Fantastic article, Matt. I used to run a tech support company for seniors and it really hit home. One of the policies we had for the company was that we would absolutely never do online or over-the-phone tech support. The reason why is because it's so hard to do - seniors just can't do it. You have to be there in the room with them because they have trouble pointing at things, using the mouse, or everything else you described.
I've often thought about exactly what you wrote here before. In fact, at one point I led a capstone team at Arizona State University to make an on-device AI that used system information to give more accurate tech support steps and even outperform ChatGPT 3.5. Of course this was a while ago, and the new versions can do much better than whatever we made, but at the time it was the best thing. It could use information like your type of computer, your software version, your screen size, and even your accessibility settings to understand how to best help you.
I really hope to see something come out of this thought.
Jordan - totally resonates with me. Trying to help my dad figure stuff out over the phone was basically impossible, so this really confirms what you experienced. Really curious to learn more about your on-device AI tech support project - do you have any links you can share?
Its an old project, i don't really have anything from it anymore. The basic idea is that you source context from the system info and browser to allow for more accurate support
Fantastic article, Matt. I used to run a tech support company for seniors and it really hit home. One of the policies we had for the company was that we would absolutely never do online or over-the-phone tech support. The reason why is because it's so hard to do - seniors just can't do it. You have to be there in the room with them because they have trouble pointing at things, using the mouse, or everything else you described.
I've often thought about exactly what you wrote here before. In fact, at one point I led a capstone team at Arizona State University to make an on-device AI that used system information to give more accurate tech support steps and even outperform ChatGPT 3.5. Of course this was a while ago, and the new versions can do much better than whatever we made, but at the time it was the best thing. It could use information like your type of computer, your software version, your screen size, and even your accessibility settings to understand how to best help you.
I really hope to see something come out of this thought.
Jordan - totally resonates with me. Trying to help my dad figure stuff out over the phone was basically impossible, so this really confirms what you experienced. Really curious to learn more about your on-device AI tech support project - do you have any links you can share?
Its an old project, i don't really have anything from it anymore. The basic idea is that you source context from the system info and browser to allow for more accurate support