2023-01-12 AI Committee Meeting Minutes

 

Public Page

 

 

 

Date

Jan 12, 2023

ANTITRUST STATEMENT

As participants in this meeting, we need to be mindful of the constraints of antitrust laws. There shall be no discussions of agreements or concerted actions that may restrain competition. This prohibition includes the exchange of information concerning individual prices, rates, coverages, market practices, claims settlement practices, or any other competitive aspect of an individual company’s operation. Each participant is obligated to speak up immediately for the purpose of preventing any discussion falling outside these bounds.

Agenda

  • Welcome/Networking

  • Antitrust

  • Meeting minute Review

  • Review Objectives

  • Review Workflows

Meeting Minutes

  • Antitrust Accepted

  • Meeting Minutes Reviewed Accepted

  • Welcome to 2023

    • Reviewed the Committee page structure on cieca.com

    • Reviewed the Committee Objectives

    • Reviewed the Workflows

  • Meeting Minutes

    • Raj Shared a document he prepared to go over the data acquisition and the standards that go with confidential and security of data.

      • Document has 4 sections which talk about the acceptance of images and videos and what a customer should look at.

        • How the images should be taken, what formats, what kind of pictures and the size of the images, the resolution of the images.

        • What video format should be used.

      • Jonathan agreed that the work he does with AI has similarities to what Raj shared.

        • At this time, his company uses an APP, so the pictures are taken with the app and they do not accept other images.

        • Agrees the standards are implies to collect the RGBD data would be a good direction.

      • All images come from phones, not just iPhones, but any of the mobile devices can take pictures.

      • Question on dots

        • Could we use dots to show damage to a vehicle?

          • Like a stencil to help guide the user into taking the picture correctly.

          • No this would be like a cling that could be placed on a damaged vehicle to help estimate damages.

            • The disposition of the standards of image quality is more around encoding compression except flags, right? So, at any point in time when you're looking at assets that you're trying to label, that's what you're trying actually to identify through a notion of, you know, kind of an attribution. If there is a dot here, you need to look at that that is outside of the scope of what the quality is rendered.

            • These footprints are standards that can be adhered to and formats to the devices. It could be an Android, iOS smartphone, tablets. They should adhere to a certain criterion because of what we need to do is as we go into these domain filatures, we also have to educate our each other as to what, what impact and implications would deem necessary to address.

            • We should adhere standards, right means that if compression is adhered to an image or a video compression, you can likely see the AI disposition being impacted and therefore the decision points being impacted and what those decision points would be. It's more around standardizing, standardizing that across the spectrum of the consumer space or what we are trying to adopt here.

            • Basically, we are trying to bring the kind of and formats or you know the standards that every it should basically work uniformly with anybody out there who is dealing with the damage assessment of the vehicles there. It doesn't have to be a proprietary kind of an you know image types or the technologies because otherwise it's going to be limited only for that particular kind of and use case or a customer.

      • The goal of this committee is to basically determine how do we expand this across the industry? It should be collision, it could be the insurance industry, it could be the rental car, it could be the fleet management, anybody. And that is where we need to basically, you know, enforce this kind of in standards that everybody who has a smartphone should be able to take the picture should be able to take the videos. The only thing is the most important aspect of the AI is what kind of data gives us the best kind of an accuracy? Because toward the end, all it matters is it's accuracy.

        • What makes those accuracies better? It depends on to how good the data acquisition is, and that's where, you know, we need to basically bring all these standards together.

    • Workflow

      • In the past we started with FNOL and there was some changes made to help identify CIECA messages.

        • The question is there is so many different workflows based on do we have pictures or do we not have pictures. Should our messages identify if picture is included?

        • We looked back and pictures were treated as attachments to the message, there were fields that captured the date of the picture, but nothing about size how to use the pictures.

        • Looked at the BMS and found those messages impacted were FNOL, Total Loss, Estimate.

    • Code List

      • Do we need a code list that will be able to be used to show the best resolution of the requirements for a picture?

        • We could have a minimum requirement, but we don’t want an upper bound.

        • It's up to the vendor of the AI provider to what limits they require.

        • There is no limit to the number of pictures.

          • There are 8 touch points on a car, can we limit to this? Not at this time.

        • A 360 degree video is great for AI.

        • There needs to be implementations between the parties to set the standards.

    • CIECA Role

      • We are not trying to set the standards for the number of pictures or resolution, but CIECA is trying to identify the data that is needed and how we capture this and integrate with each other. This is when you're transferring this information between providers. This is this is what it's called. This is how it should be configured.

      • These sessions are brainstorming so we could see what kind of data we need. Right now we're not even coming up with a recommendation for it might come out as a recommendation after we're done with everything but this kind of helps us identify if we need an array or if we need, you know, how many file attachments we need or stuff like that, just to make sure that we can handle that complexity and the BMS.

Up Next

  • Welcome/Networking

  • Antitrust

  • Meeting minute Review

  • Review Raj Document

  • Look at BMS Data

Action items

Decisions

 

Participants

  • @Paulette Reed

  • Ginny Whelan

  • Tim Ronak

  • Ken Eagleson

  • Fred Iantorno

  • Stacey Phillips

  • Jonathan Pyle

  • Raj Pofale

  • Gne Lopex

  • Kimberly Sorlien

  • James Spears

  • Edmond Schneider

  • Williams Sims

  • Greg Jensen

  • Abhijeet Gulati

  • Tom

  • Cinthia Carolina Guiterrez Oliverous

  • Don Porter

  • William Brower

  • David Willett

Participants in the meetings are noted for your information.  If you have questions on the committee’s activities, please contact a recent attendee. https://cieca.atlassian.net/wiki/spaces/CPSC/pages/1371340801