Tiger Spotting Expert
In the last column, as we finally we started to introduce Generative Artificial Intelligence, it seems to some in a rather banal way. So, is it advisable to use this simple, quite a down-to-earth analogy when trying to look at all the complexity of that rather new, and unexplored phenomenon? Isn’t that humiliating for the readers too?
I do not think so!
The model/representation of Generative AI as brain VS mouth/ears is simple, first of all to interest us, and draw us into a better understanding of core principles of Generative AI phenomenon. Is this model appropriate, I am not sure, because I will quote one of my favorite scientists, great English statistician, saying ‘All models are wrong, but some are useful’.
Based on his, one could argue, rather weird understanding of value of modeling, (since ALL of them are wrong!), I will go little deeper, putting one matching wired skill on a list of many skills we have do develop if we want to successfully apply Generative AI. From last column I already suggested a few — ability to comprehend/understand, plus excellent communication skills, both in listening and writing/communicating to the ‘brain’.
And on the very top goes your tiger spotting abilities, which as most skills, you can constantly practice:
George E.P. Box’s quote, “Since all models are wrong, the scientist must be alert to what is importantly wrong. It is inappropriate to be concerned about mice when there are tigers abroad,” reflects the pragmatism of his scientific opus. So precious and rare to find nowadays.
He emphasizes that as models are inherently simplifications of reality and thus always contain some degree of error.
However, the utility of a model lies not in its absolute correctness but in its ability to provide useful insights about the phenomena it aims to describe.
Thus, if our ‘poor’ brain-mouth/ears Generative AI model helps us to understand its internal process of communication and emphasize the importance of superb communication skills over some other valuable, but less important characteristics, these are the skills we need to develop, and or model is jut fine. It serves its unique purpose. To better understand Generative AI.
On the other hand, this model is really lame, and has no room for improvement, enlargement, or drill down, which is also just perfect.
Because, even world smartest physicists are still working on Universal string theory, a theory of everything, a unifying concept that explains all the fundamental forces of nature and ties together quantum mechanics and Einstein’s theory of relativity.
And our simplest and lamest model just works, serves its purpose 😊.
I am of course exaggerating, which I often do, but I will use much simpler, well known example. The Model of our Solar System, remember Copernicus from column one?
We are all fully aware of it, it was shown to us in elementary school, and if is one of well known, one could argue one of the most popular and best models of our civilization. But, just recently I was talking to a friend of mine, maritime Captain, and he pointed out that this model was not applicable to them as students of Naval Academy. They were introduced to similar, enhanced, 3D model, where planets a rotating around the sun (for us, laics, in beautiful, but completely messy order).
Still not perfect, but much closer to reality, according to the specific needs of the trade. This model is fascinating, so beautiful, please check them somewhere on Internet you would be amazed.
In the same time, there is absolutely no need to change the old, well-known educational model, that fully serves purpose of teaching us basic principles of our Solar System.
I don’t know if you are aware that, during our formal and informal education, we are very rarely learning from the real world, we often use models — it’s more intuitive, perfectly serving to point — to understand some natural phenomena.
And of course, we have different models of the same phenomena according to the ones purpose.
Worrying Selectively
Another beautiful pragmatic advice from George E.P. Box, is to clearly point out what are most important concepts, dimensions and/or attribute and create a model, checking out whether it is valid, and whether this model serves your specific purpose — your goal, as we pointed out its importance few columns ago.
So, yes, we always have to worry, check, improve, … but never be afraid to construct our own model, how silly it can be (the brain-mouth/ears Generative AI model), and then check again all tigers, especially if we miss some big ones.
His philosophy on models has influenced fields far beyond statistics, reinforcing the idea that models are tools, not truths.
Understanding their limitations and focusing on significant errors allows for better decision-making in complex systems.
