Many amateur data scientists have little respect to math/stat behind all computational models
Math/stat contains the modelers' logic and intuition to real world data
Good data scientists are ones with excellent intuition

On SIAI's website, we can see most wannabe students go to MSc AI/Data Science program intro page and almost never visit MBA AI program pages. We have a shorter track for MSc that requires extensive pre-study, and much longer version that covers missing pre-studies. Over 90% of wannabes just take a quick scan on the shorter version and walk away. Less than 10% to the longer version, and almost nobody to the AI MBA.

We get that they are 'wannabe' data scientists with passion, motivation, and dream with self-confidence that they are the top 1%. But the reality is harsh. So far, less than 5% applicants have been able to pass the admission exam to MSc AI/Data Science's longer version. Almost never we have applicants who are ready to do the shorter one. Most, in fact, almost all students should compromise their dream and accept the reality. The fact that the admision exam is the first two courses of the AI MBA, lowest tier program, already bring students to senses that over a half of applicants usually disappear before and after the exam. Some students choose to retake the exam in the following year, but mostly end up with the same score. Then, they either criticize the school in very creative ways or walk away with frustrated faces. I am sorry for keeping such high integrity of the school.

Data Scientific Intuition that matters the most

The school focuses on two things in its education. First, we want students to understand the thought processes of data science modelers. Support Vector Machine (SVM), for example, reflects the idea that fitting can be more generalized if a separating hyperplane is bounded with inequalities, instead of fixed conditions. If one can understand that the hyperplane itself is already a generalization, it can be much easier to see through why SVM was introduced as an alternative to linear form fitting and what are the applicable cases in real life data science exercises. The very nature of this process is embedded in the school's motto, 'Rerum Cognoscere Causas' ((Felix, qui potuit rerum cognoscere causas - Wikipedia)), meaning a person pursuing the fundamental causes.

The second focus of the school is to help students where and how to apply data science tools to solve real life puzzles. We call this process as the building data scientific instuition. Often, math equations in the textbooks and code lines in one's program console screens do not have any meaning, unless it is combined in a way to solve a particular problem in a peculiar context with a specific object. Unlike many amateur data scientists' belief, coding libraries have not democratized data science to untrained students. In fact, the codes copied by the amateurs are evident examples of rookie failures that data science tools need must deeper background knowledge in statistics than simple code libraries.

Our admission exam is designed to weed out the dreamers or amateurs. After years of trials and errors, we have decided to give a full lecture of elementary math/stat course to all applicants so that we can not only offer them a fair chance but also give them a warning as realistic as our coursework. Previous schooling from other schools may help them, but the exam help us to see if one has potential to develop 'Rerum Cognoscere Causas' and data scientific intuition.

Intution does not come from hard study alone

When I first raised my voice for the importance of data scientific intution, I had had severe conflicts with amateur engineers. They thought copying one's code lines from a class (or a github page) and applying it to other places will make them as good as high paid data scientists. They thought these are nothing more than programming for websites, apps, and/or any other basic programming exercises. These amateurs never understand why you need to do 2nd-stage-least-square (2SLS) regression to remove measurement error effects for a particular data set in a specific time range, just as an example. They just load data from SQL server, add it to code library, and change input variables, time ranges, and computer resources, hoping that one combination out of many can help them to find what their bosses want (or what they can claim they did something cool). Without understanding the nature of data process, which we call 'data generating process' (DGP), their trials and errors are nothing more than higher correlation hunting like untrained sociologists do in their junk researches.

Instead of blaming one code library worse performing than other ones, true data scientists look for embedded DGP and try to build a model following intuitive logic. Every step of the model requires concreate arguments reflecting how the data was constructed and sometimes require data cleaning by variable re-structuring, carving out endogeneity with 2SLS, and/or countless model revisions.

It has been witnessed by years of education that we can help students to memorize all the necessary steps for each textbook case, but not that many students were able to extend the understanding to ones own research. In fact, the potential is well visible in the admission exam or in the early stage of the coursework. Promising students always ask why and what if. Why SVM's functional shape has $1/C$ which may limit the range of $C$ in his/her model, and what if his/her data sets with zero truncation ends up with close to 0 separating hyperplane? Once the student can see how to match equations with real cases, they can upgrade imaginative thought processes to model building logic. For other students, I am sorry but I cannot recall successful students without that ability. High grades in simple memory tests can convince us that they study hard, but lack of intuition make them no better than a textbook. With the experience, we design all our exams to measure how intuitive students are.

