Technology insight
and understanding the true strengths of a family of technologies which is evolving quickly. Introducing any new technology into the workplace will require a learning cycle; with machine learning that process is more important than ever and requires extensive involvement of the employees who are subject-matter experts in order drive true transformation. Extensive communication needs to a core part of an AI programme. Make internal changes visible to staff, address concerns quickly, and be willing to switch direction when necessary. The integration of digital employees alongside human
linguists to support, train and continually improve AI systems. Businesses should not only be planning for AI investments but also looking to develop individuals for those new jobs. Successful implementation of cognitive technologies will require a range of different skills – many of which may be in short supply. When it comes to AI, neither businesses nor employees should fear the future. Businesses will benefit from early staff education and retraining as the digital workforce expands. From productivity gains to improved customer experiences, the benefits of embarking on an AI journey far outweigh the potential bumps in the road. However, the responsibility lies with business leaders to engage and motivate employees, preparing them early for AI’s potential impact as we move into an AI- enabled world. n
employees is a change in corporate culture as much as it is a change in process. Address the skills gap early with training Automation of tasks is inevitable, with or without AI, and it will continue as long as technology improves and we have the resources for further development. In turn, this will trigger re-definition of roles and, in some cases, will mean that jobs which exist today will disappear. Just as quickly, new roles for people will emerge needing, for example, business analysts, cognitive engineers, user experience designers and
...implementation of cognitive technologies will require a range of different skills...
Supercomputers
Mike Nicholas wonders whether you’ve got one at home S upercomputers are used for a wide range of computationally intensive tasks in various fields, including can do math equations. In November last year, TOP500 published a list of the world’s fastest supercomputers. The top five listed were: Sunway TaihuLight (China); Tianhe-2 (China); Piz Daint (Switzerland); Gyouko (Japan); Titan (USA).
gflop is estimated at USD 18.7 billion (or 152.9 adjusted to 2017 USD values.) ● In June 2013, Sony’s PlayStation 4 had a peak performance of 1.8 teraflops, with each gflop costing USD 0.22 (0.23 in 2017 values). ● By October 2017, the estimated cost per gflop had fallen to USD 0.03, for a platform built from commercially available parts at an overall cost of USD 2,050. Now imagine one million people simultaneously playing World of Warcraft on the internet. Without a supercomputer, how else could all the calculations be performed and delivered almost instantaneously for the players? In 2017, Microsoft launched the Xbox One X, at a price of USD 499, offering six teraflops of graphical power exceeding the 4.2 teraflops offered by Sony’s PS4 Pro. (I’ll let you do the math.) Perhaps you do have a supercomputer at home? n
quantum mechanics, weather forecasting, climate research, oil and gas exploration, molecular modeling, and simulations (e.g. early moments of the universe, airplane and spacecraft aerodynamics, detonation of nuclear weapons, nuclear fusion). Throughout their history, they have been essential in the field of cryptanalysis to breach cryptographic security systems and read encrypted messages. Colossus, the first ‘supercomputer’, which was designed to crack the German code during the second world war, could read up to 5,000 characters a second. Although impressive for its time, today Colossus would only qualify as a calculator. Supercomputers are ranked using a metric called ‘flops’ (floating point operations per second), which is a measurement of how fast the machine
The Sunway TaihuLight machine has achieved 93 petaflops. (A ‘petaflop’ is 1,000 teraflops – and a ‘teraflop’ is a trillion flops.) You might think supercomputers are very expensive, but perhaps it depends on the metric. Wikipedia (https://en.wikipedia.org/ wiki/FLOPS) charts the approximate cost of hardware per gigaflop (gflop) since 1961 as follows. (By the way, a gflop is a meagre one billion operations per second.) ● In 1961, the IBM 7030 Stretch performed one flop every 2.4 microseconds. As there are one million microseconds per second, its cost per
27
| Professional in Payroll, Pensions and Reward |
Issue 38 | March 2018
Made with FlippingBook - Online magazine maker