A 2016 statement from Alphabet defined Google’s parent company as an ‘AI first’ company. While that declaration certainly didn’t mark the beginning of Alphabet’s foray into AI, it did indicate where the strategic focus of the world’s third biggest company by market capitalisation lay. Alphabet’s long term ambitions lie in developing the latest technology in the world of AI across commercial sectors. Google’s online advertising revenue is still the company’s cornerstone. AI technology produced by Alphabet’s data scientists is certainly being applied to existing internet services but its executives are developing and planning to commercialise new applications.
The sector that currently holds the most promise for the commercial application of Alphabet AI is healthcare. The company’s ‘Medical Brain’ health research unit is developing ‘neural network’ AI algorithms that are built to ‘learn’ and reach their own conclusions from data rather than simply crunch it according to set rules. This kind of AI also means data fed into the AI doesn’t need to meet a single set format. Any kind of document, including files with notes in untidy handwriting, can be given to the AI and it will be able to pick out the information before archiving it into its internal database. This reduces the time spent on current predictive analysis software models by as much as 80% as the bulk of the work comes from standardising data formats.
The result is an algorithm that can quickly process vast amounts of disparate medical records on a patient, combine it with the data of their current vital readings and make eerily accurate predictions on medical outcomes. A case study recently published by the Medical Brain team in the Nature journal detailed how the algorithm processed 175,639 data points on a female patient admitted to hospital with late stage breast cancer. It’s statistical analysis of her death risk before leaving the hospital came out just below 20%. Unfortunately, the patient passed away within days.
While the particular case study covered had a sad outcome, the value to the health care system is obvious. With time, additional data that Alphabet has on individuals, such as information from Android smartphones on walking and other exercise time, indications of mental decline and even weather and traffic conditions in the locations they frequent, is expected to be added into predictive models on individual health.
There is already plenty of evidence that Medical Brain’s algorithmic models are significantly more accurate than existing software and the technology is still at an experimental stage of development. However, tools being worked on that can predict symptoms and ailments, how long hospital stays are likely to be, the chances or re-admission and even impending fatality show clear, some say eerie, promise.
Alphabet has not yet reached any conclusions on the future commercial model for its AI healthcare technologies but they would be expected to be sold into hospitals and clinics in some form and the health insurance sector is another obvious commercial opportunity.