Intuition that frees a data scientist

In my Machine Learning class for tree models, I always emphasize that a variable with multiple disconnected effective ranges in trees has a different spanned space from linear/non-linear regressions. One variable that is important in a tree space, for example, may not display strong tendency in linear vector spaces. A drug that is only effective to certain age/gender groups (say 5~15, 60~ for male, 20~45 female) can be a good example. Linear regression hardly will capture the same efffective range. After the class, most students understand that relying on Variable Importances of tree models may conflict with p-value type variable selections in regression-based models. But only students with intuition find a way to combine both models that they find the effective range of variables from the tree and redesign the regression model with 0/1 signal variables to separate the effective range.

The extend of these types of thought process is hardly visible from ordinary and disqualified students. Ordinary ones may have capacity to discern what is good, but they often have hard time to apply new findings to one's own. Disqualified students do not even see why that was a neat trick to the better exploitation of DGP.

What's surprising is that previous math/stat education mattered the least. It was more about how logical they are, how hard-working they are, and how intuitive they are. Many students come with the first two, but hardly the third. We help them to build the third muscle, while strenghtening the first. (No one but you can help the second.)

The re-trying students ending up with the same grades in the admission exam are largely because they fail to embody the intuition. It may take years to develop the third muscle. Some students are smart enough to see the value of intuition almost right away. Others may never find that. For failing students, as much as we feel sorry for them, we think that their undergraduate education did not help them to build the muscle, and they were unable to build it by themselves.

The less chanllenging tier programs are designed in a way to help the unlucky ones, if they want to make up the missing pieces from their undergraduate coursework. Blue pills only make you live in fake reality. We just hope our red pill to help you find the bitter but rewarding reality.

AI talks turned the table and become more pessimistic
It is just another correction of exorbitant optimism and realisation of AI's current capabilities
AI can only help us to replace jobs in low noise data
Jobs needing to find new patterns and from high noise data industry, mostly paid more, will not be replaceable by current AI

There have been pessimistic talks about the future of AI recently that have created sudden drops in BigTech firms' stock prices. In all of a sudden, all pessimistic talks from Investors, experts, and academics in reputed institutions are re-visited and re-evaluated. They claim that ROI (Return on Investment) for AI is too low, AI products are too over-priced, and economic impact by AI is minimal. In fact, many of us have raised our voices for years with the exactly same warnings. 'AI is not a magic wand'. 'It is just correlation but not causality / intelligence'. 'Don't be overly enthusiastic about what a simple automation algorithm can do'.

As an institution with AI in our name, we often receive emails from a bunch of 'dreamers' that they wonder if we can make a predictive algorithm that can foretell stock price movements with 99.99% accuracy. If we could do that, why do you think we would share the algorithm with you? We should probably keep it for secret and make billions of dollars just for ourselves. As much as the famous expression by Milton Friedman, a Nobel economist, there is no such thing as a free lunch. If we have a perfect predictability and it is widely public, then the prediction is no longer a prediction. If everyone knows the stock A's price goes up, then everyone would buy the stock A, until it reaches to the predicted value. Knowing that, the price will jump to the predicted value, almost instantly. In other words, the future becomes today, and no one gets benefited.

AI = God? AI = A machine for pattern matching

A lot of enthusiasts have exorbitant optimism that AI can overwhelm human cognitivie capacity and soon become god-like feature. Well, the current forms of AI, be it Machine Learning, Deep Learning, and Generative AI, are no more than a machine for pattern matching. You touch a hot pot, you get a burn. It is painful experience, but you learn that you should not touch when it is hot. The worse the pain, the more careful you become. Hopefully it does not make your skin irrecoverable. The exact same pattern works for what they call AI. If you apply the learning processes dynamically, that's where Generative AI comes. The system is constantly adding more patterns into the database.

Though the extensive size of patterns does have great potential, it does not mean that the machine has cognitive capacity to understand the pattern's causality and/or to find new breakthrough patterns from list of patterns in the database. As long as it is nothing more than a pattern matching system, it never will.

To give you an example, can it be used what words you are expected to answer in a class that has been repeated for thousand times? Definitely. Then, can you use the same machine to predict the stock price? Aren't the stock market repeating the same behavior over a century? Well, unfortunately it is not, thus you can't be benefited by the same machine for financial investments.

