Integral Mind
|
Lines of Evidence
Traditional technologies, and statistical AI in particular, can't be proven logically - while we can observe them operating in specific cases, we never know what future performance will look like. Success in one context might not hold in any other.
In contrast, this AGI has been proven in every way in which it is possible to do so. Comprehensive evidence has been provided across all possible lines - successful operational deployment, academic proof, social proof, and logical proof.
Because the logical proof shows that necessary properties hold in all cases, it is sufficient to prove the AGI for all future problems - seen or unseen.
Operational Deployment
As shown below, our AGI has been fully demonstrated to work in the real world. All core capabilities (simulation, including of the human mind, causal reasoning, explanation, decision support, info fusion, and so on) have been successfully demonstrated within every branch of the military (excl. Space Force), DARPA, IARPA, the Intelligence Community, State Department, and other governments and organizations.
Academic Validation
Full peer review has been provided for a large number of working example cases in top venues including AAAI, ICDM, KDD, Neural Networks, IEEE Symposium Series on Computational Intelligence, HumTech, and Cognitive Science. These papers show the system working on a wide range of real-world problems.
Social Proof
In addition to multiple government and contractor attempts to 'break' the system (which they could not do), one of our core Pentagon customers exhaustively studied a sample problem and conducted a full review of all properties of the system so as to be able to convince himself that the system did indeed work as indicated and that all properties held. This review was successful in all respects, and subsequent to this both he and and his immediate supervisor (Ph.D./Special Forces) executed documentation attesting to our successful achievement of AGI.
In addition to seeking to invest $1MM, this customer also delayed his retirement for some time so as to be able to continue to work with Integral Mind.
Beyond this, based on deep reviews and successful performance, the AGI was selected by DoD technologists for inclusion within the Defense Intelligence Information Enterprise (DI2E) Reference Architecture.
Subsequently, senior Office of the Secretary of Defense officials requested that we work to get this technology to specific identified parties as they believed it would be of unique import.
And in one government lab, a team lead wrote that this work was "… a paradigm unlike any extant work … what [Integral Mind] has contributed is a significant breakthrough".
Full Logical Proof
A full logical proof has been provided proving that the system will work in all future cases, regardless of task, for all problems currently known or unknown. This proof is achieved with a simple 2-premise syllogism; for any syllogism, if the premises are both valid and sound the conclusion must be accepted.
The first premise derives a definition of AGI from the intelligence literature and first principles - as the proof shows, there appears to be only one possible definition.
The second premise then clearly shows that our system meets the definition.
Having properly derived the only feasible definition for AGI, and then having shown that we meet the definition, the only allowable conclusion is that AGI has in fact been achieved.
This proof is self-contained and can be validated by anyone.
Selected Past Deployments
Intelligence Community: Simulated and predicted the thinking, decisionmaking, and actual behavior of world leaders over one year at 94% accuracy (previously entirely infeasible).
> Received Agency approval for use in government decisionmaking
Proved the AGI could:
> Accurately, and successfully, simulate the human mind (and by extension, that its principles were strong enough to do so)
> Successfully predict behavior by simulating the mind
> Simulate worldviews using general intelligence information as a foundation
Intelligence Community: Automatically combined multiple sources of information; performed real-time reasoning on questions of interest to the US Government.
Proved the AGI could:
> Fuse multiple sources of information into one common picture
> Conduct real-time thinking and problem solving
> Create explanations crossing multiple siloes of information
Non-Governmental Organization: Provided successful simulation of policymaker thinking in support of advocacy efforts.
State Department: Provided user-centric program design and evaluation services.
Received feedback that our simulations were so accurate that ‘it was as though we were in their heads’.
Proved the AGI could:
> Simulate beneficiary experiences and worldviews
> Apply beneficiary understandings to program design
> Suggest significant improvements to program designs
> Validate programs by simulating effectiveness within fully-accurate virtual environment
Special Forces: Integrated AGI-driven full-semantics Natural Language Processing and AGI reasoning capabilities in order to autonomously determine which foreign-language remote site data would be of interest to the US Government.
What the AGI did:
> Used understanding of the meaning of documents and government goals to discover what information would be useful and why
> Used knowledge to simulate the meaning of words in context and reason about how those meanings would impact the interests of the government
Proved the AGI could:
> Understand natural language, including cases where original texts had been machine-translated into English
> Understand government goals and interests
> Simulate the impact of information on goals and interests
DoD: Successfully simulated cyber systems in support of discovering potential attacks on complex systems in view of offensive and defensive scenarios.
