https://snap.berkeley.edu/snap/snap.html#present:Username=joecooldoo&ProjectName=Clive%20AI
(Name is a work in progress)
Hi! I'm building an AI that can have a simple conversation. I need a list of questions and answers to help train the AI. Anyone in?
A starter question would be "What's your name?"
That's what most people say to begin conversations. Well, most youngsters. As you grow older (This happens with most people...) you get more socially awkward. I guess.
Here's a list:
(I'll add more soon but these are just basic things.)
That's already in the AI, but it could use more responses. I'll soon give a link so people can talk to the AI (That's how it gets smarter)
I do need questions AND answers to make things work all the way.
Are you planning on using a cloud or BrowserDB to store all the names and questions and things?
Also I think you should code some sort of AI which can take questions from other chats online and then learn from those questions. But the moderators say that would be using JS to hack and could lead to a (permanent) ban so I suggest you shouldn't.
Currently, there is no such thing as a real AI. Most AI chatbots use clever hacks to fool us humans into thinking it's a good AI (and thus it could pass, or "pass", the Turing Test)
Answers may vary...
No, there is a spammer who likes to spam any cloud thing I'll make (he spams cuss words, meaning that the AI would learn to cuss).
Good idea, but again, I don't want the AI to learn cuss words if the chat contains any.
Have you heard of the term "API"?
Mine would never get close to that test...
What kind of hacks?
I guess you're right. Because of Tay.
Yes, I have.
Agreed.
So there are a few AIs out there which use clever hacks like diverting questions, being overly condescending, etc.
All my AI does is find similar questions in a database, and answer with the most relevant one. And if it can't find any within a certain percentage of accuracy, It creates a new item for use in learning. Then it occasionally asks questions to get answers for the question that has the least answers.
I put the link in the main post.
I have a question. How did you make the custom category?
C'mon. That was the state of the art in 1966 (ELIZA), but today even the (super annoying) chatbots on commercial web sites legitimately extract meaning, within limits, from what people say to them, let alone Siri and the like.
Seymour Papert used to complain about the "superhuman human fallacy," wherein people would complain that chess-playing programs weren't real AI because (back then) they couldn't beat chess grandmasters. The point is, even then, chess programs were better than most human chess players. (Similarly, you could make a case that commercial web site chatbots aren't any worse than the human customer support people who just read from a script: "Did you try turning the computer off then on again?")
People put too much weight on the Turing test. Turing made that up as the criterion for convincing an AI skeptic, but it's not the criterion for getting actual benefit from AI research; we're way past that milestone.
What's the date?
A: [date]
What's the time?
A: [time]
What's my username?
What's my user?
A: [username]
im sorry but that would be hilarious
Nothing counterfactual about it. That actually happened when Microsoft (I'm pretty sure) put ML chatbot on the net.
I have never heard of ML Chatbot, know Windows devices have Alexa.
Note that I didn't capitallize the "c." I meant that generically. You've probably never heard the name of the one I mean (and I don't remember it) because it only lasted about two days before someone pointed out to management how racist it had become.