Two types of data - Low noise vs. High noise

On and near the Wall Street, you can sometimes meet an excessively confident hedge fund manager with claims on near perfect foresight for financial market movements. Some of them have outstanding track records, and surprisingly persuasive. In New York Times archive back in 1940s, or even as early as 1910s, you can see people with similar claims were eventually sued by investors, arrested due to false claims, and/or just disappeared from the street within a few years. If they were that good, why then they lost money and got sued/arrested?

There are two types of data. One set of data that you can see from machine (or highly controlled environment) is called 'Low-noise' data. It has high predictability. Even in cases where embedded patterns are invisible by bare eyes, you either need more analytic brain or a machine to test all possibilities within the possible sets. For the game of Go, the brain was Se-dol Lee and the machine was Alpha-Go. The game needs to test 19x19 possible sets with around 300 possible steps. Even if your brain is not as good as Se-dol Lee, as long as your computer can find the winning patterns, you can win. This is what has been witnessed.

The other set of data comes from largely uncontrolled environment. There potentially is a pattern, but it is not the single impetus that drives every motion of the space. There are thousands, if not millions, of patterns that the driver is not observable. This is where randomness is needed for modeling, and it is unfortunately impossible to predict accurate move, because the driver is not observable. We call this set of data 'High-noise'. The stock market is the very example of such. There are millions of unknown, unexpectable, and at least unmeasurable influences that disable any analyst or machine to predict with accuracy level upto 100%. This is why financial models are not researched for predictability but used only to backtest financial derivatives for reasonable pricing.

Natural language process (NLP) is one example of low noise. Our language follows a certain set of rules (or patterns), which are called grammar. Unless you are uneducated or intentionally out of grammar (or make mistakes), people generally follow grammar. Weather is mostly low noise, but it has high noise components. Sometimes typhoons are unpredictable, or less predictable. Stock market? Be my guest. There have been 4 Nobel Prizes given to financial economists by year 2023, and all of them are based on the belief that stock markets follow random processes, be it Gaussian, Poisson, and/or any other unknown random distributions. (Just in case, if a process follows any known distribution, that means it is probabilistic, which means it is random.)

Potential benefits of AI

We as an institution hardly believe current forms of AI will make any significant changes in businesses and our life in short term. The best we can expect is automation of mundane tasks. Like laundary machine in early 20th century. ChatGPT already has shown us a path. Soon, CS operators will largely be replaced by LLM based chatbots. US companies actively outsourced the function from India for the past a few decades, thanks to cheaper international connectivity via internet. It will still remain, but human actions will be needed way less than before. In fact, we already get machine generated answers from a number of international services. If we complain about a program's malfunction on a WordPress plugin, for instance, established services email us machine answers first. For a few cases, it actually is enough. The practice will become more popular to less-established services as it becomes easier and cheaper to implement.

Teamed up with EduTimes, we also are working on a research to replace 'Copy Boys/Girls'. Journalists that we know from large news magazines are not always running on the street to find new and fascinating stories. In fact, most of them read other newspapers and rewrite the contents as if they were the original sources. Although it is not an important job, it is still needed for the newspaper to run. They need to keep up the current events, accoring to the EduTimes journalists from other renouned newspapers. The copy team is usually paid the least and seen a death sentence as a journalist. What makes the job more sympathetic on top of the least respect, it will soon be replaced by LLM based copywriters.

In fact, any job that generates patterned contents without much of cognitivie functions will gradually be replaced.

What about automotive driving? Is it a low-noise pattern job or a high-noise complicated cognitive job? Well, although Elon Musk claims high possibility of Lv. 4 auto-driving within next a few years, we don't believe so. None of us at GIAI have seen any game theorists have solved multi-agent ($n$>2) Bayesian belief game with imperfect information and unknown agent types by computer so that the automotive driving algorithm can predict what other drivers on the road will do. Without the right prediction of others on the fast moving vehicles, it is hard to tell if your AI will help you successfully avoid other crazy drivers. The driving job for those eventful cases needs 'instinct', which requires another set of bodily function different from cognitive intelligence. The best that the current algorithm can do is to tighten it up to perfection for a single car, which already needs to go over a lot of mathematical, mechanical, organisational, legal, and commercial (and many more) challenges.

Don't they know all that? Aren't the Wall Street investors self-confident, egocentric, but ultra smart that they already know all the limitations of AI? We believe so. At least we hope so. Then, why do they pay attention to the discontentful pessimism now, and create heavy drops in tech stock prices?