While cyber systems are best understood as causal (cybereffects are created by combining obscure causes and effects) and contextual (as attacks only work within, and create effects within, certain contexts), before this AGI it was not possible to handle either of these in computational systems.
For DoD, we demonstrated that the AGI could represent, reason about, explain, find, detect, and defend against new attacks.
Proved the AGI could:
> Vs. traditional statistical practices, understand cyber as a causal system
> Represent, combine, and reason about all domains of cyber inputs and effects
> Discover new attacks
> Defend against previously unseen attacks, in part by answering questions such as:
> Does this make sense?
> What is this doing?
> Do we want to allow that?
Government Lab: Revolutionized approach to commonsense reasoning and applications in military and disaster relief scenarios.
Proved the AGI could:
> Understand, reason about complex real-world scenarios
> Represent and apply causal, nuanced commonsense knowledge
> Perform cognitive functions in real time in disaster scenarios
> Leverage crowdsourced commonsense knowledge
> Automatically translate knowledge from propositional format to atomic format
DARPA: Performed on DARPA CALO (previously noted as ‘most advanced AI project ever undertaken by USG’ by Wikipedia and the predecessor of Siri from SRI) at Carnegie Mellon (USNWR top Computer Science).
> Developed new natural language processing methodology via Radical Construction Grammar and published paper on new parser for this in Lecture Notes in Computer Science (Springer).
> First system able to actually understand text and apply this in support of cognitively humanlike (semantic) parsing.
> Computer used real-world causal reasoning to discover and implement solutions to administrative challenges.
Proved the AGI could:
> Provide the necessary knowledge representation and semantic understanding, reasoning, matching, categorization, and sense definition services required for full-semantics language processing
> Provide necessary cognitive services for supporting Radical Construction Grammar and full-semantics processing
> Successfully reason on tasks equivalent to those that would be performed by a human assistant, including resource allocation and administrative processes
DARPA / IARPA: Automatically parsed and understood cyber logs via use of semantics without training. Performance beat out two large primes’ statistical solutions.
Proved the AGI could:
> Solely by reference to what made sense in context and in view of how protocols/systems operate, and without rules, autonomously discover the formats of cyber logs, parse them, and prepare these for automated entry into cyber defense systems
> Use proper understanding-based solutions to deliver better performance than that possible from statistical systems
> Enable full-semantics knowledge-based AGI processing in constantly-changing complex contexts
Predicted future cyberattacks by understanding the content of posts on cyber forums and reasoning about the combined effects of these.
US Government specifically requested more of our time on project and noted that we were key personnel for the project.
Proved the AGI could:
> Use language semantics to identify words and phrases of interest in cyber forums
> Combine multiple semantic data points to discover emerging semantic patterns
> Determine if effects are to be expected, and, if so, which effects
> Reason about effects and severity
> Rate and explain expected effects
FEMA-offshoot tech group: Built full-semantics Tweet understanding pipeline in support of emerging situation detection, sentiment detection, and real-time reasoning support for incident commanders.
Proved the AGI could:
> Apply linguistic and commonsense knowledge in order to determine stories told by incoming Tweets
> Determine what was happening and the severity of this
> Determine what actions were required in current moment, including what problems existed, how problems could be solved, and severity
> Compute changes and alerts with respect to the foregoing
Predicted core details of first Iran nuclear agreement (published in HumTech 2015 proceedings with cite to New York Times)
Proved the AGI could:
> Simulate human thinking and political processes to the extent necessary to determine likely negotiation positions
> Simulate the content and outcome of negotiation turns
> Converge on negotiation outcomes matching those that took place in reality
Developed AGI-driven advertisements and persuasive messaging
Proved the AGI could:
> Simulate viewpoints and perspectives
> Understand what semantic elements may or not be reasonably expected to lead to persuasion (and why)
> Create semantic elements capable of persuasion
The foregoing only summarizes some of the previous applications of the AGI. Information on many others, however, can be found in the peer-reviewed literature and in materials accessible only to US Government personnel, including but not limited to: messaging, simulation, emotion, psychology, social media, cross-domain nuanced representation and reasoning, nuanced representation of cultural concepts and worldviews, persuasion, fine-grained representation of social science phenomena, peacekeeping, conflict framing, International Relations, cross-cultural interaction, psychological and cultural effects on decision-making in uncertain, emotionally-charged environments, negotiation (including culture and emotion and their effects on negotiation strategies), full-semantics language processing, cognitive linguistics, construction grammars, cognition, and multiple other topics.