humayun Sheikh, an early investor who backed DeepMind’s launch, told CNBC that commercializing the company’s AI software was difficult, adding that without Google, the company would “likely have failed”.
Sheikh, who claims to have spent five years discussing the idea behind DeepMind with Hassabis before it was recorded, said, “The concern, the question marks, have always been in commercialization. How do you do it?”
Sheikh said National Grid may have had concerns about getting into a deal with a large company like Google.
He added, “I don’t think the model that DeepMind or any of those big machine learning and AI companies are using will work … unless it’s delivered as a service. But then the problem is with the data , the GDPR problems. ” The GDPR is a set of data protection and data protection regulations that were introduced by the European Union in May 2018.
Talks between DeepMind and National Grid eventually failed because they could not agree on the financial details, according to a source familiar with the matter who wanted to remain anonymous due to the sensitive nature of the discussion. “The money DeepMind was asking was outrageous,” the source said. “Most of their work is in-house and billed by Google,” added the source. “They sell the work of their AI engineers at inflated prices and not at the price that the market estimates for their production.”
When asked, a DeepMind spokesperson said, “We looked at the application of AI to optimize the UK’s power grid early on. These mutual efforts have been very collaborative and have resulted in many shared ideas on how technology can improve grid efficiency and resilience. These Exploration is now complete and we have no further work planned at this time. “
Gary Marcus, CEO of Robust AI robotics company and co-author of Rebooting AI, which takes a critical look at the industry and suggests ways forward, told CNBC that the technology for National Grid may not have worked well enough to do this justify costs.
“Their primary technique, in-depth learning, works best in well-controlled environments like board games and can grapple with the complexities and unpredictability of the real world,” Marcus told CNBC.
Sheikh added, “The technology may not have worked because it wasn’t really that mature.”
National Grid declined to comment.
In a podcast interview published in October, Hassabis reiterated that DeepMind’s AI software “could be applied on a grid scale,” suggesting he hasn’t given up. “We’d like to try that at some point and save energy on a national level,” he said.
While things did not go according to plan with National Grid in the UK, DeepMind could seek talks with other governments.
Leave driving forces
With around 1,000 employees, DeepMind’s workforce is divided into those who focus on research and those who focus on applying DeepMind’s AI. Research and publication do not solve real world problems, which is why the applied branch was established. Like the DeepMind Health division, which was acquired by Google last year, DeepMind Energy is aligned with the applied unit of the company.
The applied unit was headed by DeepMind co-founder Mustafa Suleyman but left in December 2019 and shocked many colleagues and supporters of the company. Known by friends and colleagues as the “Elk”, the entrepreneur, who has been described by colleagues and the media as an activist and visionary, now has a political role at Google. Suleyman declined to comment.
Several members of the DeepMind Energy team left the company shortly before or after Suleyman left. Gao left the company a few months before Suleyman to start his own start-up with DeepMinder Vedavyas Panneershelvam, while DeepMind Energy’s research engineer Jack Kelly also left to start his own start-up.
The driving forces behind DeepMind’s focus on climate change were Gao and Suleyman, two people with knowledge of the company who preferred to remain anonymous due to the sensitivity of the subject, told CNBC that it may be inevitable that DeepMind’s work in this area afterwards would slow down their disposals. The DeepMind Energy team that worked on some of DeepMind’s largest climate projects is almost non-existent today.
DeepMind said it had not scaled back its efforts on climate change, saying CNBC could not disclose related financial details.
A source from CNBC claims Hassabis decided to take some of the company’s funds for climate change and reallocate them to other areas.
Last month, the company announced that it had developed AI software called “AlphaFold” that can accurately predict the structure proteins will fold into in a few days to solve a 50-year-old “big challenge” to solve that could solve the problem way to better understanding of diseases and drug discovery. However, some scientists have questioned whether DeepMind “solved” protein folding.
On a call to journalists, Hassabis said, “The ultimate vision behind DeepMind has always been to build general AI and then use it to better understand the world around us by significantly accelerating the pace of scientific discovery.”
A DeepMind spokesperson added, “We have made some huge strides and made an impact, including increasing the projected value of Google’s wind power by about 20% and reducing the amount of energy used to cool Google data centers by up to 40% as well overall energy efficiency by 15%. “
“Now Google Cloud is offering this to commercial and industrial customers as a platform solution, helping companies around the world to make their facilities more sustainable.”