Actions
Engagement
  • Author Author: rscasny
  • Date Created: Date Created
  • Last Updated Last Updated:
  • Views 136 views
  • Likes 1 like
  • Comments 6 comments
Related
Recommended

Poll: What areas will see the most benefit from AI at the Edge Technologies in the next 5 years?

Artificial intelligence (AI) has, for many years, operated within data centers in the Cloud, which had the necessary power to perform compute-intensive tasks that AI algorithms require to process vast amounts of data.  But the latency, privacy, security, and bandwidth challenges associated with cloud computing, especially for time-sensitive decision making, has spurred the evolution and development of AI at the Edge of the network. We have only seen the beginning of the AI innovation at the edge.

 

Poll Question: What areas will see the most benefit from AI at the Edge Technologies in the next 5 years?

Anonymous
  • in reply to Jan Cumps

    Yes you are right It is indeed difficult to document the AI model. But in the recent past lots of devices are getting their FDA approval for their AI/ML models.

    Most of the times you need to provide a much detailed documentation of Risk Analysis and Risk Mitigation. If you are familiar with medical devices certification you might have heard about ISO13485

    Because of pandemic reasons its free to read now (https://www.iso.org/obp/ui#iso:std:iso:13485:ed-3:v1:en ).

    To your question about code, ISO13485 requires your entire software design as per ISO62304 ( This should document about how you came up with a particular model ) once you have a model you should test and validate it.

    Risk analysis and mitigation should be done as per the guidelines give by ISO14971. Also additional documentation as required if you use SOUP (Software of unknown pedigree) or OTS (Out of the shelf) software. ISO13485 design and development SOP should detail this.

     

    Apart from ISO13485 , FDA requires the submition the clinical trial details. Trial phase happens once they verify your initial application. ( Generally systems with AI its really hard to find predicate devices so a normal 510K submission wont work)

     

    Regards,

    Aswin

  • in reply to aswinvenu

    Ashwin, how do you see the hard-to-put-on-paper magic of AI work together with the strict code security requirements of the medical sector?

    Medical is one of the sectors where you typically have to match each line of code to a requirement and a validation. The MISRA mistery

  • I work for bio medical firm here in India. I am very keen on the changes happening on the bio-med research field. I can see people are already started taking tele-monitoring very seriously.

    I think the pandemic has proved to the world that everyone should have atleast basic knowledge of your own body. There is a lot of research going on consumer medical devices and most of them are highly based on AI/ML models.

    It spreads across a small scale relatively simple heartrate or breathing rate based cardiac/respiratory issue prediction models to much more sophisticated early stage cancer detection using optical/ultra sound scanners.

    So yes I think AI/ML inference on the edge will be the next big thing on Medical field. Also AI co-processors will be a key selling feature for the upcoming models of MCUs and processors.

     

    -Aswin

  • marketing - AI will figure out what we want before we know what we want, and exactly when we can afford it, and how and where to pitch it to us

    information systems - our phone and its underlying infrastructure will augment our brains - providing whatever information we want, whenever we want it and maybe providing it by the time we decide we want it - in augmented reality - interacting like an AI person with unlimited resources

  • first 7 voters each voted for a different area. I think this is a first off where the range spread is so broad.

    I voted utilities.

  • You need an all the above option.

     

    As the technology matures it is being used in many places, regardless of the risks for doing so.

     

    DAB