"Imagination first and then knowledge, Watson!", by Howard Rankin
We must study how humans think and succeed, by Michael Hentschel
Our survival is totally dependent on our ability to predict, by Grant Renier
✋Visit IntualityAI Magazine past issues
|
A great start for the month!
|
|
|
"Imagination first and then knowledge, Watson"
|
Holmes and Watson were having dinner: after being extravagantly hired to provide security at the mansion of a very wealthy, well-known figure. This celebrity was concerned that someone was out to murder him within the next two days. He had spent the day outlining his fears to the detectives.
“Well, we collected a lot of good information, today. Is that why you are so quiet, Sherlock?"
|
|
|
|
"You’re considering all the data we have collected?”
“No, Watson, I am not reviewing the data. I am imagining,” replied Holmes.
“Imagining what?”
“Surely you know me by now, John. As the great Albert Einstein said, ‘Imagination is more important than knowledge.’”
“Surely you need both?” replied the doctor.
“Do you know how the Theory of Relativity was born?: Einstein was sitting under a tree and saw beams of light descending from the clouds like a magical escalator. Then he imagined, ‘I wonder what it would be like to ride on a sunbeam?’ Inspired by that idea he used his knowledge of physics to answer the question. Imagination first and then knowledge, Watson.”
Watson then asked his friend what was it that he was imagining.
“I was imagining various scenarios: I was trying to predict the potential perpetrator’s movements and activities. If he tried to access the property from the extensive back yard, what would be our best precautions? For example, if he tried to come in from the front, what is our best defense? These predictions allow us to prioritize our defenses.”
“Well, that’s all well and good but how do we know if those predictions are right?” asked Watson.
“Having predictions allows us to view any information we have differently: Predictions are a function of intuition and data. However, you are right Watson. If over time those predictions change, we have to adapt accordingly. The feedback loop is essential as it allows us to moderate our predictions over time. That’s the key. That’s what makes intelligence – adaptation. Without the constant feedback you're just churning over the same information and don’t know how to weigh it.”
We are making predictions all the time: even if we don’t realize it. You might want to call them expectations. Most of the time those expectations are unconscious, and transpire naturally and habitually. It is only when those habitual expectations are not met that our awareness is aroused and quickly focused on the unexpected. Research shows that we are more quickly consciously alerted to deviations from expectations.
This is how we learn a new skill: As you first start driving a car, you are likely to, for example, turn the wheel too far to the left. Your expectation as you made that movement was that the car would stay on the side of the road rather than mount the sidewalk. However, that feedback is key as it tells you your expectations are wrong. Now you incorporate that feedback into your next expectations, adapt and can now learn to properly steer the car.
Our expectations are indeed products of experience: feedback about past experiences and associated data -- and imagination.
|
by Howard Rankin PhD, Science Director, Intuality Inc, psychology and cognitive neuroscience
|
|
|
We must study how humans think and succeed
|
To achieve human-desired outcomes via AI: we first study how humans think, predict and succeed.
Human behavior and success are dependent on the interplay: of many complex components of our human “Art of Prediction,” one of humanity’s greatest aspects of active intelligence:
- cognitive human biases (human conscious and subconscious influences, as discussed at length recently)
- quality of information (awareness of incomplete data, true and fake news)
- chosen depth of predictive analysis (the amount of effort taken to calculate an expectation)
|
|
|
|
- probabilistic decision-making (all human predictions are less than 100%, future events are 100%)
- imperfect but satisfactory heuristic shortcuts (humans take shortcuts, for better or for worse)
- instinctive insightful intuition (humans perceive selectively, often based on experience)
We want to know our assets and liabilities in facing risky challenges and opportunities: Complex combinations of some of the above factors comprise much of what we think of as indicators of intelligence. Human decisions are highly dependent on continual predictions and constant re-evaluation. All of these factors are intertwined and constantly influencing each other, leading to a highly dynamic and ever-changing landscape of decisions.
AI is not yet humanized to predict and act similarly to and on behalf of humans: Machines can juggle and mimic past human actions, but cannot yet learn, duplicate, or improve the results of human behavior without knowing future outcomes of specific actions and specific impacts of human thought. Those outcomes are probabilistic and only humans can presently deal with the uncertainty of human behavior.
