How AI is being used to improve chips & software
Moore’s Law, the observation that the processing power of a computer doubles roughly every two years, is a topic of constant debate. One camp believes that the concept is gradually becoming obsolete, whereas the other upholds that the law still applies. Admittedly, it’s becoming clear that the growth rate of central processor units has slowed. At last month’s GPU Technology conference, Jensen Huang, CEO of NVIDIA, gave a presentation which argued exactly that. However, he also heralded a new wave of progressive software which, instead of relying on existing chip development techniques, harnesses the power of Artificial Intelligence. New AI based software could lead to the continuation of Moore’s Law – let’s look at how. . .
Accelerating software development with AI
Paradoxically, AI software could be the next step in making AI software. In theory, this will make computers much smarter. According to Reza Zadeh, Stanford University professor and CEO of software startup Matroid, computing power is “a bottleneck” for machine learning. This is something which tech giants are readily addressing. At Google, the AutoML project is automating one of the most challenging parts of AI software design. AutoML hasn’t just matched the ability of human software developers – it has exceeded it. Similarly, Microsoft is touting the application of reprogrammable FGPA (field programmable gate array) software. Future chip solutions could even take inspiration from biology – IBM, for instance, is building prototype chips that use spikes of current, just like the neurons in our bodies. Ultimately, these projects represent yet another real world application of AI that could be incredibly beneficial for technological expansion.
How will AI affect software development?
The development of machine learning software for machine learning software makes perfect sense. Chip developers have committed themselves to improving performance whilst shrinking transistor size. In 2015, the International Technology Roadmap for Semiconductors stated that this strategy would no longer be economically viable. None the less, the computer chip industry has fought to stay relevant, finding alternative ways to improve power. Adding AI to the mix is another thing that might keep Moore’s Law ticking along nicely. This isn’t just about making better computers – it’s about answering the serious skill shortage that has resulted from the demand for AI specialists.
Google CEO Sundar Pichai has stated that machine learning software will make it easier for less skilled programmers to use the technology. But, if successful, it isn’t hard to imagine AI software development programmes disrupting employment patterns by replacing actual developers. Even so, the experiments of bigger companies will encourage young companies and established firms to explore the possibilities of AI software too. As with all innovations, interest will lead to investment and adoption. In time, this technique could eventually change how all software is made.
The application of Artificial Intelligence within software development demonstrates the need to find new ways to create powerful processors. This could involve reprogrammable chips, as well as inspiration from the natural world. Before AI software can really enhance the creation of, well, itself, there are considerable barriers to overcome. Computer chips will need to undergo radical changes in order to support the expansion of AI, and at the moment only the biggest tech companies have access to the necessary resources. It’s also somewhat ironic that even those with the perfect skillsets to find employment in the ongoing tech revolution will also be at risk of automation. However, this is still very much an ambitious idea. Machine learning specialists are still vital to tech companies – at least for now.
How might changes in chip development impact your business or industry? Will AI software keep Moore’s Law alive? Could AI software programmes replace software developers? Comment below with your thoughts.