A.I secretaries : Pros and cons

We already live in a world where you can ask you phone to do things like look up something on the internet or put something in our diary.

Fast forward slightly to a point we all have a A.I on our person, in our homes, at work. A personal secretary in software form.

What are you thoughts on what we would gain from this? What would we lose? Just because the technology can (and probably will) exist, should it? Can companies be trusted not to abuse this technology?

Most obvious downside: exploitable vulnerabilities that need patching.

Well, there'd obviously be less jobs for secretaries.

What on Earth would I want a secretary for? Especially a fake one? Can't even have an affair with that, so why bother?

Chances are you have AI on your phone already. What level of AI are we discussing for the future?

edited 27th Jul '12 10:34:28 AM by Michael

capable of having a conversation and a higher level of understanding than a cheap call centre (basically, able to understand a wide variety of accents, slang and possessing a degree of initiative)

edited 27th Jul '12 10:47:55 AM by MCE

On one hand, It'd be terribly convenient. Something capable of reminding me of all the various things I need to do, prioritizing things according to how important I would think they are (Webcomic Updates > Work Deadlines) and helping me to find things I want to find..

Basically, a smartphone++

On the other hand, when the robots decide to take over the world, I'll already be at their mercy.

This is exactly what I need. Something that can learn what I want kept track of better than I know myself. The problem is privacy. I dont want it sharing my profile with anyone.

Yeah it'll inevitably be yet another backdoor to your personal information for whoever made the thing decided to close a shady backroom deal with.

Not just personal information- if it's a smart AI, it will know me better than I do. And so will Google Ads, if it isn't set up right...

Well it'd be more useful than out-sourced call centres if you can make it work (no wait time to get to the crappy service that corporations offer).

PRO: A program that is "sentient", learns and adapts to me as time goes on and becomes an extention of myself. This will VASTLY improve my ability to function, speed up tasks like paying bills, filing forms, booking flights, calculating and presenting data in a form I prefer. Even remembering aniversaries and events. A godsend.

CONS: AI is a crapshoot threat obviously. If this thing goes off line, I'd feel like someone lobotomized me, as I'd be cut off from something that was part of me. Whoever built/maintains the AI (it sure as hell won't be me) will effectively have me at their mercy. Should someone hijack it, I'm doomed. My accounts, contacts, personal data could be compramised in one go.

The problem isn't even that your information could be compromised. Obviously an actual secretary can go saying stupid things or something and do the same thing.

The problem is that with a mass-produced AI, compromising your information to other companies would be an unlisted feature.

I'd just like to say that your system doesn't get compromised like that and AI is not a crapshoot. Stop using Hollywood hacking! :P

It'll be roughly the same as logging into a bank account, except your program is offline so it's even more difficult to hack, and it just manages your information like a super calendar.

What I'd like:

• It figures out my most routine purchases and either composes a schedule for me (such as buying groceries at a single place because I'm lazy but do so in a manner that reduces expenses at the quality level I want)
• It can conduct online purchases for me at the cheapest price for goods I'm waiting on but don't want to get until they're cheap enough (at the quality level I want)
• Remind me of things I may need to do (pay rent, pay mortgage, get cash for a vacation etc)
• Automatically handle customer complaint issues for me ("HI FOR SERVICE IN ENGLISH PLEASE PRESS ONE", "AUTOBOT PRESSES ONE.")

Well, there are still potential security vulnerabilities in any software, especially any software that connects to the internet (as it would have to do, for example, to make online purchases). These can be ameliorated, but ultimately you do have to trust that the people involved in writing the AI are both competent and trustworthy. Would you trust the people behind Microsoft or Facebook to make an AI secretary that didn't report back home on its own initiative?

Pro: Might actually be able to convince me to get my arse to class and remember my appointments.

Con: Doesn't come with a short skirt or the possibility of after-hours nookie. Also, not something I'd use much. Finally, I generally hate computerized "help" systems, and always turn such features off in any program I can.

edited 27th Jul '12 12:41:22 PM by Ramidel

One problem mentioned here I can expand on, does anyone remember that issue with Siri? Originally when asked "what is the best smart phone?" Siri recommended a Nokia product, Apple got wind of this and changed it. Now imagine your A.I start automatically recommending a certain shop for your products...

(http://www.bbc.co.uk/news/technology-18071342)

(http://news.cnet.com/8301-13579_3-57434326-37/nokia-apple-rigged-siri-to-name-iphone-best-phone/)

edited 27th Jul '12 2:09:18 PM by MCE

There is no way in hell it wouldn't be online. None whatsoever. Like, it would have to be online for the very basic secretarial communications and scheduling you'd want it to do on its own in the first place.

We won't have anything for quite some time that can accurately predict the temperament and habits of the person a secretary would work for, which is a huge part of what being a secretary is all about.

The entire point of a secretary isn't just appointments, it's to do all the little shit so that someone with important shit to do has more time for that type of thing.

I played heavily armed secretary to a General for a week once. It consisted of setting appointments, getting coffee, copying/scanning/faxing, and delivering messages. Now a computer can do some of that, but a computer cannot yet go get you coffee, or take paperwork to a fax machine or copier.

Plus, instead of doing it himself, he could do something like say "Hey Sgt, call the commander of MSG and tell him I can't meet him at 10 for the budget meeting." With an AI, he would have to actually compose a proper message and essentially tell the AI to deliver it, as there is a certain professional manner in which you speak. Instead he could forget about it, and I could get on the phone for a composed 2 minute professional conversation.

There's just a lot of personal touches that you really can't get with a computer, and won't be feasible anytime soon.

While it would probably take a robot arm or two to fetch coffee, anything that can be predicted can be programmed since most human interaction takes place within certain expected templates.

Well I was picturing this was far into the future where your AI wouldn't be a piece of junk.

Besides, you get what you pay (or don't pay) for with the Siri example. Get an AI from open source groups and it'll be fairly neutral, get one from major corporations and you get your problem. The Sobeys (or Fresh n Easy) AI!

Also, it's more likely that you get piece-meal AI for now. One that does your grocery optimization. One that does scheduling. And so on. Until someone combines them into packages when they become good enough.

I'd still rather deal with people.

That is another potential problem, if A.I's can be customised more people may prefer to deal with them instead of humans, this is already happening to a certain extent.

I don't see it happening any time soon though, or at least I'll be old as dirt by the time this starts to happen.

The worst idea ever would be AI human resources, but I can totally see corporations going for it.

Human human resources are already horrible people 9/10 times, all we need now is fucking G La DOS firing people based on stupid statistics.

edited 27th Jul '12 5:25:22 PM by Barkey

Assuming it could get cheap, A. I. Secretaries primary benefit is to people who did not have secretaries now. It could help ordinary worker remember when their favorite sport on TV, their wives birthday, etc. people who already have human secretary will retain more useful human secretary.

I once read that in some asian countries hiring driver is cheaper than installing GPS on car. so A.I secretary success will depend on how cheap it is, and how cheap hiring people as secretary.

Actually, if they're actual AI capable of scientific and statistical analysis, they're less likely to fire people because sales are low, because that shit doesn't improve sales. So, rather than just following orders, it would spit out an error.