Decision Making
Apart from understanding the world and its phenomena — chemistry, physics, geography, history, … there is practically no school discipline that is not ‘flooded’ with models, for the same reason - to better understand and also easier remember, so you can learn and reproduce the elementary school knowledge. Which we still do.
And to drill down, build new/better ones if needed later. Vast majority of scientific papers are building, discussing some form of models, so everyone else, can hop on, and try to find ‘missing’ tiger.
But, if we have to find a business field, complex one like finance, there is one person standing out, making fortune (mostly) on models. His name is Charlie Munger, esteemed investor and late vice chairman of Berkshire Hathaway, one of most profitable investors company in a world.
Did he made better model for assessing the financial prospect of some company, did he secretly construct a financial decision-making model of some lucrative, specific industry?
No, he shared his secret of successful investor publicly, very often, and is so simple it can be put in one sentence.
He advocates using a latticework of mental models from various disciplines, including economics, psychology, and physics, to better understand complex situations and make more informed decisions.
This multidisciplinary approach allows one to cross-check and validate decisions from multiple angles, reducing the risk of significant errors.
Basically he was applying various, very diverse model on specific company or situation, fearlessly looking for tigers not in picture.
As one cannot completely understand such complex field of investment, one can still run a serious of model checks in order just to effectively lower the uncertainty.
Simple as that, in a very best spirit of G.E.P. Box!
He was so skillful and versatile, that his long-time friend and business companion Warren Buffett described him as best 30 seconds mind in a world.
‘He goes from A to Z in one go. He sees the essence of everything even before you finish the sentence’.
By applying as many various models appropriate, looking for unrevealed tigers, lowering the uncertainty.
In a nutshell, he was cautious, risk-averse thinker in a heavy risk environment. Munger famously focuses on ‘avoiding stupidity’ rather than seeking brilliance, preferring to pass on opportunities that are not clearly within his circle of competence.
Knowing More
· All Models are Wrong. In his paper ‘Science and Statistics’, from 1976, George Edward Pelham Box, a renowned British statistician, emphasized that models are inherently simplifications of reality and thus always contain some degree of error. However, the utility of a model lies not in its absolute correctness but in its ability to provide useful insights about the phenomena it aims to describe. Box’s work underscores the importance of identifying and focusing on critical errors within models that can significantly affect outcomes, rather than getting lost in minor inaccuracies.
Box’s approach to modeling was both philosophical and practical, grounded in the belief that while no model can fully encapsulate reality, a well-constructed model can still be “good enough” for decision-making and prediction. His contributions, particularly in the fields of time series analysis, experimental design, and quality control, laid the foundation for modern statistical practice. By advocating for a balance between mathematical rigor and practical application, Box advanced the art of modeling, encouraging scientists to continually refine their models in the face of new data while staying focused on the most impactful errors.
· Modeling for Decision Making. Charlie Munger, the esteemed investor and vice chairman of Berkshire Hathaway, emphasizes the power of mental models in decision-making. A mental model is a simplified representation of how something works in the real world.
Munger’s practical application of models fully aligns with George E.P. Box’s philosophy: while no model is perfect, combining several imperfect models provides a more reliable basis for decision-making. Munger uses models not to predict outcomes with absolute certainty but to navigate uncertainty effectively. He often emphasizes the importance of understanding where models break down, mirroring Box’s advice to focus on what is “importantly wrong.” For Munger, good decision-making is less about avoiding mistakes entirely and more about avoiding catastrophic mistakes — those ‘tigers’ that can have devastating consequences.
Charlie Munger’s approach is deeply rooted in the idea of ‘worldly wisdom,’ which integrates insights from diverse fields. He uses models to challenge assumptions and test ideas, always keeping an eye on potential risks. This holistic view enables more robust and resilient decision-making, especially in the unpredictable world of investing.
In Search for Knowledge publication
Mastering Insightful Dialogue with Gen AI
<PREV Understanding Generative AI
NEXT> AI — Simplified to The Core
