So, what is DeepMind working on? How is it developing the technology? And what does it have in store for future generations? DeepMind has already been featured in the Nature and Science journals a few times, and many of its papers have been openly available. But what is it doing now that will make them so important? Let’s find out! We’ll also explore what their competitors are working on.
While they have already been creating some pretty impressive AI, DeepMind has a more commercial agenda. Rather than just focusing on AI agents that can do specific tasks, the company is also working on general AI systems that will be able to leverage breakthroughs in science, including physics, climate change, and neuroscience. Despite these ambitions, the company has struggled with a difficult choice between scientific research and making products.
Google acquired DeepMind Technologies in 2014 for $500 million. The company’s projects range from solving structural biology problems to assisting with nuclear fusion research. However, it was recently caught up in controversy due to its relationship with the UK healthcare system. It has since pulled back from deploying AI in front-line healthcare. DeepMind was founded by Demis Hassabis and Mustafa Suleyman, and currently has approximately 1,200 people working on both sides of the Atlantic.
Google DeepMind has entered the healthcare market with its first major project. In a collaboration with the Royal Free London NHS Foundation Trust, DeepMind received patient data that could help doctors better diagnose and treat acute kidney injuries. While DeepMind is not known for its health-specific domain expertise, the company has been able to garner the trust of healthcare providers through its power and prestige. So, what is DeepMind Working on Now?
Aside from its research in medical applications, DeepMind has also begun to use its AI to tackle climate change. Their AlphaGo program combined reinforcement learning and Monte-Carlo tree search to beat world champion Lee Sedol, and has been used in speech recognition and image search. As a result, AlphaGo has replaced 60 rule-based systems at Google. This is a great achievement for a small team!
While DeepMind’s AlphaFold AI program has been used in a variety of places in Google, it is still largely a research project. Its AlphaFold project is an excellent example of the power of artificial intelligence in handling messy problems. The research team will continue to develop and improve AlphaFold until it’s ready for mass use. If it meets its goals, DeepMind will definitely be able to beat any competition in the field.
Until October 2016, DeepMind had not disclosed what it was working on. At that time, it was not clear how much of this data it was gathering. Nevertheless, the company has stated that it will share these algorithms with the community once they are developed. The researchers will also share any new modes of analysis with the general public, though they will not publish algorithms that they have trained. So, how can we know for certain what DeepMind is working on?
In addition to DeepMind’s own research, some of the engineers at the company are leaving to launch their own startups and non-profits. According to its website, 11 of its alumni have already launched their own companies – some of them are backed by Alphabet and others have their own ventures. In addition to the two main companies mentioned above, there are other projects that DeepMind is working on. These include business intelligence development, cybersecurity, and customer relationships.
The company’s latest project is Streams, which is a smartphone app that analyzes the results of blood tests to determine if a patient has suffered an acute kidney injury. This condition can cause major health problems and even require dialysis, transplant, or death. In the UK alone, more than 40,000 people die from acute kidney injury every year. DeepMind is working to make these tools more accessible and useful for patients.
The new algorithm dubbed AlphaFold2 has started making an impact on scientific research. The AI program is so accurate that it can predict the shape of proteins to the nearest atom. The team has published detailed descriptions of the system and released the source code. They have also partnered with the European Bioinformatics Institute to create a public database for protein structures. The database currently contains 800,000 entries and is projected to reach a hundred million entries by the end of next year.