Guess the Wall Street hates to see Silicon Valley to be paid too much. American East often think the West too unrealistic and floating in the air. OpenAI's next round funding may surprise us in a totally opposite direction.

Boot camp is for software programming without mathematical training
MSc is a track for PhD, with in-depth scientific research written in the language of math and stat
We respect programmers, but our works are significantly varying

Due to the fact that we are running SIAI, an higher educational institution for AI/Data Science, we often have questions about the difference between Boot Camps for AI and MSc programmes. The shortest answer is the difference in Math requirements. Masters track is for people looking for academic training so that one can read academic papers in that subject. With PhD in the topic, we expect the student to be able to lead a research. From Boot Camp, sorry to be a little aggressive here, but we only expect a 'Coding Monkey'.

We are aware that many countries are shallow in AI/Data Science that they want employees only to be able to best use of Open AI's and AWS's libraries by Rest API. For that, boot camp should be enough, unless the boot camp teacher does not know how to do so. There are nearly infinite amount of contents for how to use Rest API for your software, regardless of your backend platform, be it an easy script languages like Python or tough functional ones like OCaml. Difficulties are not always indicators of determinants in challenges, and we, as data scientists at GIAI, care less about what language you use. What's important is how flexible your thinking for mathematically contained modeling.

Boot camp for software programing, MSc for scientific training

Unfortunately, unless you are lucky enough to be born as smart as Mr. Ramanujan, you cannot learn math modeling skills from a bunch of blogs. Programming, however, has infinitely many proven records of excellent programmers without school traninng. Elon Musk is just one example. He did Economics and Physics in his undergrad at U Penn, and he only stayed one day in the mechanical engineering PhD program at Stanford University. Programming is nothing more than a logic, but math needs too many building blocks to understand the language.

When we first build SIAI, we had quite a lengthy discussion for weeks. Keith was firm that we should stick to mathematical aspects of AI/Data Science. (which doesn't mean we should only teach math, just to avoid any misunderstanding.) Mc wanted two tier tracks for math and coding. We later found that with coding, it is unlikely that we can have the school accreditted by official parties, so we end up with Keith's idea. Besides, we have seen too many Boot Camps around the world that we do not believe we can be competitive in that regard.

The founding motto of the school is 'Rerum Cognoscere Causaus', meaning 'the real cause of things'. With mathematical tools, we were sure that we can teach what are the reason behind a computational model was first introduced. Indeed, Keith has done so well in his Scientific Programming that most students no longer bound to media brainwashing that Neural Network is the most superior model.

Scientists do our own stuff

If you just go through Boot Camps for coding, chances are that you can learn the limitations of Neural Network just by endless trials and errors, if not somebody's Medium posts and Reddit comments. In other words, without the proper math training, it is unlikely one can understand how the computational logics of each model are built, which makes us to aloof from all programmers without necessary math training.

The very idea comes from multiple rounds of uneasy exposures to software engineers without a shred of understanding in modeling side of AI. They usually claim that Neural Network is proven to be the best model, and they do not need any other model knowledge. And all they have to do is to run and test it. Researchers at GIAI are trained scientists, and we mostly can guess what will happen just by looking at equations. And, most importantly, we are well aware that NN is the best model only for certain tasks.

They kept claim that they were like us, and some of them wanted to build a formal assocation with SIAI (and later GIAI). It's hard for us to work with them, if they keep that attitude. These days, whenever we are approached by third parties, if they want to be at equals with us, we ask them to show us math training levels. Please make no mistake that we respect them as software engineers, but we do not respect them as scientists.

Guess aforementioned story and internal discomfort tells you the difference between software engineers and data/research scientists, let alone tools that we rely on.

We screen out students by admission exams in math/stat

With the experience, Keith initiated two admission exams for our MSc AI/Data Science programmes. At the very beginning, we thought there will be plenty of qualifying students, so we used final year undergrad materials. There was a disaster. We gave them two months of dedicated training. Provided similar exams and solved each one of them with extra detail. But, only 2 out of 30 students were able to get grades good enough to be admitted.

We lowered the level down to European 2nd year (perhaps American 3rd year), and the outcome wasn't that different. Students were barely able to grasp superficial concepts of key math/stat. This is why we were kinda forced to create an MBA program that covers European 2nd year teaching materials with ample amount of business application cases. With that, students survive, but answer keys in their final exam tell us that many of them belong to coding Boot Camps, not SIAI.

