DeepMind is one of Alphabet’s AI projects that will ‘solve new problems’ and ‘advance the cause of mankind’, according to their home page.
The good news is that the artificial intelligence company made twice as much revenue in 2019 as it did in 2018. The bad news is that its gains mean nothing when compared to $572 million invested last year to keep it operational.
Now DeepMind is over $1bn in debts and its cofounder, Mustafa Suleyman is on leave after what the company calls, “ten hectic years”.
These years were rife with controversy, the most notable was the company’s partnership with Britain’s National Health Care where it acquired patient data for ‘research purposes’. Least to say, this was deemed illegal and caused quite a commotion in the media.
Does this mean Alphabet placed its bets on a lost cause? Experts believe this isn’t the case. Compared to other large-scale research-based projects like the Large Hadron Collider which costs around $1 billion per year to maintain, $500 million seems paltry.
Some scientists worry that DeepMind has been investing too much in a single research area called deep reinforcement learning. This technique is used for recognizing patterns that have to do with reinforced learning or gamification, which rewards people to score in a game or complete a new level.
DeepMind rose to popularity in 2013 when it presented a research paper that demonstrated how a single neural network system could be trained to play video games like Space Invaders and Breakout. This paper piqued the interest of Google which later acquired DeepMind in 2014.
DeepMind has proven successful enough that its trained AI beat a human in a game of chess, but did little more to uplift Google’s spirits. To its credit, DeepMind applied deep reinforcement learning to reduce cooling costs for Google’s server.
Can DeepMind create a profitable product in the next 10 years? The answer remains yet to be seen.