back
11 comments
Their architecture doc, http://forum.botlibre.org/forum-post?id=11890569 : 'The Bot's Mind is defined by its Thought processes. The Mind receives input objects from its sense, which are queued in the Memory's active memory. The Mind has both conscious and subconscious thought processes, which can be thought of as threads.'

Is this for real, could an expert perhaps comment on the sophistication of this compared to the state-of-the-art? Love the Love.java, https://github.com/BotLibre/BotLibre/blob/master/ai-engine/s...

It seems like an old way to solve the problem. I would be quite surprised if it gave good results compared to recent advances such as the LSTM-based paper “A Neural Conversational Model” [0]. In fact, I doubt it is even competitive compared to Cleverbot.

It does try to compare itself with existing alternatives, but it does not characterize the pros and cons of each alternative and does not compare results.

(I am not an expert, but did take classes on the subject of pattern recognition and machine learning.)

[0]: http://arxiv.org/pdf/1506.05869v1.pdf

For anyone getting up to date with language AI, taliesinb recommended this paper as "a real eye-opener" in the discussion about Google's Go AI https://news.ycombinator.com/item?id=10984973 :

"A Cognitive Neural Architecture Able to Learn and Communicate through Natural Language"

http://journals.plos.org/plosone/article?id=10.1371%2Fjourna...

It uses a more complex (I daresay more advanced) model than the LSTM paper you linked, a model that can be trained to learn and use language constructs (in his words, "mimic certain kinds of child-level cognition with TINY amounts of training data"). Definitely worth a read.

This ANNABELL model doesn't require much training input to go from blank slate to conversation-capable (1587 input sentences -> 521 output sentences; see Appendix S1 for examples). This high learning efficiency might imply that it indeed resembles language-processing architecture in humans (which it was modelled after). At the low level it's also neurons, but organised and connected into specific, planned subsystems.

A key point is that the central executive (i.e. the core, which controls the flow of data between slave systems) is a trainable neural network itself, which learns to generate the "mental actions" that control the flow of data between slave systems (like short-term and long-term memory), rather than rely on fixed rules to control the flow. This apparently allows the system to generalise.

+1 Thanks for the link, I just added this to my eval list.
Fascinating reading. This is probably of interest to the wider HN crowd.
Source code: https://github.com/BotLibre/BotLibre
Thanks. Recently, I have been looking for the source.

I've looked at bots from Program-E which I think has since stopped in development and also Program-O which the source isn't that great to look at and also slowed in development since 2014?

Therefore I have been building my own bot which is aiml 1.0 in Golang, just to learn. Apparently, there isn't one, probably due to lack of interest?

I will be interested to see what else these guys have done.

As another commenter pointed out. It would be interesting to combine ML with the decision making of the response and base a future response on the context of the conversion. It's something I am eager to tackle myself when my own bot is more fleshed out and Aiml 2.0 compatible.

Testing a bot from these guys, show me they still have far to go: http://www.botlibre.com/bot?instance=145&dynamicChat=Chat

So for example:

You: Hi

Bot: How are you?

You: I am Fine thanks.

Bot: I don't approve of your objectives, but I love your methods.

This is a failure from my perspective. The bot should know the context and when I say I'm fine its relating to the greeting.

Really, I am looking forward to see if any progress in this area can be made.

> Free Open Artificial Intelligence for Everyone

... with access to computing resources.

No kidding. This engine has over 90 Java classes! It uses Ant for its build system (state of the art in 2004), and seems to require at least Java 5. I doubt that this could run at all on anything less than a Raspberry Pi 2. Glad you're helping all these microaggressors check their gigaFLOPS privilege.
Unless you're reading Hacker News Print Edition, I think that's assumed.
NOT AMUSED STOP PLEASE CONSIDER THE ENVIRONMENT BEFORE PRINTING THIS TELEGRAM STOP