Covering Disruptive Technology Powering Business in the Digital Age

If you talk about technology today, you can bet that the topic of artificial intelligence would come up at least once in conversations. Almost all aspects of technology today are looking to have some form of machine learning; be it for data analysis, storage, backup or cybersecurity. Looking at use cases, AI and machine learning can be applied to almost any industry. From AI-driven facial recognition software in CCTV cameras to autonomic vehicles, through to agricultural innovations and even surgical and tumor analysis in the medical industry.

Machine learning automates analytical model-building based on data analysis. It learns from data, identifies patterns and makes decisions with minimal human intervention.

According to Josh Simons, Senior Director and Chief Technologist for HPC Machine Learning Program Office, VMware, there are three pillars for machine learning in VMware which is for smarter business, smarter VMware and smarter products and services. For smarter business, it is to ensure customer ML workloads run well on VMware platform while smarter VMware sees how ML is applied to improve VMware efficiency and productivity. From this, ML is injected into VMware products and services to enhance automation, scale and efficiency.

Having said that, Josh explained that ML can be separated into different stages of learning. The table below explains it.

Interestingly, Josh pointed out while ML is heavily used today, ML does not mean human-level learning and AI does not mean human learning intelligence. He said that both AI and ML rely heavily on data. The more data it looks through, the better its capabilities. Josh explained that ML has some limitations whereby it can be easily tricked into believing data.

He highlighted a couple of examples in which the ML algorithms can be easily be tricked. One of which is a picture of a road sign. Autonomous vehicles rely on data heavily and identification algorithms from the camera onboard the vehicle. If a road sign is clear, not smudged, the AI of the vehicle should be able to identify it easily. However, if the same road sign was a bit faded or rusty, the AI might identify it as something else, unlike a human eye which is still able to differentiate and read a sign even if it’s not clear.

Readings and data from these algorithms are crucial in any AI device. And it’s not just for autonomous vehicles. There have been reports and concerns on AI bias as well, especially when it comes to facial recognition. Reports have shown that AI is likelier to recognise fairer skin individuals compared to darker-skinned tone individuals.

“We have to be careful about how we use these algorithms. We have to avoid bias in the models being built. Training data is crucial to ML algorithm to create a model. We follow practices on ML but it’s not sufficient to just build and deploy. We need to have a test set.”

In other words, just as how VMware is looking more into AI for its products, the amount of data and tests is crucial to ensure the right decisions are made. AI is making a difference, but it is important to ensure that it’s making changes the right way.

(0)(0)