• Steve Jones - SSC Editor - Tuesday, November 21, 2017 8:53 AM

    HighPlainsDBA - Monday, November 20, 2017 2:56 PM

    If the bots are leveraging machine learning to quickly and accurately adapt to their end users preferences then I'm impressed. However, I'm inclined to think that what is needed more is a 'killer' app and not killer bots so-to-speak. Maybe a synthesis of both is the best option? Knowing and planning (good programming) for the usage scenarios that your end-users find relevant makes for successful software (in part at least) so my question is: Do your fancy SkyNet Death bots know what I'm looking for when phrased in my peculiar and oblique style and in the context I intend? Or do I have to try it  15 different ways? How quickly and accurately do they adapt I wonder?  Machines don't do well when it comes to vague concepts, but I can see where searches that leverage your preferences/history etc. can make better guesses as to what I'm looking for but that really isn't all that new or radical. I've read some science articles that describe how ML has greatly improved image recognition programs but does anyone know of software that actually comes close to behaving like a good butler? That's what I need. 🙂

    You might be surprised how many are being used now. They're not trying to replace a person, but replace a narrow domain of knowledge that a person might respond to. Bots are being used by large airlines and financial companies, and they may not handle every query, but they often handle a many, and they respond quicker, making a decision to hand off to a human.

    We use them. We don't get perfect responses, but good enough. Better than the press 1, press 2 nonsense on the phone, and we lived with that for years. Bots are going to take over plenty of spaces, though not completely. What they'll do is give huge leverage to the humans employed in terms of getting work done.

    If you are thinking you'll have the Star Trek computer, or a Bicentennial Man intelligence, you're aiming for a perfect bot. We aren't there yet, not sure we will, but we don't need to get that far. Plenty of humans are employed that aren't great at their jobs or can't handle the interactions. Bots will be different, and have different imperfections.

    I could use more real world education on the subject, have you come across any posts/publications that you would recommend? I'd be particularly interested in how they are employed and managed in a system and what makes them different from all the other loops. Thanks for the commentary Steve and Happy Thanksgiving!