I am Hardcore? No... WE are hardcore! Code: github.com/devanhurst/ktane_v... Follow me: / devrph Charlie was developed in Ruby using Pocketsphinx for speech interpretation and espeak for TTS.
I mean, it kinda makes sense. The human, while imprecise, values their life more knowing that they can't simply have their memories copied and loaded onto a fresh body in the event of their demise.
Not to mention the fact he's talking really fast with little space. Sometimes I had trouble understanding him. And then we got Siri and Alexas shity voice control which can barely understand you when you tell her to play a song.
Okay so I suppose I will kill the fun by saying the ai here is taught to recognize his voice only and understand bomb defusal commands (which are not that many actually), while home assistants have to hear different accents from different people and have thousands of commands to understand?
@@Faithy2386 The only AI that this kind of bot implements is ironically in the speech recognition algorithm. The actual bomb defusing instructions are probably just the manual instructions hardcoded into the program.
@@whynotanyting I wonder why he chose to go "white wire, black wire, black wire, white wire etc", that's just more words for the speech recognition to trip over. Why not just go for "white, black, black, white, etc"?
I thought the same thing watching this. I looked at the manual though to see how you figure it out, and it's actually pretty easy to memorize what buttons you have to press for Simon says so long as you have no strikes.
That's actually very standard for mid level defusers, and even some low level ones. If you don't have any strikes you only have to memorize two maps, both of which are extremely easy to remember. Hell I almost exclusively play expert and I can do it.
pretty sure there is a bomb in the base game that you cant defuse with the expert cause it takes to long to watch the simon says(there are like 6 simon says modules), so you have to memorie the patterns
If his computer is fast enough he could use image recognition, since the modules are clearly visible, that way it would be without hacking the game, thus more legit. But that kind of bot would be way harder to make since this one is "just" finding values in the defuse manual based on what he tells it.
Nocturne I feel like that might be way too iffy even if it’s plausible. There’s probably an easier way, the only reason I suggested the resources accessed was bc that’s not hacking the “game” per say just looking at tasks.
-So...the test that allows us to know if the bomb works had positive results. The test that allows us to know if the average man can survive a 12cores explosion...not so much. This was a triumph
+dragonddarkPSN I'd say about 25 hours. I spent 15 making algorithms that solve the puzzles and building a text-based interface. Wasted a bunch of time on error-checking for the text commands that ended up thrown out when I integrated voice. Maybe spent about 10 getting speech to work and optimizing it.
+AirplaneRandy "Airplane" Randy wants to make a toy bomb. It'd be a lot of fun trying to ship that over plane. Or better yet, bring it as take on luggage. TSA would have a ball with that.
Wow, this seems like a code that would take months to develop or even years. Not only that but the amount of clarity he would need for the computer to not hear something else and just go down a hole where nothing is going right! Good job.
I mean, it's probably just a state table. If "Button" then "button", wait for input - If "red, detonate" then "press at mod 5 seconds" Still very impressive though. And you are technically correct (the best kind of correct) in that it has taken speak recognition software years to get this good.
I'm really curious: Which voice recognition software did you use? It seems to work really well. You might have customized it as it only needs to understand a tiny range of words and answers, but customizing VRS sounds like a lot of work. This is impressive, glad I found this video :)
Ok, I thought this was like a machine learning thing, but this seems to be just the bot accesses information in the manual. Still pretty impressive/fascinating, especially considering you got NLP going without a hitch
Imagine this same type of technology, but for REAL defusing bomb drills. A robot doing all of the work, and if he fails, at least no lives will be lost, just experience learned. Think about it. (P.S: I’m still amazed at how a bot helped you to give you instructions on the game!)
I love the part at 3:50 where it thinks you gave it a button, and now you have to defuse the bot so it doesn't get more confused. This would be hilarious to do to a friend if you have a miscommunication! I suppose it's so you don't accidentally tell the bot that it misinterpreted you and it deletes its memory about a module that actually did exist? But the silent acceptance of the mistake is so funny.
+Austin Paxton The repo is public, and you could find it on github pretty easily seeing as I use my real name. :P I'm working on refactoring a bit and writing a readme, but you're welcome to fork it now if you'd like! Thanks for the feedback!
Amazing video dude! Is the speech control transferable to other applications? I just watched a video were a youtube made a moving ipad holder for a handicapped child. It also works with speech control, but his is very rudimentary! Would be awesome if you could help him out. Please check out the channel "fynn kliemann". It is the most recent video. You will understand it probably even though it is in german.
Cool! Is an internet connection necessary for this library? Because I think that is a limiting factor for the device, it needs to work offline with a rasberry pi.
You will need an internet connection to download it, but once you have it on whatever device you want, it should work with no internet connection. It's a fairly low level library though, so you'll need quite a bit of knowledge about how speech recognition actually works before jumping in.