03-19-2017 02:51 PM
03-19-2017 04:06 PM
Not sure either to love them or to hate them!
Chatboots, even with the usage of AI, will have some "hard time" dealing with all the text "adaptations" humans tend to make.
But in another field that concerns me a bit, is the human interaction with chatboots, when they get smart enought to keep a "human-like" conversation!
03-20-2017 08:51 AM
I prefer human interactions, not technologically forced. I'm sure someday I'll be chatting with one and won't know the difference. But still don't like the idea.
03-20-2017 01:12 PM
03-20-2017 02:30 PM
It begs the question why people are building these things if there is a much more fundamental issue at play. Passing the Turing test is a nice goal, but the value isn't immediately obvious.
In today's world, trust and accountability is such a huge focus. Particularly where the organisation is large and complex. Losing all that with one misguided AI response doesn't seem worth it.
03-20-2017 04:09 PM
Because I want my conversations to be with humans not AIs. I'm the type of person if I was contacting a business and they used an AI without my knowledge, I would seriously consider moving my business elsewhere or just not do business with them at all. To me it is deception and if a company does that with an AI, then what else will they deceive me with?
03-21-2017 01:44 AM
To me the strongest argument for chat bots is their use within messaging apps like Weibo and WhatsApp. There a huge growth in the time people spend in these apps, so there's appeal in being able to transact without context-shifting to another environment.
A: Want to meet for a beer?
B: Sure, what time?
A: Trainbot, when's the next train to Oxford?
Tb: 17.05, arriving 17.50. Change at Didcot.
A: Let's meet at 6
Right now I see three drawbacks
1) As soon as the interaction becomes more complex, you're better off with a form. "When's the next train to Oxford, without any changes and with an off-peak fare" is a pain to type, but 2 check-boxes on a form.
2) The natural language processing isn't flexible enough. If you need to remember a specific keyword, you may as well use MS-DOS. It needs to be really resilient to typos and contractions too.
3) Chatbots are basically one-result search engines. Most search tech isn't good enough to always get it right first time.
03-22-2017 02:26 AM
There are so many things that can be misused with AI.
Sure it can process information faster, even read books, papers and so, on in a impressive speed, but! It is still software! And software still has bugs, and vulnerabilities!
Who would be accountable, for some disaster caused by AI ? Something like a trading error that makes someone or some company loose a lot of value ?
And if someone finds a way to temper with some AI software ?
To many questions to little answers!
04-02-2017 04:40 PM
04-03-2017 12:58 AM
I'm not sure I understand that example. Even if you took AI out of the equation and imagined a human PA in the conversation it wouldn't make sense to book a train ticket without some validation that it was needed.
Its not inconceivable that the tech could do it, and it is worthwhile thinking though what we actually want from dialogues like this. However, in terms of near-future chatbots, the costs of false positives are high if it makes a purchase and AI is far from being able to do the necessary inference reliably at the moment.
04-03-2017 04:16 PM