SILVIA Conversational AI Platform

Hello all,

My company has developed an AI system called SILVIA that has some state-of-the-art conversational capabilities, as well as a rich set of tools for developing SILVIA brains for deployment within various applications.

We have recently licensed our technology to a company that will be using our Unity version of SILVIA for their applications, and at least one other licensee of ours is interested in expanding their license to include a Unity version.

In the process of “packaging up” this Unity version for deployment to our licensees, we are now much closer to finishing up a “commercial” version of SILVIA for Unity developers.

This post/thread is intended as an initial stab at getting feedback from the Unity developer community, and to gauge overall interest in our system as an enhancement to Unity.

Thanks for your time.

no link ?

To really judge how interested we would be in this, you’d need to give us at least an overview of what the system will do, especially relating to game development. If it’s a general conversational AI like some of the ones on the web then I can only see limited uses, but maybe there are specific features I can’t foresee.

The OP now has the link to http://cognitivecode.com

Sounds expensive.

Alright… you manage to smooth out the voice and there might be no way to realize you’re talking to AI!!

I’d love to speak with it, sounds freaking amazing. You know what’d be great, to somehow use that AI to make game characters interact with the player, that’d be revolutionary. I mean, in games conversations are pre-generated, but with this ai type both computer and player speech would vary according the context and would never be the same, right?
The only thing is that the voice is still choppy as computer generated voice always have been since I have memory. Does this AI take too much cpu to work?

We are very interested to leverage this tech with current projects in the pipeline. (Virtual training)

Please keep us updated on the status of this remarkable addition to the Unity toolset!

Currently, yes.

However, part of our idea behind commercializing the Unity version of our SILVIA technology is to make it more affordable for wider range developers.

We can’t commit to any particular price point until we do more market research, this thread being part of that process.

Gaming is an important market for us, and talking in-game NPCs are an obvious fit for our technology. :wink:

Regarding performance, the SILVIA Core runtime is VERY lightweight, both in memory usage (as little as 1MB per Core) and in terms of processor load.

We have virtually instantaneous response times from our AI, even on ARM-based mobile devices such as the iPhone. In the context of Unity or other 3D engines, having one or more SILVIA Cores running in the background has no appreciable effect on game performance.

That’s very impressive! (and exciting) If you ask me, this is not only great for human-ai interaction, but for ai-ai interaction as well, imagine the posibilities :smile:
Can’t wait to see what people do with this.

It sounds interesting, but I also would need more information.

Vimeo is blocked here in China, and the fora.tv presentation streams too slowly to be of any use.

It would be helpful if there was some text to read through, and if support for Unity is being added, why not have an AI available through a Unity web player?

Very interested! Have you decided on licensing - ie per app vs single purchase? I saw the SDK is cross platform. Is the IDE Win only? Can you give us more info on how SILVIA works with Unity? A video of that certainly would be of interest. Cheers and congrats!

Hi Pete,

For the commercialized Unity-specific version of SILVIA, we haven’t yet decided on the runtime distribution licensing model. However, for the IDE/SDK, we’re looking at a per-user one-time purchase.

Our most likely scenario is, you could build as many products as you like with the SDK, but you would have to pay Cognitive Code a licensing fee per unit of your product sold. Whether this license is a flat fee per runtime (adjusted based on platform), or based on a percentage of your product price has yet to be determined.

I know that having runtime licensing isn’t quite as clean as a pure one-time or per-product purchase, but our goal is to offer most, if not all of our feature set to Unity developers, without devaluing our larger B2B customers’ existing licenses.

Re. integration and platforms:

Our SILVIA Studio IDE runs on both PC and Mac, and our runtime library, the SILVIA Core for Unity, runs on PC, Mac, the web player, and iPhone/iPad.

At a high level, how it works:

SILVIA Studio is used to create/edit/test SILVIA brains, which are composed of simple and complex behaviors and linguistic/conceptual data. At any time, you can export these brain files as a Unity-specific package to your working Unity project folder.

Runtime integration of SILVIA with the Unity IDE is very easy, via Unity script access to the SILVIA API, and we have full-featured examples that will help developers quickly get up and running with attaching conversational AI to Unity game objects, including characters.

When the time comes, and we’re closer to a “SILVIA for Unity” product launch, we will make available the documentation and video demos for that product.

Cheers

I think it looks amazing and am very interested in using this for RPG games.

Thanks for the answers CC. Sounds as slick as it looks. My 2c on licensing: there’s nothing wrong with a runtime model. A flat fee being affordable for smaller devs, like myself, will depend on the final price point. So, I can’t say whether that vs percentage is more attractive. Just some feedback. Looking forward to what happens with it.

I can make the following limited comments:

  1. Our current B2B software price is a significant multiple of Unity’s price, although we sometimes absorb some or all of those costs if a customer or partner contracts a significant amount of custom development work from us.

  2. We believe that an attractive base price for “SILVIA for Unity” should be significantly less than the cost of Unity itself. When I say “base price”, let me clarify by saying that we are considering a two-tier “indie” and “pro” model, similar to Unity’s.

Before you ask, I’m NOT saying SILVIA for Unity “indie” would necessarily be free. Just inexpensive enough to make it a no-brainer if you really need conversational AI in your game.

That sounds very reasonable considering the current pricing of add-ons and prefabs for Unity. I imagine the research and development involved in creating your software was/is significantly higher than most if not all Unity add-ons/prefabs so that general price range would give you a high quality to price ratio.

Sounds really great. I certainly wasn’t expecting free :slight_smile: The projects I have in mind would require 3rd party investment to bring to market. Paying out a 5-6 figure flat fee to prototype in order to secure funding would be tough to swallow for my size shop. In that case, percentage of sales would work better. That’s really what I was getting at. Anyway, it was just some feedback since you were looking for it. I’ll be watching. Thanks!

Just to add to the feedback. I think this is great and has a lot of potential in my realm which currently is Training and Collaboration in a virtual world. making a more dynamic world that responds to your speech input would have huge benefits. However the ROI (Return on Investment) for adding a system like this to a virtual world for a large (50,000+ employees) company like the one I work for would be hard to justify.

In other words if I spend lets say 100,000 USD a year on virtual worlds and can recover that cost each year via the virtual world (saving on travel expenses for online virtual training is one example) Then by adding this speech system I am not adding to how much I can recover year over year. I am simply making the system more user friendly. That doesn’t significantly change how many people will use it, it simply changes HOW they use it. (Training is mandated so more people wont train just because this system is in place) SO if this system has a significant cost that will create a huge barrier for adoption. Why would I spend 20x (just a guess) more than I spent on the engine for a simple add on that may or may not increase my potential ROI. That is just a significant risk I think.

An alternative might be very significantly reduced cost for prototyping licenses where someone like myself can make a prototype with little or no investment using your system and send it through the project management to get buy in from the project leaders and above. This would reduce the risk on our end and if all goes well it would allow you to sell more licenses and it would give me a more robust training environment.

Just my 2c.

Thanks for the feedback. In essence, I believe your questions/points are covered in my previous post.

But I would like to add that we do currently offer lower cost prototyping licenses for R&D within an organization.