WelcomeA A A
New features added to the Forum:
- Private Message System
- Answers Topic (indicating that a post answers the original post in topic)
- Post Thanks (allowing the thanking of posts and a point system)
- New Search Engine (displaying search results by post instead of by topic)
I have been looking at what people are doing with Jarvis themes (Iroman's suit computer AI) that are made to work with Rainmeter (Rainmeter is a windows app that lets you use and make custom themes). Some of these are nicer than others, some are used with speech recognition and one even seems to have some AI. I'd like to make one of these until I can afford Denise to train my speech recognition. I need to find a way to add a form of AI into this. See the videos and you'll see what I'm after.
This fellow made his own Jarvis voice and even seems to have some AI. have a look. http://www.youtube.com/watch?v=4RRexvv_vvk
I have no idea if this is just a gimmick to fool people or if it's real or what type of AI he uses. He doesn't seem like a scammer but you never know.. This video is 2 years old and I have recently requested he send me a copy of his work as he says he will in the video. I will let you know if I get a response from him.
Here is one showing just the voice commands while searching the web.
I want to create a combination of the two above programs. Like the first video, it doesn't have to have the Jarvis theme, just the functionality. Most of these guys use the Microsoft speech macro app to set up the commands.
Question: I have instructions for getting Rainmeter, a Jarvis theme, the Microsoft Speech Recognition set up and the Microsoft Speech Macros app but how to add Alice or any form of AI? (I assume I wont hear back from this fellow above) I am using Windows 8.
For those who want the Jarvis Rainmeter theme see this site for instructions: http://www.dgboost.com/download-jarvis-theme-for-windows-7/
Edit: I have no idea why the post display's one Youtube link and not the other or why it's breaking up my sentence above with spaces between words.
That's a really good find and seems do-able using macros and keyboard shortcuts tied into voice commands.
Unfortunately, the technology that powers the real Jarvis in the movie does not exist. Least not publicly. The problem with doing this with Denise using older systems is the fact that natspeak and dragon are resource hogs. I have 8 gigs of ram in my laptop and it still weighs down my system from time to time. If I disable dragon and natspeak, Denise runs fairly smooth.
I too will look into this. Perhaps maybe some macros can be created to better assist Denise.
Thanks for the reply Dominique but I think you've missed my point of the post. <grin> I want to make this.. something to tie me over in lieu of Denise since I can't afford her yet. This will get me learning about and practicing training my computer with speech recognition.
What I need to know is: Is there a way, for a Windows PC, to incorporate a form of AI such as ALICE or ALICE 2.0, into such a project?
That's why I posted it in this forum instead of the dedicated forums for Denise… because it wasn't Denise related.
I have no skill at programming – since I gave up on Visual Basic years ago LOL. Basic HTML is all I can do – this was stuff before tables were popular and there were no scripting languages for HTML. Think about 15 years ago.
I understand Pandorabot is a web hosting service for bots. That's no good. I want something run solely from the PC without having to rely on a server. There are times I may not be connected to the internet and still want this project to function.
If I could learn AILM for offline use to use with my small project, I may consider it if I have to but I'm a user not a programmer. I see tons of people spend months programming their bots – I have even read many of your posts on this subject – and I believe it should not be so. That's the developers job, not the users job. Guile for instance or another developer should have already spent years doing this work before they release their chatter bot with AI project. Verbots suffered from this as well. The devs didn't do the work and thus they left it up to the user community using the SDK to build the knowledge bases. I was hoping someone would spend a few years making a large knowledge base (teaching the bot so it's AI would be more human like) that users could use but it never happened. I'm afraid much the same thing is happening to Denise from what I read – these are the things users wait on until the devs can produce a bot that really has strong contextual learning. I know that's light years away from PC AI today. The large knowledge base foundation has never been built unless you include Alice 2.0 – which I understand isn't even available to someone like me.
( I do understand I need to spend time teaching the bot things I need it to know from my perspective, but that Knowledge base I'm talking about would be a general encyclopedia of knowledge. Teaching the bot extra things from my end would happen in the course of everyday conversation, not via scripting or using an SDK.)
That said, would learning AIML help me incorporate ALICE into such a project as I have described for offline use?
