5 AI TRENDS TO LOOK AT OUT FOR IN 2019
We tended to witness a dramatic rise within the platforms, tools, and applications supported Machine learning and AI. These technologies not solely compact software system conjointly the internet business however also different verticals like health care, legal, producing, automobile and agriculture.
Here are five AI trends to look at out for in 2019:
1) the increase of AI-enabled chips
Unlike alternative software, AI heavily depends on specialized processors that complement the CPU. Even the quickest and most advanced CPU might not improve the speeds of training an AI models. Whereas influencing, the model wants further hardware to perform advanced mathematical computations to Hurry tasks like object detection and automatic face recognition. In 2019, chips makers like Intel, NVIDIA, AMD, ARM and Qualcomm can ship specialized chips that speed up the executions of AI-enabled applications. These chips are going to be optimized for specific use cases and situations associated with computer vision, natural language process and speech recognition. Next generation applications from the health care and automobile industries can think about these chips for delivering intelligence to end-users.
2) Convergence of IOT and AI at the sting
In 2019, AI meets IOT at the sting computing layer. Most of the models trained within the public cloud are deployed at the sting. Industrial IOT is that the prime use case for AI that may perform outlier detection, root cause analysis and predictive maintenance of the instrumentation. Advanced ml models supported deep neural networks are optimized to run at the edge. they’re going to be capable of handling video frames, speech synthesis, time-series information and unstructured information generated by devices like cameras, microphones, and different sensors. IOT is geared up to become the largest driver of AI within the enterprise. Edge devices been equipped with the special AI chips supported FPGAs and ASICs.
3) ability among neural networks became key
One of the vital challenges in developing neural network models lies in selecting the correct framework. Data scientists and developers got to choose the correct tool from a inordinately of decisions that embrace Caffe2, PyTorch, Apache MXNet, Microsoft cognitive Toolkit, and TensorFlow. Once a model is trained and evaluated during a specific framework, it’s powerful to port the trained model to a different framework. The lack of ability among neural network toolkit is hampering the adoption of AI. to deal with this challenge, AWS, Facebook and Microsoft have collaborated to make Open Neural Network Exchange (ONNX), that makes it attainable to recycle trained neural network models across multiple frameworks. In 2019, ONNX can become a vital technology for the trade. From researchers to edge device makers, all the key players of the scheme can think about ONNX because the customary runtime for inferencing.
4) automatic machine learning can gain prominence
One trend that is about to modification the face of ML-based solutions basically is AutoML. it’ll empower business analysts and developers to evolve machine learning models that may address complicated situations while not looking the standard method of coaching ml models. When addressing an AutoML platform, business analysts keep targeted on the business drawback rather than obtaining lost within the method and progress. AutoML absolutely fits in between psychological feature Apis and custom ml platforms. It delivers the correct level of customization while not forcing the developers to travel through the frilly progresses. in contrast to cognitive apis that are typically thought-about as black boxes, AutoML exposes constant degree of flexibility however with custom information combined with moveableness’.
5) AI can modify DevOps through AIOps
Modern applications and infrastructure are generating log information that’s captured for categorization, searching, and analytics. the huge information sets obtained from the hardware, operating systems, server software and application software will be aggregative and related to to search out insights and patterns. once machine learning models are applied to those information sets, IT operations remodel from being reactive to predictive. When the ability of AI is applied to operations, it’ll redefine the approach infrastructure is managed. the appliance of ml and AI in IT operations and DevOps can deliver intelligence to organizations. it’ll facilitate the Ops groups perform precise and correct root cause analysis.