Important trends in AI/ML you might be missing
By the end of 2020 Artificial Intelligence will deploy 48% of global CIO’s as per Gartner. As AI is improving day by day in many fields such as, Health care, Banking, Finance, Cyber security, Fashion not only these many inventions are carried on easily.
There are three trends that may be unnoticed at the moment but will have big long-term impacts:
Specific hardware and cloud providers are changing the landscape
As we can see cloud platforms are revamping their offerings to include Artificial Intelligence and Machine learning services. Machine Learning solutions becoming more demanding in nature, the number of CPUs and RAM are no longer the only way to speed up or scale.
More and more algorithms are being optimized for specific hardware than ever before, be it GPUs, TPUs, or “Wafer Scale Engines. Companies will limit their use of CPUs, to solve only the most basic problems. Besides these new hardware/chips may allow use of AI/ML solutions that were earlier considered slow.
Innovative solutions are emerging for, and around, privacy
Machine Learning models are not true black boxes, they infer the model inputs based on outputs over time, hence leads to privacy leakage. Challenges in data and model privacy will force companies to embrace federated learning solutions.
Federated learning helps the user’s data that never leaves their device/machine. These ML models have a small enough memory footprint to run, are smart enough on smartphones and learn from the data locally.
Robust model deployment is becoming mission critical
There are many upcoming models which are innovated by many people, but not necessarily experts on how to deploy them with model safety, security, and performance in mind. Artificial Intelligence / Machine Learning to identify threats like audit network security, physical security,, etc. To rectify such threats, companies need to put more emphasis on model verification to ensure robustness. Some companies are already using adversarial networks to test deep neural networks.
