This article is part of our series exploring the business of artificial intelligence
DeepMind, the UK-based AI lab that seeks to develop general artificial intelligence, has finally become profitable, according to the company’s latest financial report.
Since its acquisition by Google (now Alphabet Inc.) in 2014, DeepMind has struggled to break even with its growing spending. And now, he’s finally giving his parent company and shareholders encouraging signs that he’s earned his place among Alphabet’s constellation of profitable businesses.
This could be wonderful news for the AI ââlab, which has bled significant amounts of money throughout its life.
But the financial report is also surrounded by vagueness that suggests that while DeepMind did find its way to profitability, it did so in a way that makes it inextricably linked to Google’s products and business model.
A turnover multiplied by three
According to the DeepMind file, he raked in Â£ 826million in revenue in 2020, more than three times the Â£ 265million deposited in 2019. Over the same period, his spending increased slightly to $ 717million. from Â£ 780 million. The company ended the year with a profit of Â£ 44million, compared to a loss of Â£ 477million in 2019.
The dossier does not provide many details on DeepMind’s sources of revenue, except for a paragraph that says: âThe company generates revenue through a service agreement with another company in the group for the provision of services. research and development. “
DeepMind does not sell products or services directly to consumers and businesses. Its customers are Alphabet and its subsidiaries. It is not clear which of DeepMind’s companies caused the increase in its revenues.
A source who spoke to CNBC said that DeepMind’s sudden increase in revenue could be “creative accounting.” Basically, this means that, since Alphabet and its subsidiaries are the only customers of DeepMind, it could arbitrarily change the price of its services to make it appear as if it is becoming profitable. DeepMind has not commented on the claim.
Sell ââreinforcement learning
DeepMind’s main area of ââinterest is deep reinforcement learning, a branch of machine learning that is very useful in scientific research. DeepMind and other AI labs have used Deep RL to master complicated games, train robotic hands, predict protein structures, and simulate autonomous driving. Scientists at DeepMind believe that advances in reinforcement learning will eventually lead to the development of AGI.
But research on deep reinforcement learning is also very expensive, and its commercial applications are limited. Unlike other deep learning systems, such as image classifiers and speech recognition systems, which can be directly ported and integrated into new applications, deep reinforcement learning models often need to be trained. in the environment where they will be used. This imposes technical and financial costs that many organizations cannot afford.
Another problem is that the type of research DeepMind is engaged in does not directly translate into profitable business models. Take, for example, AlphaStar, the reinforcement learning system that mastered the real-time strategy game StarCraft 2. It’s an impressive scientific feat costing millions of dollars (which was likely subsidized by Google, which owns vast cloud computing resources). But it has little use in applied AI without being reused (to the tune of millions more).
Alphabet has adapted DeepMind’s RL technology in some of its operations, such as reducing power consumption in Google’s data centers and developing technology from Waymo, Alphabet’s autonomous driving company. But while we don’t know the details of how the technology is applied, I guess Alphabet is outsourcing some of its applied AI tasks to DeepMind rather than integrating the technology from the AI ââlab directly into it. his products.
In fact, a separate division of DeepMind is engaged in applied AI projects for Google and Alphabet, but this effort is not directly related to the AGI research carried out by the main laboratory DeepMind.
The costs of AI talent and research
As big tech companies like Facebook, Microsoft, and Apple show interest in deep learning, hiring AI talent has become an arms race that has driven researchers’ salaries up. Top AI researchers can easily earn seven-figure salaries at large tech companies, making it difficult for academic institutions and nonprofit research labs to retain their talent.
In 2020, DeepMind paid Â£ 467million in personnel costs, almost two-thirds of its total expenses. The company has around 1,000 employees, a small percentage of which are highly paid scientists, researchers and engineers.
The rising costs of AI research and talent will present DeepMind with heightened challenges as it moves forward. It will depend on Google to fund its operations and subsidize the costs of its research.
Meanwhile, as a subsidiary of a publicly traded company, it will be scrutinized to determine the profitability of its technology. And for now, its only source of profit is Alphabet, so it will become more and more dependent on the purchase of its services by Google. This in turn may push DeepMind to direct its research into areas that can quickly turn into profitable businesses, which is not necessarily in line with its scientific goals.
For a company that pursues the long-term dream of general artificial intelligence and whose stated mission is to “advance science and benefit mankind,” the distractions of short-term profits and additional gains can be sidelined. ‘prove to be harmful.
The closest example I can find for the job that companies like DeepMind and its near-rival OpenAI are is Bell Labs, AT&T’s former research group. Bell Labs was a subsidiary of a very large for-profit company, but its work was unrelated to next quarter profit targets or shareholder incentives. Although generously rewarded for their work, its scientists were motivated by scientific curiosity, not money. They were looking for foundational ideas that pushed the boundaries of science, creating innovations that would not pay off for years and decades to come. And that’s how Bell Labs became the birthplace of some of the ideas and technologies that changed the twentieth century, including transistors, satellites, lasers, fiber optics, cell phones, and information theory. . Bell Labs had the freedom to discover and innovate.
For now, Alphabet has proven to be a patient owner of DeepMind. He gave up Â£ 1.1bn debt in 2019 and helped DeepMind report positive profits in 2020. It remains to be seen whether Alphabet will remain generous and true to DeepMind’s mission for the long term – and it does. is a long term. But if Alphabet’s patience runs out, DeepMind will find itself without clients, without funding, and without fierce competition from tech giants who will want to poach its talented scientists to achieve fundamentally different goals.