I believe it is a bit unfair to the developer community to say no one tried to build a very large knowledge base. The trouble is that the complexities grow exponentially as the number of categories increases. And at some point, the law of diminishing returns kicks in. You are doing more and more work for stuff a user will find less and less often. Also AIML has fundamental limitations as to the matching of context; in the parsing of the current sentence and in the analysis of the last 3 or 4 sentences. No AIML engine even looks at a entire paragraph for overall context; too many variables to analyse in 5 or less seconds. This is why a "2nd level AI layer" is so crucial to the Denise project. AIML by itself can not hack it. New AI algorithms are required to improve the understanding of context.
To answer your question, yes there are a number of Offline AIML open source engines you can choose from. Search the ALICE web site. AIML 2.0 has signifigant improvements in context matching. Denise is being converted to that new version now. If you want a custom system, you are going to have to learn AIML. There is a teach mode in Denise, but it is still taking 'baby steps' and is not nearly as robust as programming AIML directly. I would get a book about AIML programming and research whether you want to get into doing it. I can general say AIML programming is very challenging, it is a lot like writing a book. You need to be a good story teller, you need good attention to details, You need to understand the principles of programming, and especially understand issues of indexing. That is to anticipate what questions or ideas your readers / users are going to have about what you are saying.
I may have to wait till Denise is fully loaded with AIML 2.0. or just practice my speech recognition without an avatar bot with AI. I wanna clarify the reasons why I don't want to be the one scripting a bot.
It's all about Magic. If you learn a magic trick it takes all the mystery out of it, the fun, the sense of wonder. I believe these "learning" chatbots should learn in the course of normal conversation with the user and any encyclopedic knowledge base they have should be done by the developers because I don't want to know whats behind the curtain. I don't wanna know the Wizard is a short bald man with bad breath. LOL.. I bought the secret to the Magic Linking Rings when I was 12. I have regretted it ever since.
I have programed a small Interactive Fiction ( text based adventure) game using the Adrift engine and after I finished it, I had no interesting in playing it – I already knew the outcome.
Thank you for explaining why people have had trouble in the past trying to make such a large knowledge base. I disagree with your statement "You are doing more and more work for stuff a user will find less and less often." (assuming I understand it correctly) I don't think anything is a waste of time, it's all knowledge that may come in handy at some point even if it's old knowledge. In fact, I think it's more important for a bot to be more chock full of older knowledge than newer knowledge or current events. These are the things that would help the bot seem to have a fuller richer personality – assuming the language used is good enough to simulate real learning and or contextual references.
In any event, I don't expect this AI bot to be a great conversationalist.. I just want it to integrate with the speech recognition and or macros in a seamless manner.
For instance the macro will call on the bots name lets say Cindy as part of it's operating procedure to open a web browser. I will say, " Cindy, open Firefox" Of course the PC will really be executing the command But, Cindy will hear me say this and respond in kind with: " Opening Firefox" and so on. There may be a better way to do this, depending on other responses I get in this thread about software others know about. I'm not a programmer, just a user and am looking for a way to make this work for PC. I can always to this without any avatar at all, but I like the idea of an avatar to help provide the illusion of the PC having a personality. This too may spoil a tiny bit of the magic but I can live with that to get something that works.
Thank you for your very thoughtful response. Yes every AI developer wishes for their AI to have breadth and depth to it's understanding of the world. (Known as the "Knowledge Base" KB) But the truth is there are only so many hours in a day and you only have XXX number of day's to complete the project. Also at a certain point details get to a fidelity that the AI Can't handle anymore. Beyond that point it is futile to add more details, the AI can not "understand" the differences anyway. Actually avoiding these situations also makes the AIML coder an 'Illusionist'. You keep the user away from these situations and misdirect their attention elsewhere. Skillfully using anthropomorphism will make the AIML bot appear much more intelligent than it actually is.
Regarding your last paragraph, I think you will find that Denise is the only product that will fulfill your design goals. Her various bells and whisles (when fully implemented) will provide a very robust illusion of intelligence and maximize anthropomorphic user perceptions. And of course she will have some ACTUAL intelligence for controlling PC applications. No one else to my knowledge is developing a real time animated avatar fully under the control of the AI, and able to perceive the real world and interact with it / you. (Well, except the Japanese that is)
Thank you for the reply and expanded explanation on the limitations of modern AI in creating a large KB. It will be interesting to see how well Denise functions once she's fully updated compared to other popular PC AI.
