codetrap wrote: ↑Feb 18th, 2011 1:58 pmSo you advocate replacing a pharmacist with a purpose build $31Million dollar supercomputer that can play Jeopardy that's roughly the size of 10 refrigerators, and costs who knows how much to actually run. Sure. Sounds like a great plan. Also, there's one thing a pharmacist does that can not currently be replaced by a computer. Confirm understanding of instructions. All you have to do is look to IVR systems to see how poor they are. Most people hit 0 to talk to a real person rather than try to navigate the options and talk to the computer. Think an automated drug dispensing machine is going to be any better? Think that machine will be able to converse with your Grandmother and make sure she really understands that she can't take her meds with her morning grapefruit, and her glass of pomegranite juice, creating an interaction that will kill her? Yeah, I can see that happening.
Until the state of the art gets to the point where an AI actually exists (which it doesn't), and that AI can understand and interact on a meaningful level, stuff like automated drug dispensing is not only stupid, it's dangerous. Anyone that advocates it with the current state of the art simply doesn't have a clue what they're talking about.
IVR's work that way by design. Saves call routing times for 95% of callers. Chances are by the time you press zero you've been funnelled a few levels into the IVR. Multinationals have finely tuned algorythms that make this process as accurate and cost-friendly as possible for them. (I have >10 years of experience in BI creating these algorythms). PS: Most people do not hit 0. A very very small minority does.
I'm not talking about building a physical supercomputer like Watson. Watson was built to be a self-contained database to more closely abide by the jeopardy rules, I'm talking about the level of AI being made available. Why else would IBM invest so much time energy and resources in the WCG?