AI Thoughts on the Build Keynote

  • Comments posted to this topic are about the item AI Thoughts on the Build Keynote

  • This is the third Microsoft Build I've attended digitally, and it is also the most fascinating one I've attended. The advent of generative AI has really made this Build much more significant, at least to me. During this long weekend I intend to watch some sessions I wasn't able to attend.

    For me, the most significant thing I saw, which I believe Kevin Scott introduced, was an interaction by a user with a chat client (probably ChatGPT) where the user was asking the chat client questions about some data related to their company. At one point the user asked the chat to generate a table of the data, which the chat client did. WOW! That is BIG!! This is a moment that reminds me of when the Internet first came about. Before then the software I wrote was for Windows applications. When the Internet and HTML came about, it was a huge disruption in how we wrote applications. That demo of a chat client generating a table of data, on the spot at the user's request, is exactly like when the Internet/HTML came into being. I am not one of those who preach doom and gloom to all our jobs being brought about by the advent of AI, so we'll all lose our jobs and must start living under a bridge. What I do see is that soon it will not be necessary to generate a report of data for a user to review. Instead, the necessary skill will be writing APIs, because these AI's will need to be able to refresh the data, they use to generate those tables and graphs on the fly.

    Kindest Regards, Rod Connect with me on LinkedIn.

  • And these AI's should do that efficiently so that their queries don't block other services and users, or exhaust server resources.

  • The demos are interesting, but they're canned and limited. I think we'll get into strange places where the AIs need access to internal data, but we need to be sure they aren't sharing this data outside the company. We also will need to have some ways to verify things and learn to trust the AIs.

    I expect that IT people will end up doing a lot of prompt engineering to keep the AIs working in the way we want them.

  • The demos are interesting, but they're canned and limited. I think we'll get into strange places where the AIs need access to internal data, but we need to be sure they aren't sharing this data outside the company. We also will need to have some ways to verify things and learn to trust the AIs.

    I expect that IT people will end up doing a lot of prompt engineering to keep the AIs working in the way we want them.

  • The one thing I did not understand is if the Windows Co-Pilot is a trained AI assistant or if it is one we have to train (like correcting chatGPT's answers). If it's the former, it will be interesting to see how correct it is. If it's the latter, I might use the Windows search instead.

    • This reply was modified 1 year, 5 months ago by  Raphra.
  • That's an interesting question, Raphra, I don't know. I assumed that it would be the first one, but I could be wrong.

    Kindest Regards, Rod Connect with me on LinkedIn.

  • I suspect it's pre-trained, but I hope it adapts to me over time. If it doesn't, I think it's semi-useful, but I would hope the nuances of how I work are picked up over time.

Viewing 8 posts - 1 through 7 (of 7 total)

You must be logged in to reply to this topic. Login to reply