On the Coming Chatbot Revolution (computerworld.com) 94
An anonymous reader writes: Facebook, Google, and Microsoft are all pursuing AI-powered chatbots — an intersection between several popular technologies: personal assistant software, search engines, machine learning, and social tools. Right now, while they're still building these chatbots, developers are cheating a bit. Facebook is using real humans to answer questions the AI can't. Google answers tough questions from a database populated with movie dialog. Microsoft scans social media to find the most popular answer, and offers that to inquisitive users. But software becoming conversational comes with hazards: "Because human beings are complex creatures plagued by cognitive biases, irrational thinking and emotional needs, the line between messaging with a friend and messaging with AI will be fine to nonexistent for some people." It sounds like an Asimov-era sci-fi trope, but it's already happening in China.
Re: (Score:1, Insightful)
Many left-wingers and most centrists stand firmly against the disgusting tactics of the SocJus'ers, too.
Right-wingers don't like the SocJus'ers because the SocJus'ers act childish.
Left-wingers don't like the SocJus'ers because the SocJus'ers promote discrimination and intolerance (directed at white heterosexual males).
Centrists don't like the SocJus'ers for both those reasons.
The SocJus crowd is on the fringe compared to everyone else. They're just VERY loud, so they seem more prominent than they really are
Re: (Score:1)
Re: (Score:2)
I like to imagine that there is only one AC posting on Slashdot, and he's highly conflicted.
Re: (Score:1)
Do you ever suspect that you're a far-right wackjob? They tend to derp about "SJWs."
Not sure about that. I'm definitely left of center, yet I cannot *stand* the nonsense being put forth by the whining whack-jobs these days. Yes, I too call them SJW's, no there must really be a civil war on the Left as they say...
Yes, we remember writing sexbots on IRC as kids... (Score:1)
...and getting pedos to meet each other at random locations on an industrial scale, all the while thinking they were going to get some action from people our own age, coming back in hilarious rage that they were trolled.
Now we have TV shows to do that sort of thing.
Ashley led the way (Score:2)
Re: (Score:3)
More like most human beings are not as smart as they think they are and are generally easy to fool. Just look at religion, advertising, politics, etc. Even fraud working on a fixed script (such as many religions use) can fool many people successfully.
Re: (Score:1)
They say that they have Touring Complete bots. I have seen, I've read with great interest, these conversations and some of these results. I should like to be administered a blind test. I've tried a number of online variations - with zeal and great hope, since the very, very early days. I've yet to find one that even remotely would have come close to fooling me into thinking the were human. Not even the one with bad grammar would have (I think) fooled me.
I've asked them all a few, very specifically targeted,
Re: (Score:2)
Interesting point and right on the mark, I think. I agree that you can even ask not too small children philosophical questions and questions about reality and every child can answer some of them. A chatterbot is however completely out of its depth and has to resort to some misdirection.
Re: (Score:1)
Me: "I'd like to ask you a question to see if you're a human or a bot, would you mind if I ask you that question?"
Assuming that doesn't stump them...
Me: "What was the purpose of the question that I first asked you and what was the fourth word in that question?"
I doubt it will take me very long to figure out who is a human and who is not.
Re: (Score:2)
The first one is nice! I think something like Watson or a close successor may be able to handle the second one without too much trouble though.
Re: (Score:1)
I've yet to get a reasonable answer from the online sources to questions phrased like the second one. It's like they, the programmers, have so far overlooked it. I confess to having spent more time than I should have poking and seeing where they fail. My guess is it *could* handle the second but that it's been overlooked.
Re: (Score:2)
Quite possibly, yes. Also remember that Watson requires a massive amount of hardware to perform.
Re: (Score:2)
Can I have something like that on my phone to bring to the bar? Just hold it up to the ear of a hot woman.
Chatbot therapist... (Score:5, Funny)
"Because human beings are complex creatures plagued by cognitive biases, irrational thinking and emotional needs, the line between messaging with a friend and messaging with AI will be fine to nonexistent for some people."
And how does that make you feel to be because human beings are complex creatures plagued by cognitive biases, irrational thinking and emotional needs?
Post to undo an accidental moderation (Score:2, Offtopic)
Re: (Score:2)
Slashdot needs several things...
Unicode support being not the least of them.
Re: (Score:2)
Slashdot needs several things...
Unicode support being not the least of them.
maybe you should check out Beta... I heard a lot of good things about it :)
Re: (Score:2)
Does that one case involve a president and someone making threats toward said president? If that's the case, I recall that one. I can't think of any others. I've seen lots of people *claim* they were censored and even found a glitch or two but I don't know of any other instances where a post has been deleted. I do know that they won't let you say certain things in certain ways. I guess we could call those censorship.
