9 Jul 2016 |
| suriyadeepan changed the history visibility to "world_readable" from "shared". | 02:37:08 |
suriyadeepan | Hello anyone who has stumbled upon this group. Have you read the book AI : A Modern Approach? Are you reading it now? Post your notes and questions here. Let's conquer this monster together. | 02:40:16 |
suriyadeepan | A short course on Intro to AI by Norvig and Thrun https://www.udacity.com/course/intro-to-artificial-intelligence--cs271 | 02:43:49 |
16 Aug 2016 |
| suriyadeepan changed their profile picture. | 13:04:53 |
22 Sep 2016 |
| Vincent joined the room. | 16:14:18 |
6 Oct 2016 |
| Vincent set their display name to Vincent. | 07:30:35 |
27 Nov 2016 |
| Vincent set a profile picture. | 09:03:05 |
4 Dec 2016 |
| suriyadeepan changed their profile picture. | 15:25:36 |
25 Jan 2017 |
| suriyadeepan changed their profile picture. | 10:16:07 |
4 Feb 2017 |
| @Spacex:matrix.org joined the room. | 05:05:43 |
5 Feb 2017 |
@Spacex:matrix.org | I used to teach the book | 16:29:44 |
Vincent | Any expert in language model? | 16:32:20 |
@Spacex:matrix.org | Certainly | 16:32:43 |
@Spacex:matrix.org | HMM for example | 16:34:24 |
Vincent | Well, do u think deep learning models are the solution to text comprehension? | 16:37:13 |
Vincent | A lot of experts think natural language understanding is not probabilistic | 16:37:59 |
Vincent | For example, Lstm rnn is just a sequence mapping function so it can't understand the meaning without the context | 16:39:07 |
Vincent | Hence, it is only for simple translation only. | 16:39:31 |
suriyadeepan | hello | 16:40:34 |
suriyadeepan | there are complex models built with lstm that considers context, while decoding/answering/generating | 16:41:13 |
@Spacex:matrix.org | I think pure engineering techniques and probability models have already failed to solve the core of natural languages complex problem ..the use of phoneme is the main issue for the failure of previous systems and will not allow more progress toward understanding / cognition or the creation of a complete intelligent agent | 16:48:52 |
Vincent | I heard that semantic network is a major breakdown in this area | 16:51:06 |
Vincent | It will solve the issue deep learning can't for language | 16:51:37 |
@Spacex:matrix.org | Deep learning has made good progress so far but I need the knowledge representation part has not been deeply discussed | 16:53:17 |