Hey guys, well I have spent a couple of weeks re-building opencog/docker to Ubuntu 20.04 and ROS "noetic" release, concentrating on re-vamping the "eva-ros" container initially.
I ran into various issues with old python code, easily upgraded. Also some issues with conflicting configuration parameters, I think those are all solved or solvable.
The issue that has become insurmountable is the change in the Blender implementation from 2.79b to 2.80rc3. The key change was they dropped support for the Internal Renderer in favor of Cycles and later Eevee.
I have tried to convert the textures using the Materials add-on they supply, but it failed to convert the eyes, and was less than perfect on the skin tones as well. If this one issue could be handled by a skilled Blender user, we might be able to "save & load" up to the current Blender release.
The current Blender release has the distinct advantage that it runs in an Ubuntu 20.04 docker container "out of the box". The older blenders, including 2.79b do not. There has been a lot of churn with X11 acceleration, 32bit to 64bit, and the new Wayland support, and the default Ubuntu 64bit setup confounds the older blender versions.
With a properly running Blender, the Eva project could be re-vamped. Is it worth it? All I can convey is my enthusiasm for the project when I trailed it back from the "Sophia the robot" interviews. Whilst I understand that basic research into AI learning does not need the heavy load of supporting robots or human like interaction, I strongly believe that having such a teaching example fully functional would do wonders for the entire field: many people are inspired by the hype around AGI and androids,
Providing fledgelings with a setup such as was developed for Eva would be a powerful kick-off for anyone wanting to dig into the field, especially young people, who would definitely get a kick out of running their prototype AGI hookups in the Eva rig. There IS something inspirational when she comes to life, breathes, smiles etc. Granted the rig is NOT itself a demonstration of AGI, but the hookups back to the AtomSpace should allow people to make all sorts of experiments that point in that direction.
I think the Eva project is a great landing place for enthused talent. It just needs to get over this current hump, which has defeated me at the moment.
Perhaps someone in this group has access to some Blender people we can engage to remedy this situation?
Regards,
--Mark