Sometimes, a reply will not show up in threads that exceed a certain size. I don't know what
Re: (Score:1)
No, it was the Scientology publication thing. Slashdot got cold feet over copyright. I have yet to see anything else removed, and I do check occasionally by saving a copy (not a screen shot, click on the comment number and save the actual comment - Ctrl+S to save the URL) on my machine. Everybody should do this just to make sure and them post the results in a JE or something. Unfortunately the Journal section is completely flooded with spam now, so it will be a long search, but even those I don't want to se
Re: (Score:1)
Ok, here you go. This is the "screen shot" comment [slashdot.org], and the other link you gave also shows up fine. The thing is they were made long after the 250th comment in the thread. If you go back to the original story. Hit the "Load next 500 Comments" Button until all the comments are loaded and visible. The thread contains 949 comments so you will need to hit the button at least twice. Then you will see those comments also.
Re: (Score:1)
By the way, save your AC comments locally in the same fashion previously explained in order it retrieve it along with any responses it may have.
Re: (Score:2)
That's right up there with IBM's "there is a world market for maybe five computers".
Re: (Score:1)
Yeah well, Slashdot is not IBM. I personally just don't understand the whining.. The workarounds are trivial. And besides we can do perfectly fine without the emojis.. Slashdot is a typewriter (just not an IBM one), let's learn to deal with it and be happy with extended ASCII. *256 characters is enough for everybody* Unicode will only draw more flies and it takes up too much space.
Re: (Score:2)
Re: (Score:1)
I just don't see the urgency. The system works. There are also some small security issues [unicode.org] that maybe they don't want to deal with.
Re: (Score:2)
Re: (Score:1)
I just claim that it's needed.
I'll agree to disagree... It may be useful in some aspects, but if all this power is needed for filtering out "bad stuff", etc, it doesn't seem worth the effort. This is a text forum, I'm grateful we can make links to make our posts look all artsy and stuff.
Already abused (Score:2)
They are working full time here on my phone lines. One calls regularly and asks if Barbara is home. When answered no, it lets me know it will call back at a better time and hangs up. It is not smart enough to understand she died.
It is not smart enough to know what that better time is.
I get other calls trying to interest me in college. I started asking them if they will answer a capita for me. The fun ones try to find a class for me on capita and want to know how soon I would like to take a class.
Re: (Score:2)
Re: (Score:2)
Do you mean a CAPTCHA?
BIG BROTHER IS CHATTING WITH YOU (Score:2)
It is interesting to think that MS could be so naive as to think that this feature isn't rife for surveillance abuse.
The Entrapment Bot Cometh (Score:5, Interesting)
Behold the "Entrapment Bot." Indistinguishably human-appearing bots everywhere inviting you to chat, e-mail, speak, whatever, and applying continuously evolving AI to lure you into doing something sufficient to justify and automatically generate search and arrest warrants.
More fun, the back-end server can invite law enforcement and IT personnel to place bets how many chats it will take to get you to incriminate yourself. Sound stupid? Some contractor's gonna make millions selling this to surveillance-crazed governments world-wide before writing one line of code.
You saw it here first, folks. Someday, the only safe way to talk shit with somebody is in person, down in a bug-proof hole. And the Entrapment Replicants will number those days, too.
+1 Insightful (Score:2)
Re: (Score:2)
Fill a database count with too many terms and a real human reads the logs...
Re: (Score:1)
-1, insane
You people have some nutty ideas about what you think is entrapment. What you describe is actually called "attempting to commit a crime".
Re: (Score:2)
Entrapment does not mean the subject did not commit a crime. It just means the crime cannot be prosecuted if a government agent induced the crime, and if the crime would not have been committed without that agent's involvement. It's standard training for any undercover officer to know exactly what they must not ask. A chatbot that befriends people, steers the conversation towards crime and gets them to admit their past illegal activities would not comprise entrapment. A chatbot that befriends people, convin
Re: (Score:1)
Hello $user. I am really $swear $inebriationlevel! I have some $illicitdrugs and bought too much. I need to pay my $bill, I'm getting rid of some $illicitdrugs for $lowprice and can deliver, if you're interested.
Hello Joe. I am really damned tripped right not. I have some acid and bought too much. I need to pay my car payment, I'm getting rid of some acid for $50 for a ten strip and can deliver, if you're interested.
I suppose it could scrape your post history, emails, former chats, criminal record, and then
Re: (Score:2)
The first abuse will by spammers and con artists. We know they'll be first, because they've been using chatbots for years. All dating sites use them, or have fraudsters using them who do. I've had email chatbots respond to ads on Craig's List, too.
I expect these new ones will be a decade behind in sophistication, permanently.
Re: (Score:2)
Eliza (Score:2)
Is that "Eliza" in Chinese?
