Jonathan Edwards
Project OwnerJonny will be the main worker on this project. He will do all the research and all the development work on this largely sequentially structured project.
Our company, Temporal Computing, proposes to enhance the performance and scalability of the Hyperon framework by integrating cutting-edge temporal computing technologies with specialized hardware. Temporal computing offers an alternative to traditional and quantum computing by representing input data as a "passage of time" rather than traditional binary. This method allows for processing efficiencies and energy savings by utilizing simpler memory manipulation devices. Our vision is to leverage these advantages to provide a significant speed increase and power reduction in AGI computations.
This RFP seeks to explore and evaluate innovative hardware paradigms, ranging from small-scale IoT devices to large, decentralized systems, to optimize AGI workloads in the OpenCog Hyperon framework and handle advanced processes such as cognitive synergy and hyperdimensional computing. The focus is on assessing hardware paradigms and emerging architectures (e.g. neuromorphic processors, associative processors such as TensTorrent’s APU, etc). The goal is to enhance computational efficiency, scalability, and cognitive synergy in AGI systems. Part of this should involve interacting with the Hyperon team who've built the existing and in-development MeTTa interpreters.
Milestone description: Conduct literature review on MeTTa MeTTa compilers Hyperon technologies temporal computing and the potential synergies. Run workshop with the Hyperon team who've built the existing and in-development MeTTa interpreters. Engage with industry experts to identify priority areas for improvement. Identify key challenges and opportunities for integration.
* Key Deliverable: + Comprehensive analysis of current state of art hardware for respective computational requirements. This will be an academic document of around 10-15 pages with around 100 references. Work has already started on the temporal computing aspect in collaboration with the University of York. We seek to augment and support this in the content and thinking
$15,000 USD
Develop and refine temporal computing insights and algorithms. We already have a start on this process as we have the building blocks for some of the key matrix operations. We will extend these to align with this project brief by integrating temporal IP with existing Hyperon idealogies.
+ Recommendations for integrating temporal technology into the Hyperon/PRIMUS stack most likely building on out advanced algorithms for linear algebra. The main thrust of this work will be integration of the temporal ideology with that of Hyperon.
$15,000 USD
* Milestones description: + Design and develop POC hardware solutions for key AGI components. Because of time pressures this will likely utilise extending processes (open source verilog design) and IP that have been successfully deployed in temporal computing the business. + Conduct rigorous testing and validation of hardware prototypes. This can be done through simulation using cycle accurate simulators such as verilator. + Refine and iterate the design based on feedback from stakeholders. We will engage with other Hyperon/Primus users to establish next steps in moving the POC to solve a beachhead technical challenge.
* Key Deliverable: + Construction of a demonstrator prototype showcasing the potential of temporal computing in AGI applications. + Proof-of-concept hardware solutions for a key aspect deep learning and cryptographic, most likely linear algebra.
$44,999 USD
Run a presentation / final stage workshop on the key findings, opportunities and deliverables with a cross section of the appropriate stakeholders (initially agreed at engagement but could add other parties as appropriate) covering the staged deliverables and the next steps we recommend for optimal ROIs extraction.
Report and recommendations document - workshop day via video conference.
$5,000 USD
Reviews & Ratings
Overall
3.3
4.0
4.3
Topics of this response were of great interest to the judges. Strongly encouraging participation in future rounds, perhaps along the lines of this proposed work.
Please create account or login to write a review and rate.
Overall
4.0
3.0
0.0
Accept! Proposal outlines the problem and generally follows RFP goals and objectives. I like that the team wants to spend time reviewing the codebase of MeTTa compilers and other PRIMUS components to identify the pain points and suggest the improvements for proposed architecture. The downside I am not sure at all that temporal compute technologies will work well with PRIMUS and how much code will need to be changed.
Overall
3.0
5.0
0.0
I like this proposal, and I would like to see it funded. Maybe we can create a followup RFP that this fits into? But this RFP is really for putting together a super careful in depth review of the AGI hardware space, rather than for taking one particular approach to hardware and fleshing out its application to Hyperon in great detail. Related to this, I also am not sure really whether temporal computing speeds up the operations that are most critical to speed up for Hyperon in particular. For Hyperon the biggest bottleneck is access to the memory graph according to complex patterns, which is not obviously super sped up by temporal computing (though maybe it is). I can more easily see how ActPC and ActPC-Geom stuff (reinforcement learning ish stuff) would be massively sped up this way actually....
Overall
3.0
5.0
0.0
Very interesting conceptual idea for compute, but not particularly focused on speeding up the sorts of graph pattern matching calculations we are interested in. It seems temporal computing could mostly speed up addition and multiplication calculations.
© 2024 Deep Funding
Join the Discussion (0)
Please create account or login to post comments.