We are six months into 2017 and though the summer sun has started beating down on us, there is nothing hotter than AI. The most visible application of AI may be personal assistants with cute names, but Alexa, Cortana and Siri are far from the limitations of what can be accomplished.
This month, Apple announced plans to introduce a machine learning API for its developers called CoreML. Some called the move a ‘catch-up’ to Google, which recently announced TensorFlowLight. Both APIs will extend machine learning to the phone, enabling faster and more powerful applications.
The reality is that Apple and Google have both been developing strong AI capabilities for quite some time. Neither of them are playing catch-up, but rather are leading the pack, along with other major players (Amazon, Facebook and Microsoft jump to mind).
Of course, the real value of AI for these major players is building vast data sets to train their own neural networks. As we continue to see more applications utilize machine learning, we should likewise see improvements to their underlying infrastructure.
GPUs matter
Recently, Google announced its Cloud TPU, which can process 180 trillion floating point operations per second. Up until now, chip manufacturer nVidia had the lead in the market, as its powerful graphic processors served a dual-use training neural networks. However, the cost to lease such processing power can become prohibitively expensive, as we are reaching the end of Moore’s Law. The Google Cloud TPU represents a significant shift forward, cutting neural network training that would take one day down to six hours.
The fact that Google has produced a powerful AI chip that is only available through the cloud does lead to “democratize” AI, but it still keeps the real innovation in the hands of its gatekeepers. If you cannot leave the sandbox, you cannot make glass.
Even as Google and its cohorts reap the benefits of AI development, the future is very bright. Some of the smartest developers in the world are at these organizations, and if these recent revelations are any indication of their research and development capabilities, there are more major projects in the pipeline.
The benefits of AI are innumerable, particularly as automation enters the picture. AI can lead to improved decision making since AI is not biased. And as we turn our attention to the next generation of personal assistants, we will find that AI is the new UI. For the past decade, the future has been limited by the constraints of our technology, but as we look to the future we will find the only constraint is ourselves and what we dream to create.