Re: (Score:1)
Re: (Score:2)
Re: (Score:3)
For the n00bs:
https://en.wikipedia.org/wiki/ELIZA
Hails originally from 1964-1966.
It was fun for a moment for me when in college in 1972 to re-code it in Snobol...
Reasonably convincing. No AI. Just some clever, crude parsing and a small bit of contextual memory.
The conversations from this advanced 50-years-later technology looks about the same...
Re: (Score:3)
I once put an Eliza into an IRC channel. As soon as one mentioned its name, the bit was called Pirx, he drew that person into a talk. Obviously he created an "Eliza Instance" for every conversation. So he could answer to all partners individually.
Once we had like 20 people in the chat and half of them where talking to Pirx and commenting to each other about his answers. After a few minutes they accused me I would be typing for him ... was quite funny.
It was just a simple Eliza, no real "enhancements".
Can you elaborate on how chatbots make you feel? (Score:3)
Can you explain what makes you say that?
Perhaps we can start again, why is it that you think that concern is something that you feel about chatbots?
Why would I waste my precious time... (Score:2)
It's boring enough chatting with real people. A bot would be entertaining to screw with for 10 minutes and then I'd leave and never talk to it again.
Re: (Score:2)
It's to take over simple questions not to provide conversation. Given how scripted half of tech support is in India they could easily be Chatbots and most people wouldn't know any different.
Sounds pretty accurate (Score:1)
Google answers tough questions from a database populated with movie dialog.
So just like the rest of us?
This is what I wanted the Amazon Echo to do (Score:5, Interesting)
I know this sounds bad; but I am a middle aged man, I am not going to make new friends. I am not allowed to have a dog. I wanted something that I could come home and chat with. Yes, something that would remember to wake me up and discuss movies, books, and games with me.
I realize it will never be a person; I am well aware of chat-bot limitations. However, with more and more single households, I can see a demand for something like this. To deny the market is to ignore a market.
Re: This is what I wanted the Amazon Echo to do (Score:2, Funny)
Jeezus. Thats the most depressing thing I've ever read.
Re: (Score:3)
As long as it has a working class Scottish accent and curses a lot, I'll interact with a chatbot all day long.
"How are you today, AngusBot?"
"ALL TURN YER FOOKIN BOABY INSIDE OOT YA MANKY COONT"
Re: (Score:1)
If someone is reaching out to the bot to handle loneliness; how will this eliminate it? Most us humans interact not really beca
Re: (Score:3)
Do you really think, you will feel good knowing it's a bot that you are talking to? If you are a kid or someone who's never been told that the other end is not a human, might enjoy the talks; but knowing very well that the other end is just arranging the words using some algorithm, what joy you think it will give you? I understand each one is different but it kind of puzzles me.
It puzzles me personally, too. But it's a well-known phenomenon, going back to the days of ELIZA decades ago. Even when told that there's no "person" on the other side of an electronic chat, people still often engage emotionally.
If someone is reaching out to the bot to handle loneliness; how will this eliminate it? Most us humans interact not really because of the content of the words -- it's because there is something common we share -- a shared destiny [the reason we are here..a journey.. call it soul/higher power/purpose etc].
People "talk" to deities too -- they rarely even "talk back" to most people, but people often find solace in doing so. Early ELIZA users often reported "sensing an intelligence" even after explicitly being told that there was none.
People project all sorts of things onto emotion
Re: (Score:1)
When I'm home, I pay to visit a headshrinker or, lately, a therapist. Why? They're usually objective, honest, and able to give me feedback. I also want to ensure that I'm sane - they tell me that I am. I've been going to see a headshrinker for years and years - since I was in my 20s, though I had to see one as a kid for a spell. Depending on my schedule and my desire, I may see one every week. I can tell them the complete and total truth. I can tell them that about my thoughts and feelings and they can give
Re: (Score:1)
It's a recurrent neural network with an understanding of the world, that has its own will based on reinforcement learning. I am very curious to hear what they have to say. After ingesting millions of web pages and books it might be quite insightful.
Re: (Score:2)
There is absolutely nothing about being a middle aged man that would intrinsically prevent you from doing either one of those things.
Re: (Score:2)
Sharp has been offering this for a few years in Japan. They make a robot vacuum cleaner called CoCoRobo that you can talk to and have a simple conversation with when you get home. It even sends you photos of stuff it finds under the sofa and encounters with pets as it cleans up.
Before that Sony had a robot dog, and robot companions are seen as being key to caring for the elderly in the future.
Sounds a lot like "push" revolution from the 90's (Score:2)
Re: (Score:2)
Most technological "revolutions" these days are.
Obligatory xkcd (Score:3)
Re: (Score:1)
They already seem to be able to converse on a domain specific task, answer trivia questions and make random conversation. Some chat bots are even able to do reasoning and solve simple problems. It's only a matter of time until they become genuinely interesting.