Heute 25

Gestern 89

Insgesamt 39231229

Montag, 27.09.2021
eGovernment Forschung seit 2001 | eGovernment Research since 2001

The next evolution in AI technology will be driven by a cloud-edge hybrid model making selective use of both cloud and edge resources to maximize the value delivered by AI while staying within latency, cost, security and privacy constraints.

In everyday parlance, the phrase “on the edge” usually denotes extremes of some kind. But in the context of Artificial Intelligence and Edge computing, “AI on the edge” is starting to become the norm rather than the exception. Edge AI is garnering attention partly due to the recent advances in edge compute capability and increasingly lightweight AI algorithms. But the real need for edge AI has risen from the changing dynamics of human-machine interactions and user engagement models.

“AI on the edge” is an amalgamation of two key technology areas – Artificial Intelligence and Edge computing. Artificial Intelligence or machine intelligence is now a commonplace technology in applications such as computer vision, language processing and complex data analysis. AI applications have traditionally been deployed on cloud servers with data being streamed from and to the source. For instance, video from a surveillance camera can be streamed to a cloud where it is processed by an AI model to detect anomalous behaviour.

On the other hand, edge computing, enables data processing on the edge of a network closest to the data source through edge software and hardware components. Mobile phones, IoT devices, smart home appliances and processors in a self-driving car are all edge devices. AI on the edge refers to AI algorithms deployed on edge devices instead of on the cloud.

AI as a technology started to gain traction with the rising ubiquity of GPUs (Graphics Processing Units) that are highly suited for complex AI models. The advent of cloud services enabling efficient training and execution of AI models really catapulted AI to the centre-stage. However, emerging requirements for both consumer and industrial applications are exposing some of the inadequacies of centralized cloud computing and making the case for edge solutions.

  • Data security and privacy: With increasing amounts of data generated by sensors and devices, privacy for personal data and security for business sensitive information are critical needs that cannot always be guaranteed with cloud applications.
  • Latency: With cloud-based AI, transmission of data back and forth from the cloud and can negatively impact latency and user experience.
  • Bandwidth and cost: Cloud based models that process large amounts of data can be crippled with bandwidth constraints and cost of transmission to and from the cloud.
  • Customization and personalization: For both consumer and industrial applications, there is an increasing demand for personalization and customization for users as well as machines. Centrally managed cloud solutions are cumbersome to customize and maintain.
  • Real time decision making: Centralized cloud-based solutions with higher latencies are inefficient for applications that require automated decision making in real time.

There is increasing compute power now available on edge devices with dedicated AIPs (AI processors) or NPUs (Neural Processing Units) capable of executing complex AI algorithms efficiently and effectively. The time is indeed ripe for AI to move from cloud to edge. The potential applications for edge AI are vast and diverse but here are a few industries that seem poised for a transformation.

  • Manufacturing – With increased automation in manufacturing, there is a need for real time automated process control using in-line sensor data. Edge AI based real time closed loop control can deliver higher efficiency, lower costs and improved quality for manufacturing
  • Automotive – Advanced Driver Assistance Systems (ADAS) which are one step towards fully autonomous cars aim to reduce accidents and ensure safety. Given the low latency of Edge AI solutions, they are at the heart of driver assistance applications that demand fast response times for driver feedback and control.
  • Media & Entertainment – Entertainment is becoming increasingly personalized and customized with the popularity of social media platforms, OTT (over the top) content and gaming. AI solutions on users’ edge devices can enable personalization of user experience based on their preferences and context without compromising data privacy.
  • Healthcare – With the recent Covid-19 pandemic exposing chinks in the armour of the world’s healthcare systems, there is an urgent need for efficient resource prioritization while maintaining high standards of care. When deployed on wearable devices or edge hardware in healthcare centres, Edge AI can help detect and diagnose diseases while furnishing insights and recommendations for healthcare providers.

While Edge AI will undoubtedly be transformative for many industries, cloud-based solutions will continue to serve applications that involve large amounts of data or require centralized processing and decision making. The next evolution in AI technology will be driven by a cloud-edge hybrid model making selective use of both cloud and edge resources to maximize the value delivered by AI while staying within latency, cost, security and privacy constraints.

---

Autor(en)/Author(s): Anusha Rammohan

Quelle/Source: ET CIO, 17.08.2021

Bitte besuchen Sie/Please visit:

Zum Seitenanfang