Humans don't expect 100% success of predictions: They always rethink the present and its consequences. If AI is presently no more than a derivative of the past, then we can only humanize and improve the AI output by teaching the AI our ways: how to predict and weigh the consequences with a risk model the way that humans consciously and subconsciously do.
Computers appear creative because past data and events can be recombined in unique ways: by combining and recombining past data in fully-scaled combinatorial ways, they can present scenarios and rationales that appear innovative, just because those combinations have seldom been seen before. And even more so, we can teach computers not only past derivatives but second-order and compound derivatives that few humans have ever thought of or will ever think to pursue. And we will enable AI to predict.
The key to properly evaluate predicted scenarios is the ability to assess and weigh outcomes: the monetizable or intangible value of the results, and especially how often and reliably a prediction becomes true, requires constant monitoring and reevaluation of observed consequences. This is what humans continually have to do, as imperfect as we are. In a way, this is simply a mathematical or econometrics challenge, which the computer can usually do more thoroughly than humans. But it is the intangible cost/benefit calculations that require more sophisticated analysis. Luckily computers can also be trained for that.
We are already successfully teaching computers to assist and enhance our written and visual Arts: (where AI is already admirably accelerating human work). We will soon teach computers to predict more effectively and accurately in many domains where losses from failures and overall success of human outcomes can be vastly improved.
We must remember Asimov’s science fiction predictions eighty years ago: computers will not be humanized or protect human interest unless they are securely programmed to respect positive human values.
|
by Michael Hentschel, CFO, Intuality Inc, anthropologist, economist, venture capitalist
|
|
|
Our survival is totally dependent on our ability to predict
|
We define the core of intelligence as the Prediction Loop: Without it, we humans and maybe life itself cannot survive.
The loop is a series of sequential functions: that allow us to reconcile, at least temporarily, the current state of things with our need to deal with the future, all the way from forming our next thought to driving to the store. If the loop is somehow missing, where we lose our ability to confront what is to come, we quite literally become locked in place. We are naked in the face of the threatening realities of our environment. Suddenly, all that is real and existential in the world can 'have at us'. We cease to exist.
Our survival is totally dependent on our ability to predict - the Prediction Loop: this DNA-level function is composed of seven sequential functions: 1) New Data real-time, 2) Input 1,000s of data feeds, 3) Prediction direction and magnitude, 4) Action taken on those predictions, 5) Result of action, 6) Adaptation to results, and 7) Biases forming intuition. This is a commonly understood sequence in human decision-making and
|
|
|
|
computer logic. We can all relate to this process at our conscious level, but it's even more efficiently operating in our subconscious without our being aware of it.
The current state of AI lacks a deep understanding of Adaptation and Bias functions: As a result, current AI applications do not include adaptation and bias learning and management algorithms that try to simulate human behavior.
Why is it necessary to simulate human behavior?: In order for AI to be true to its terms - artificial human intelligence - and be general in nature - exhibit intelligence in all kinds of applications - it cannot appear 'naked in the face of the threatening realities'. Another way of saying this is 'it cannot not know' when trying to perform the Prediction Loop. In the absence of New Data or the extreme aberration of New Data, it needs the safety of bias-driven intuition to continue to exist, and an intuition-learning function rather than inputted static biases of the developers. At the end of the day, truly general AI must successfully compete with human behavior, in all of its strange and confusing ways.
Humanized AI includes the vital function of intuitively-formulated decision-making: Replicating this primarily subconscious function and positioning it in its proper role in the Prediction Loop could bring us in confrontation with HAL, of Space Odyssey fame. But unlike that HAL, humanized AI intuition will be no more than the product of the pendrillons of prediction loops it performs, which at that volume will be impossible to us to filter. It will be a complex mixture of the good, bad and ugly.
Humanized AI will be us: It will perform well with the best of us, but no better than our individual expertise and constantly surprised by our uniqueness. Remember that Cline Eastwood won!
|
by Grant Renier, Chairman, Intuality Inc, engineering, mathematics, behavioral science, economics
|
|
|
This content is not for publication
©Intuality Inc 2022-2024 ALL RIGHTS RESERVED
|
|
|
|
