Google’s Multitasking Pathways AI is the most brain-like AI yet

Machine learning is a supremely powerful tool for data analytics which is opening up unique insights into all facets of our society. However, it’s extremely focused, and because of that, somewhat limited. Google’s new AI architecture “Pathways” is a fundamentally different way to approach artificial intelligence, through developing an AI that is more multifaceted than ever, and potentially capable of solving far more problems than any individual algorithm yet written.

As it stands, most machine learning algorithms are trained to respond to one form of data and make their assumptions based on that. Pathways is different in that it can take on board multiple types of information, potentially letting it respond to text, speech, and visual inputs, as well as make different responses based on a combination of or select individual pieces of, information — depending on what its current task is.

Such an AI would be able to understand differences in perspective, but similarities of topic. For example, if given an image of a bird, a recording of someone talking about birds, and a written article about a specific species of bird, it would be able to recognize they are all about the same topic, despite coming from vastly different sources. That could allow Pathways to be trained for a greater number of tasks than any existing machine learning algorithm, but more importantly, draw on its previous experiences learning different tasks, and to learn new tasks more quickly and effectively.

Google also claims Pathways has the potential to be a much more efficient artificial intelligence, as the algorithm will only activate relevant portions of its network to solve whatever problem it’s currently working on.

All of this is Google’s latest effort to corner a market which is expected to explode throughout the next decade. Some of the world’s largest companies, such as in the U.S., China, and Europe, are vying to become the leading producers of the most capable AI solutions. The company or country that wins that race is likely to have an enormous edge over the competition.

It’s more than just a capitalist sprint, though. The hope is that advanced AI can help solve some of the crises potentially facing humanity in the coming years, from climate change, to future pandemics, to food shortages. Advanced, unbiased intelligences like Pathways have heaps of potential to do good, too.