It's the duality—executing & thinking—that makes this project work.

It's only fair that they test us first.

(Yeah it's a joke post.) Next up: The other way around!

Level 1: Literal. Just make some responses. How are you? => I'm good/I have some problems

Level 2: Battle it out. Take the rules of level 1 and have them compete against each other. (There are also meta-rules.)

Level 3: Fact management. Some things are true, others aren't. Manage this.

That's pretty much it. Now we just need a lot of rules to work with. Oh, and there's a very good parser for German & English.

Meeting.

...is an AGI? 'cause we deliver.

Getting somewhere.

If you could talk to an AI about any topic—which one would you choose?

Bot knows (sadly).

Who knows what it will do next?

Channel's waiting dudes.

v1 - Make a basic logic using a bastard language located somewhere between Prolog and plain English (mostly it's just the latter).

input(I'm a scammer!)
  && fact(a scammer scams people)
  && relationX(scams people, scam people)
  => output(You scam people?!)

v2 - Add some $ signs and turn spaces into underscores (making the pattern dynamic).

input(I'm a $scammer!)
  && fact(a $scammer $scams_people)
  && relationX($scams_people, $scam_people)
  => output(You $scam_people?!)

v3 - Find out what relationX is (that's not that hard).

v4 - Interpret this stuff.

DONE

Estimated time frame: 2 days

Edit (2018/08/03): And it's done (see picture).

IBM and all the other guys.

My fabulous English Parser just got better, you can now help it out by using some brackets, like this:

{This time of night} it's usually quite cool.

...even though that place is the living hell for any sane person. I'm not even allowed to ask questions anymore because they hate dissidents.

Shurely an important part.

We're getting there though.

Participate!

"What, you're doing AI with 4 gigabytes?"

Yes, that's right. And 512 MB actually go to the GPU, so yeah... it's tight.

Another working AI component. Takes any sentence or phrase, like:

smart combinations of various fast methods

and outputs:

1. combinations
2. combinations of methods
3. smart combinations
4. combinations of fast methods
5. smart combinations of methods
6. combinations of various methods
7. smart combinations of fast methods
8. combinations of various fast methods
9. smart combinations of various methods
10. smart combinations of various fast methods

Neat, right? And very useful.

There'll be a channel and a bot.

I'm not sure if you have thought this through yet. This is a project that makes computers think.

...and you'll have the dialog engine you're all waiting for.

Input:

"Confessions of a Burglar Video"

(It's the name of a video on YouTube.) So... is the video confessing—or the burglar? Let's ask the parser.

1. Confessions {of {a {Burglar Video}}}
2. {Confessions {of {a Burglar}}} Video

Yup, it has found the 2 grammatical possibilities. Next up: Semantic analysis to rate these against each other.