2018 Technology Trends: Custom Hardware Is Here to Stay
When it comes to AI technology trends, people love to talk about the impact on software and services. The not-too-far future promises digital assistants that speak in everyday vernacular. Having real conversations with AI is part of an aspirational future, but less discussed, though equally crucial, AI hardware is mostly ignored. The thing is, the AI hardware revolution is well underway.
AI Centric Hardware Is Here
The Hardware in Your Hand:
There’s a standing joke among AI researchers that as soon as a piece of AI research becomes part of an actual product, it’s no longer considered AI. Thus, the field of AI is a field where the finish line keeps mover toward the horizon of human-level intelligence. This scenario makes sense if you remember a time when speech synthesis was a challenge, machine vision was in the domain of labs, and even a game of chess was considered a topic worthy of AI research.
In 2017 we don’t even think about how our smartphone cameras focus, recognize smiles, exposure time, and add color corrections. It’s easy to forget that a “good” photograph has a large part of human subjectivity in it. The amount of HDR or bokeh our cameras produce are highly tuned, occurring in micro-fractions of a second. This is all possible with image processors.
The camera in the Pixel 2 is one such example.
But if you want to put on your AI researcher hat and say: “But hey, that’s not AI anymore, I want my AI to make my photos into works of art that are worthy of Van Gogh!” Then rest assured your phone is good at doing that as well. While it’s not always AI specific hardware, the use of mobile processors in the service to AI tasks is now standard. Mobile processors save companies computation power and time that otherwise would not have scaled with an increase of users or their profit margins.
Cloud and AI Centric Processors:
While your phones are not full of AI specific hardware (though they have a little of that), there are places where this specialized equipment is poised to make a considerable impact: cloud and data centers.
Google is one of many that want to leverage hardware to make AI calculations faster. This makes sense if you compare the price of services to the cost of the device. As the price of the equipment drops, the amount of profit that services will make starts to warrant the design of custom hardware. Here, each small improvement in power or speed will translate directly into money saved and earned. It’s no wonder the likes of Facebook and Google are spending so much effort in this arena. Facebook even wants the open source community to help out in its efforts.
Autonomous Cars (and Maybe Even the Car You’re Driving):
If we want to talk about personal consumer hardware, there are few things larger than cars.
Cars started out as more mechanical systems than electric, but modern cars are full of computers. Subsystems such as anti-lock brakes, fuel injection, and cruise control are all relatively conventional up until recently. As research into autonomous cars trickles down into safety features (e.g., forward collision detection, before the actual collision occurs) what goes inside your car is now designed specifically for AI algorithms.
Companies like Nvidia, let alone Google, Apple, Uber, and Tesla are now jumping head first into the booming autonomous car field. Figuring out a solution to autonomous (driverless) cars is a circumstance where decisions made “on-the-fly” is imperative to success. Using current algorithms of data, sending those to a server for calculations, and receiving results will not be a feasible solution. Reactions to road conditions need real-time assessment, and only onboard AI hardware can satisfy this requirement.
Get ready. In many ways, faster speed and the lower costs of hardware is helping AI come to the forefront of modern products and services. However, it isn’t only a story of better-faster-more; we are using it in novel ways that will tangibly improve our lives. The turnaround time for processor design is as fast as a few months compared the years it took during the booming PC era.
And with the prevalence of AI, so comes along with it an increase of data types to segment, interpret, and analyze for future business decisions. Whether that is investing in a data analyst, implementing PR reporting, or receiving a refresher course on your media monitoring platform, you’ll need to be prepared. Now is the time to move towards setting up your campaigns and team up for success.