HAIDIE: Heuristic Analytics Integrated Data Interpretation Engine
HAIDIE — Heuristic Analytics Integrated Data Interpretation Engine
In metaphorical terms, HAIDIE rules the roost. It uses near-real-time Machine Learning coupled with the persisted Deep Learning data provided by EMPATHS to predict highly-relevant responses to your question. We do this via a set of proprietary cross-validation algorithms coupled with industry standard TF*IDF and BM25. Under the hood, HAIDIE is a set of distributed micro-services used for ingestion, analysis, association building and communication. TAME is our communication protocol of choice between these machines.
In a sense, ZEN and HAIDIE are two sides of the same coin, one being a server-side and the other being client-side. Their goal, however, is the same: to learn from the user.
In the purest sense, AI is alchemy:
“Researchers, he said, do not know why some algorithms work and others don’t, nor do they have rigorous criteria for choosing one AI architecture over another.”
The above quote is from the chief of AI R&D at Google, while speaking at an AI conference in San Francisco last December. It was a Moment of Truth for the industry, and he received a 40-second standing ovation for speaking this truth.
Too often, the buzz around Machine Learning, Predictive Analytics and Artificial Intelligence sets unrealistic expectations. To be sure, the hype around some “no tech skills required” ML tools has caused some mis-perceptions in the marketplace. First and foremost: Quantum computers are in their infancy, and lack the qubits necessary for what we consider intelligence: self-awareness and the ability to understand cause and effect. Further, these machines lack the ability to repair themselves and reproduce, two requisite factors required of sentient beings. They excel at optimizing, but at the end of the day it’s still about number crunching, albeit at a quantum pace.
And behind all this number crunching — all this Machine Learning — are a set of human-created classifiers that have trained these platforms to provide their users with the answers they seek. The problem is, the users already know the answer, and are just creating a set of constructs that the machine may evaluate to determine said answer.
At the end of the day, the bottom line is that this is not really “thinking.”
Thus do we here at Capture Club see a huge swath of opportunity for innovation. Given this, we have a lofty goal for HAIDIE and ZEN: to actually be able to “think.”
How do we intend to do this? Well, of course, some of that is classified information. From a high-level, though, we intend to try and follow the same path afforded to carbon-based life.
When we think of “intelligence” our human-centric mind tells us that only humans “think.” This is easily disproven by anyone who has spent any extended time observing animals in the wild. All creatures “think” to one degree or another. Certainly they are self-aware, and certainly have the ability to repair themselves and reproduce, to a greater or lesser degree. That qualifies them as sentient.
Beyond that, though, thinking involves building a Worldview. In the simplest terms, a Worldview is a container with a set of deeply-associated constructs. When we encounter new information, we “understand” it by fitting it into our Worldview. If it doesn’t fit, it is either discarded (not recommended) or a new construct is created in the Worldview, the data is inserted there. After that, the system looks to build out associations to the new construct, thus tying it to n number of areas in the Worldview. These associations are constantly re-evaluated for relevancy and priority, and are thus used as relay associates when a relevant query is encountered.
In short, our goal for HAIDIE is to be able to build such a Worldview, and alter it as she would see fit.
To be sure, this is a lofty goal, at best, and a Holy Grail of sorts within the industry. We don’t see any reason to think in smaller, more pragmatic terms, in this regard. There is a world of intelligence to be conquered, innovated, realized, and our goal is to contribute substantially to the knowledge base in this regard.
To innovate, to realize, and to ultimately create cognition, this remains our ultimate goal with HAIDIE.