3 ms·

The general problem with all "AI-powered assistants" is that they're not AI-powered assistants. They are voice interfaces atop tiny libraries of utilities. If

The general problem with all "AI-powered assistants" is that they're not AI-powered assistants. They are voice interfaces atop tiny libraries of utilities. If I had an AI powered assistant, it's actual data would be on my system, and not out on the internet somewhere. It wouldn't need to send my speech out. But further, it would actually be an assistant. It would be able to operate all the applications on my computer. "Cortana, copy the last three Slack messages in the current channel to a new Box document. Call it Requirements and add the date." -> Uses two separate apps for which I have open sessions, can exercise my credentials (where I complete the MFA when needed). "Cortana, search all my communications for things related to Bob. I need to trace all of his activities that I have record for. Include Sumologic logs for the applications I support." -> Searches chat logs, email, and web application logs out in my log collection. "Cortana, find out from Mom what groceries she needs today, and place Instacart orders accordingly." The "assistants" we have today can't combine actions. They are just isolated pools of menu selections. They can't replace hands-on-keyboard for the things I regularly need to do with my computers or devices. They need to become intention-processors, not just language parsers to activate a menu.

xemoka 6y ago
And not even good ones at that. They never take negative re-enforcement, you can't tell them what they just did was wrong or report a problem. You can't say "Good Bot" and let it know that the response it gave was what you expected and to continue to return results similar. Hell, if it was just 'stupid' you could keep programming new features into it and new ways to get to the same features with slightly different wording. But no. We don't even get that. Just this awful black box that "sometimes does what it's asked" and can barely reference previous incantations. Uhg.

xemoka 6y ago

And not even good ones at that. They never take negative re-enforcement, you can't tell them what they just did was wrong or report a problem. You can't say "Good Bot" and let it know that the response it gave was what you expected and to continue to return results similar. Hell, if it was just 'stupid' you could keep programming new features into it and new ways to get to the same features with slightly different wording. But no. We don't even get that. Just this awful black box that "sometimes does what it's asked" and can barely reference previous incantations. Uhg.