#03 - 2022-wk48-R1-A1 - Google Sheets

	A	B
1	Timestamp	12/6/2022 22:34:52
2	What is the title of your project?	Avatar 2 Human Learning: Net Zero and Carbon Negative Emissons,
3	How are you and your project doing in general?	Advancing quickly on the immersive project front. Progress is steady. Machine Learning experts are in high demand, we are a little delayed here. 3D modelling talent we're finding in new graduates, giving them an early opportunity to work and for us, faster model prep than we can do.
4	What have you accomplished in the previous period, and how would you describe your current status?	We've combined the four on-site worlds into one application and have added three new 'rooms': 1. A lobby for our Anya-01 avatar, which presents an introduction to Carbix's X2. In the near future we expect to connect Anya to speech-to-text and TTS to enable freeform conversation. 2. Scene choice room, to access the immersive worlds in an interesting way. 3. X2 Showroom - with "new car" lighting - to present X2-specific lessons that would benefit from a simpler environment. We've also created avatar animations and lip sync examples to improve upon M2's avatars. These behaviours will be replaced by the API service we will create. The Iceland Geothermal plant has improved terrain and surfaces. The X2 model has been improved by a recent game development graduate in Vancouver who specializes in architecture. The model's responsiveness and surfacing is much better. We've added a geothermal 'hot rocks' environment in a tropical zone. Lastly we have a lean strategy document which puts together in one place discussion of the technologies and lesson creation innovations we are looking for.
5	What are the next steps you are planning?	We are now very much into the Adaptive Learning development module. First priorities are: 1. Create an ontological structure for the industrial, chemical, and sales instruction streams. 2. Establish a dataset for the Knowledge Graph recommendation engine training. 3. Prepare a Hello World API link and an MVP for our lesson recommendations. 4. Present a demo in one of the immersive worlds of recommended lesson sequences. 5. Show how avatar behaviours might be assigned to text copy via sentiment analysis.
6	Are there any specific events, challenges or successes in the previous period that you would like to elaborate on?	Finding available ML talent is a challenge. We're seeing that ML experts are quite busy. Of the 12 individuals we've talked with this period, 8 are too busy to assist. Two weren't working in our discipline, and two didn't have direct experience. Therefore the commercial company is our best current choice, and could prove effective partners throughout our project. Adaptive Learning AI might be described as a mix of Recommendation Engine (such as Youtube or Netflix uses) combined with evaluating student progress and this appears novel. So we're starting to describe our project in commercial terms first, then bridge to the lesson-specific focus. Converting the models from precise engineering standards of detail to game-ready versions is proving a larger job than expected. This should be outsourced to talented graduates. Having all the worlds in one app reduces duplication of assets, is much easier to manage, and is a better experience. There is a bit of work involved in standardizing the appearance, but the scenes were using basically similar assets.
7	Do you need any help or support? (Please describe the kind of support and who can best provide this)	ML advisement on which networks/which algorithms is helpful, but we do understand everyone available is over-committed. On the other hand this means we're working in an exciting area of development. Suggestions for 'synthetic' training databases for our focus, if available, is also very helpful. Thank you all!