BotColony

@mb: We did not make any deliberate attempt to mislead anyone. Our company started as a R&D company in AI and Natural Language Understanding in 2001, and until 2007 we had no idea we would one day put out a game. We learned about making games by attending local chapter meetings of IGDA. We don't have any senior person on board experienced in the game industry (although we employed at one point more than a dozen Ph.D.'s). Most of our developers never worked on a game, unfortunately. The Steam store page has a field Reviews, and the same store page format is used for Indie games (that don't have reviews) as for other, AAA games (which do have reviews). Since we didn't have a game before Early Access, how could we have game reviews? The next best thing we had was articles about the concept, the technology, the story, etc. - and we quoted them in the Reviews field, since it seemed to be the only place we could mention them. There's a lot of novelty about Bot Colony - you can shoot, wreck cars, fight in a million other games, but we elected to do something different. So, I think that posting articles about something that's truly genre-defining, to explain what the novelty was, was OK. If you define review as a hands-on gameplay session by a game journalist, what we have there is not reviews, but people do write reviews of films, books, shows, etc. which are conceptual, so it's a fair use of the word. In the meantime, there are gameplay reviews on Steam of people who actually have played the game, and prospective buyers can rely on those.

@mrsmiley: It doesn't require more training, it requires a speech model for bigrams and trigrams, that Microsoft speech can take as input. We don't have the speech model in the version that Vinny used. But the best solution would be to reach a deal with Nuance (it is a commercial issue, not a technical one).

Your points about commands that should have been understood but weren't is well taken. Here's what I posted on that in the main GB thread:

"Shut" seemed like another problem, and when he tried "close the drawer", Jimmy let him know that he was already in "close" proximity. Infuriating.

*** No, Jimmy let him know that it's already closed. I agree that 'shut the drawer' should be supported. Other things that Vinny tried and should also be supported:

- 'more' (we support 'again').

- 'lay down X', 'turn this', 'repeat X', 'speak X', 'slide X', 'turn off X', 'turn off', 'shut down',

- put X back (we do support 'put X where Y was' so 'put the pepper grinder where the salt was')

- 'are we done?', 'what am I doing?' (only what we can tell from the interaction), 'tell me about your commands' (we have 'what are your commands'. Maybe not 'where I click', it's out of the experience).

- We need to improve generation (I am now for the first step ready)

- Vinny was right about the shoe rack, it is an item that had changed between photos.

- When teaching Jimmy a new command, Jimmy insists you teach him based on commands he already knows, one step at a time. Jimmy fucked up simple commands already, there is no reason to assume he would be smart enough to understand a multistep process of picking up the grinders and putting them down in each other's places.

*** As I said, we actually support 'put X where Y was', so that actually does work.

I'd like to explain the time business in robot visual memories business (the 3rd Gamasutra link explains perception-based Question Answering in detail). Where is he? or Why is he not in bed? (I agree this is what the obvious thought process would lead to) may get "I don't know" answers if the robot didn't see or hear anything about that. Then the question is what do you do next. Questions like when did you last see X? will return a specific time, that you can then use to ask a specific question. It works by narrowing down to a time-segment by elimination.

You're making excellent points (about singular paths, testing and community feedback), and trust me that this session and the conversations surrounding it give a lot of fodder for thought and improvement. About voice training - I'd even go to the distance of not making it available if the training score is not good enough. Players can type until they get a precision enabling them to have a good experience through speech.

It's not a review, it's a technical article. If you want a review, this came out recently

http://www.rockpapershotgun.com/2014/07/07/bot-colony-alpha-review/

These videos

https://www.dropbox.com/s/ginpub8ztnmhfae/Bot%20Colony%20Robot%20Visual%20Memories.mp4

https://www.dropbox.com/s/pai63dfy04530gx/Text-to-animation.mp4

show how the game performs when it's played with a trained voice profile. Can you please take a bit of time and actually WATCH them? Of course, you're welcome to buy the game on Steam, check it out yourselves and report actual problems you may find in the Community Hub on Steam (I did acknowledge the ones Vinny encountered). It is a very new kind of game. We did work very hard on it, and I find that Vinny's 'review' amplifies the negative to make points about speech to text and using language in games.

My post in the main thread on giantbomb.com enumerates the real technical problems highlighted in Vinny's session. They're manageable. I feel that Vinny did make the game look terrible, to put on a good show. Using keyboard correction, for example, would have solved 90% of the heartburn, but then he wouldn't have such a juicy show, right? He only used it once or twice at the end. Also, he asked for the robot's commands, but never used the information. Why? To issue unsupported commands?

There were several Gamasutra pieces that dealt with the Bot Colony's potential, concept, story, etc. (the first prototype was shown at GDC in 2009 and the second at E3 in 2010 - it really was a very long haul effort)

http://gamasutra.com/view/news/202072

This is more dated http://gamasutra.com/php-bin/news_index.php?story=22513

and this

http://www.gamasutra.com/blogs/EugeneJoseph/20140626/219765/Natural_Language_Understanding_and_TexttoAnimation_in_Bot_Colony.php

you'll probably find the quote there.