For now, I am just going to concentrate on learning SR and using Macros for voice command and forget about adding AI to my PC until I have better options – that is until I can both afford Denise and until she is fully updated with the AIML 2.0 (and other features)
I am still on a fact finding mission. I can use Microsoft SR with or without Microsoft macros, or I can use Dragon Naturally Speaking with or without it's macros but don't know which would work better as of yet. I understand my version of Dragon Naturally Speaking is Premium and it may have some macro limitations Professional doesn't have. The Microsoft macros may be of fuller use because they are designed to work with Windows – but I don't know that yet.. finding such details is hard. I don't want to spend hours training on and setting up macros on one system only to find the other system is better for my intended purpose – (I wont be doing a lot of dictation which I know Dragon is better at, just voice commands). I have to weigh that, as well as the amount of resources Dragon takes up compared to running the Microsoft SR and Marco engines.
I'm on a 400 dollar quad core AMD A8 4500M Trinity APU 1.9 to 2.8 ghz, laptop (AMD Trinity APU's are cool – they only scale up in speed when they are required to do so depending by the process, so they can run anywhere between 1.9 to 2.8 ghz) with 8 gigabytes of ddr3 ram with HD Radeon 7640G graphics that has 512 dedicated memory but uses up to 4 gigabytes shared system ram for extra video processing. It was purchased as a cheap gaming laptop. In that respect, it plays all my new modern games very well with good frame rate so I'm happy with it – But I understand Dragon can be a real ram hog. I haven't tried to use Dragon yet for such extensive purposes.
Coming soon: Leap Motion Controller coupled with Denise could prove to be closer to Jarvis than anything currently. However, it is Denise's lack of ability to control machinery that keeps her from being anywhere near Jarvis's abilities in Iron Man. At this point Denise cannot control and maintain the systems within a smart house. By the time the technology is developed a fuctioning android will be able to do everything Denise can and more, and smart houses will have a Denise like controler built-in.
Programming an AI is not at all like doing a game. There is no real end to it and the goal is more elusive than you can imagine. AIs at present are developed for a specific fuction. Denise's primary fuction is seach (databases, web, etc) and display (run software, display info, etc.). Minor fuctions are learn and respond to user(s). Denise simply could not fuction as a salesman or a teacher and is not really designed to be a stand alone application completely disconnected from the world wide web. Certainly, she could be programmed for this kind of fuction, but that is not the true AI that all us programers seek. It is that spark that would make an AI human. With a game you know how it ends, but with an AI the end goal is just the beginning.
Currently an AI reacts to speach, spoken or texted. But, speach is a small part of human communication. We communicate visually much more than we do vocally. We identify the meaning of each word used and adjust our response based on the context in which each word is used often deriving the context from visual cues from the other person. This is simply beyond the ability of any AI at present. AIML uses the <topic> tag to adjust to the perceived context as assumed by the programmer based on what he understood you might mean by what you might say. However, the <topic> tag is just that a topic tag and the AI is limited to patterns stored under a certain topic without searching the internet. If the AI sets the wrong topic because it failed to understand the actual context you have a minor problem a human does not often encounter when dealing with other humans. However, for a programmer to solve even some of the simplest AI problems that keep them from becoming more human-like would be a big big deal, and it still would not even scratch the surface of what an AI could become and will become in the future.
Most Users Ever Online: 73
Currently Online: cdburns
Currently Browsing this Page:
Guest Posters: 4
Newest Members: ncouponcoupon, eonlinelinks, epromperfect, edgebounty, ehnkelonn, sonsonc, Chezlula, nejeanadol, mrbigherc, linodeetermarce, mottomnamr, Luis Monteiro, Hermie, mastcole974, mastcole
Moderators: Dominique (2334), captquazar (2771), Niko_Mage (44), adam007g (74)
Administrators: leandro (4), Guile Lindroth (1062), igroupaccount (289), Marcus Vinicius (0)