Our Team

MLabs AI sees AGI as a very important strategic goal, and our senior technical team already devote a portion of their time in pursuit of this outcome. We see our collaborations with SingularityNET as an important part of this endeavor.

Company Name (if applicable)

Project details

There are a number of large, generic semantic graphs (e.g. WikiData, DBpedia, BabelNet, and so on), with an ever-increasing number of domain-specific knowledge graphs appearing every month. In addition to these more traditional graphs, there has been a move recently to using LLMs for the production of knowledge graphs.

While this capability provides a useful shortcut to a difficult problem, there are several problems with using LLMs for knowledge graph extraction:

• inconsistency - LLMs often fail to correctly resolve entities, and the output is non-deterministic
• accuracy - they can miss valid predicates, or (more often) hallucinate invalid predicates
• context - LLMs often fail to maintain a frame of reference specific to the knowledge domain
• confidence - LLMs display the same level of confidence about valid and invalid conclusions
• speed - knowledge graph extraction from even a single website can take many minutes, and is impractical for very large knowledge graphs

Our KG tooling roadmap assumes that most of the structured graph information being operated on will be noisy, of variable quality, and not subject to ontological best practices. It is likely to have been created manually, using ad hoc processes, or from the output of an LLM.

We have designed requirements that specifically address the concerns listed above. Our tooling needs to maintain a suitable frame of reference (and to identify multiple frames of reference in large KGs), with contextually resolved entities, have relationships with quantified confidence, and be able to operate computationally efficiently at the very large scales that will be required in AGI systems. There are four types of tooling:

• Context Identification
• Entity Management
• Predicate Management
• Confidence Management

This is one of two KG tooling projects submitted to the current RFP and concentrates on context identification. We will open source not only the tooling developed during the project but also the tooling worked on thus far. We will continue to develop these tools as an important component of our AGI roadmap and may deliver further enhancements to the Singularity NET community too.

Frame of Reference (Context) Identification

AGI systems will have to deal with multiple frames of reference, with different entities, predicates and conceptual nuances operating in each frame. When applied to knowledge graphs this can be reduced to three main tasks:

• Discovering sub-graphs
• Identifying frame-of-reference entities
• Assigning predicates to frames of reference

Discovering sub-graphs

Graph partitioning is an NP-hard problem, and so optimal sub-graph discovery for the large KGs that will be in operation is likely to remain technically infeasible on current hardware. There are a number of approximate methods for graph partitioning that scale suitably for large graphs but are still computationally expensive for KGs of the size we are dealing with (many millions, to hundreds of millions of nodes).

MLabs has proprietary IP in the form of a novel algorithm for efficient, approximate partitioning of extremely large sparse graphs. The method is based on Szemerédi’s Regularity Lemma and exploits the same algorithmic finesse as MergeSort. It has thus far remained an MLabs trade secret. We intend to open-source this algorithm, which is guaranteed to run in O(n logn) and produce partitions that compare favorably with those found by eigen-decomposition of the graph Laplacian, but orders of magnitude more quickly. The same algorithm can be used in recommender systems with trivial modification; MLabs will not restrict its use by the SingularityNET community for either purpose.

We will demonstrate the efficacy of frame of reference identification using our algorithm on a KG based on WikiData. We will reuse this dataset for other aspects of the project.

Frame-of-Reference Entities

Word co-occurrence graphs have long been known to possess the small-world property; they have a high clustering coefficient and can be divided into cliques. We believe that this is important for AGI efficiency as it enables analysis that is local to a frame of reference - it is worth noting that the connections between neurons in the cortex have a topology consistent with this observation. We have established that large knowledge graphs have similar properties and that nodes can be assigned to frames of reference using their frame linkage statistics. This process can be initialised using the sub-graph results from above to ensure that the assignment is computationally efficient.

Frame-of-Reference Predicates

One of the statistics that helps us assign entities to frames of reference is the type of predicate linking the entities in the frame. While some predicates appear to be universal (instance_of, subclass_of and so on), there are a large number of predicates that are specific to a subset of frames. For example, integral_of is specific to the algebra frame, subsidiary_of is specific to the corporate law frame. So the assignment of predicates to frames, and entities to frames is a collaborative inference task for which we employ Expectation-Maximisation to find a (locally) optimal assignment matrix.

Assignment of entities and predicates will be demonstrated using the same WikiData KG as in the previous activity.

Open Source Licensing

Background & Experience

Dr Bedworth is the Chief Scientist of MLabs AI with over 40 years of experience. He co-authored a 2000 AGI paper with psychologist Carl Frankel. He worked with Nobel Laureate Geoff Hinton on Boltzmann machines, and his work on Bayesian probability received worldwide recognition. Two of his patents were acquired by Apple for use in Siri.

Ibrahim received his Masters Degree in Computer Engineering and Machine Intelligence from the American University of Beirut. He specializes in optimization, search algorithms, and deep learning. He has worked at MLabs AI on novel approaches to deep learning and is currently leading the LLM workflow.

Dr Sarma received her Doctorate in Data Science and Artificial Intelligence from the Indian Institute of Technology. She specializes in NLP and reinforcement learning. She has built context embedding models, and LLMs for a number of languages, and is currently a member of the NLP team at MLabs AI, as well as handling our reinforcement learning activities.

Links and references