In the second of our multi-part series, our regular bloggers look at what they expect from 2017 in AI, obsolescence management and processor technology.
Michael Both
Deep learning and artificial intelligence are flooding the market with things like digital assistants (e.g. Amazon Echo, Google Home) and semi/fully autonomous driving (e.g. Tesla, BMW, Mercedes). This is enabled by GPU technologies which allow for extremely fast parallel processing of the input to see if it matches something in its learned neural network.
Throughout 2017, artificial intelligence will become more engrained in our daily lives, enabled by the rapidly growing deep learning compute power of GPUs, with NVIDIA leading the way. In medical environments, for example, it could be used to interpret X-ray, MRI, and CAT scan results.
Abaco’s GPU product range enables this technology to be brought into equipment operating in the harshest environments. That means, in the military/defense market, AI could be applied to an enormous range of applications. Radar/sonar identification is one example, and smarter weapons and automated weapons systems is another—but perhaps much more significant is what AI can bring as the armed forces look increasingly to autonomous vehicles.
And: who knows—in the not-too-distant future, AI may well enable the development of robot soldiers…
Richard Kirk
There are some exciting new technologies on the horizon, but a major challenge for integrators is often focused on extending the life of existing systems, rather than creating new ones, as they continue to struggle with budget constraints. This means an ever increasing interest in obsolescence management and low-effort technology insertions. Understanding the fact that reducing the "total cost of ownership" has to be planned in early and not added on at the end of a program is vital to the success of long term programs. Minimizing the impact of Diminishing Manufacturing Sources and Material Shortages (DMSMS) will be a key area for 2017—and a long way beyond.
Pete Thompson
As die shrinks get harder and harder, we will see other innovations from Intel to keep us on track with Moore’s Law. It is speculated that we will see an increase in the number of cores in CPUs appropriate for embedded use from the current four to six.
The next disruptive technology will be support for AVX-512 instructions. This doubles the processing speed of vector math. AVX-512 is already here on Xeon-Phi but this is hard to use in embedded applications due to power and packaging. AVX-512 will start to make its way into more mainstream CPs during 2017 and 2018.
