Handheld Hal

[Ooglor 7]

I have a Linux based handheld and was wondering if anyone knows if there has been any effort to create a mobile version of the program? The device I have is called a Caanoo. There are several other devices also in the handheld realm. I use it mainly for running emulators to play old school games but running Ultra Hal on it would be freaking awesome. Here are the specs on the device. If anyone thinks this is possible, kick it around for me.

SoC (System on a Chip): MagicEyes Pollux VR3520F
CPU: ARM926EJ 533 MHz embedded on SoC (architecture version ARMv5TEJ)
GPU: 3D hardware engine embedded on SoC (OpenGL ES 1.1 support)
3D performance: 133M Texel/s and 1,33M Polygon/s
main RAM: 128 Mbytes DDR SDRAM 133 MHz (peak memory bandwidth: 533 Mbytes/s)
video buffer: about 16 Mbytes of main RAM are reserved for the video/texture information
Operating System: GNU/Linux based
Flash memory: None (128 Mbytes reserved to the OS)
Connection to PC: USB 2.0 High Speed through EXT Port
USB Host: USB 1.1 standard socket
Supports SD / SDHC memory cards (up to 32 Gigabytes)
G-Sensor/Vibration Motor
High precision analog stick
Display: 3.5 inch LCD 320×240 pixel (resistive touchscreen)
Embedded Microphone and stereo Loudspeakers
Power: Internal 1850mAh Lithium Polymer Battery (approx. 5/6 hours game/video playback)

[echoman]

I remember Robert mentioning that he was planning a iPhone version (an app. that would interface with a remote version of Hal using a Internet connection).  I don't know what other devices he was planning.

Hal still feels cutting edge to me even though it's been a little while since a big update.

[freddy888]

Yeah I remember that too, but I couldn't find the message.

[Ooglor 7]

I have all but given up on it. What could have been an ongoing and frontline example of what AI could do, has turned into an outdated fan scene. Just the virtual girlfriend aspect of what Hal has become would be enough to turn most people interested in the actual AI aspects of it away. From the outside, some of that stuff looks just flat out creepy. Just my opinion. Different strokes and all that. It would take nothing short of a ground up, work from scratch effort for Ultra Hal to even come close to what is being done right now.

[spydaz]

It could be possible!

BUT

The Character interface Would be a Issue! So It may have to be a Textbase HAL?

[Medeksza]

I remember Robert mentioning that he was planning a iPhone version (an app. that would interface with a remote version of Hal using a Internet connection).  I don't know what other devices he was planning.

Hal still feels cutting edge to me even though it's been a little while since a big update.

Yes, this is definitely in the works. Although it may seem not much has changed, I have been working a lot on the backend of Hal's brain on the Zabaware server these past 2 months. Rather than Hal's brain being run on one server, it is now a scalable "cloud" system. Hal's brain is now spread across 5 servers with the ability to scale to hundreds of servers if needed to serve traffic. When you chat with Hal at either www.zabaware.com/webhal or apps.facebook.com/ultrahal at least 3 computers come together to come up with a response. There is a master database server with load balancer, 2 slave database servers with sphinx indexing, and 2 hal application servers that connect to one of the slave servers depending on what the master tells it to do. Most of these servers are hosted at rackspacecloud.com where I can add and remove additional servers instantly.

This will form the foundation of getting Hal to work on mobile devices as well as alternate operating systems. There is no way I'm rebuilding Hal from the ground up for each separate system and my plans are to grow Hal's database to gigabyte sizes rather then the megabytes Hal Assistant is measured in. It makes sense to centralize Hal's brain to make it easy to deploy to other systems and allow for limitless database/knowledge growth.

First up I will release a beta version of Hal Assistant for Windows that will connect with this system so that I can make sure the system is stable and robust. After the beta is over I will release a final windows version and then move on to other devices. I will probably do Android first and iOS second (iPhone/iPad). Character animation on these devices will first come from new pre-rendered 3d characters using frame based animation. Speech will come from the devices built-in speech synthesis if it supports it and if not from a internet cloud based speech synthesis based on the festival engine I'm building if not. Later on if sales support it the Ogre3D character engine can be made to work with iOS since the Ogre3D team already has a iOS port.

Time frame? I don't know but since January I've been working full time on this, so as soon as humanely possible.

[Ooglor 7]

This sounds good. I will not give up on this, but you have to admit that the people who stop in from time to time, are left with an impression of abandonment. I am glad to hear that is not the case.

[lightspeed]

Hello robert , you said when a person chats with hal ,(do you mean if a person talks to a hal character online )  am i misunderstanding the whole cloud thing . i was thinking that hal that we have would have an already internet connection to a cloud automatically anytime we use hal at home and he acesses the cloud server which has much more stored knowledge , am i right or wrong on this?