From year 2025 and onwards, we will have one admission exam for MSc AI/Data Science (2 year) in March, after 2 months pre-training in Jan and Feb. The exam materials will be 2nd year undergrad level. If a student passes, we offer an exam with one notch up in June, again after 2 months pre-training in Apr and May. This will give them MSc AI/Data Science (1 year) admission.

Students who failed the 2-year track admission, we offer them MBA AI program admission, which covers some part of the 2-year track courses. If they think they are ready, then in the following year, they can take the admission exam again. After a year of various courework, some students have shown better performance, based on our statistics, but not by much. It seemed like the brain has its limit that they cannot go above.

Precisely by the same reason, we are reasonably sure that not that many applicants will be able to come to 2-year track, and almost no one for the 1-year track. More details are available from below link:

• Program comparison - Swiss Institute of Artificial Intelligence (siai.org)

Top brains in AI/Data Science are driven to challenging jobs like modeling
Seldom a 2nd-tier company, with countless malpractices, can meet the expectations
Even with $$$, still they soon are forced out of AI game

A few years ago, a large Asian conglomerate acquired a Silicon Valley's start-up just off an early Series A funding. Let's say it is start-up $\alpha$. The M&A team leader later told me that the acquisition was mostly to hire the data scientist in the early stage start-up, but the guy left $\alpha$ on the day the M&A deal was announced.

I had an occation to sit down with the data scientist a few months later, and asked him why. He tried to avoide the conversation, but it was clear that the changing circumstances definitely were not within his expectation. Unlike other bunch of junior data scientists in Silicon Valley's large firms, he did signal me his grad school training in math and stat that I had a pleasant half an hour talk about models. He was mal-treated in large firms that he was given to run SQL queries and build Tableau-based graphes, like other juniors. His PhD training was useless in large firms, so he had decided to be a founding member of $\alpha$ that he can build models and test them with live data. The Asian acquirer with bureaucratic HR system wanted him to give up his agenda and to transplant the Silicon Valley large firm's junior data scientist training system to the acquirer firm.

Brains go for brains

Given tons of other available positions, he didn't waste his time. Personall,y I also have lost some months of my life for mere SQL queries and fancy graphes. Well, some people may still go for 'data scientist' title, but I am my own man. So was the data scientist from $\alpha$.

These days, Silicon Valley firms call the modelers as 'research scientists', or simliar names. There also are positions called 'machine learning engineers' whose jobs somewhat related to 'research scientists', but may disinclude mathematical modeling parts and way more software engineering in it. The title 'Data Scientists' are now given to jobs that were used to be called 'SQL monkeys'. As the old nickname suggests, not that many trained scientists would love to do the job, even with competitive salary package.

What companies have to understand is that we, research scientists, are not trained for SQL and Tableau, but mathematical modeling. It's like a hard-trained sushi cook(将太の寿司, shota no sushi) is given to make street food like Chinese noodle.

Let me give you an example in real corporate world. Let's say a semi-conductor company, $\beta$ wants to build a test model for a wafer / subsctrate. What I often hear from those companeis are that they build a CNN model that reads the wafer's image and match it with pre-labeled 0/1 for error detection. In fact, similar practices have been widely adapted practice among all Neural Network maniacs. I am not saying it does not work. It works. But then, what would you do, if the pre-label was done poorly? Say, the 0/1 entries were like over 10,000 and hardly any body double checked the accruracy. Can you rely on that CNN-based model? In addition to that, the model probably require enourmous amount of computational costs to build, let alone test and operating it daily.

Wrong practice that drives out brain

Instead of the costly and less scientific option, we can always build a model that captures data's generated process(DGP). The wafer is composed of $n \times k$ entries, and issues emerge when $n \times 1$ or $1 \times k$ entries go wrong altogether. Given the domain knowledge, one can build a model with cross-products between entries in the same row/column. If it is continuously 1 (assume 1 for error), then it can easily be identified as a defect case.

Cost of building a model like that? It just needs your brain. There is a good chance that you don't even need a dedicated graphics card for that calculation. Maintenance costs are also incomparably smaller than the CNN version. The concept of computational cst is something that you were supposed to learn in any scientific programming classes at school.

For companies sticking to the expensive CNN options, I always can spot followings:

• The management has little to no sense of 'computational cost'
• The manaement cannnot discern 'research scientists' and 'machine learning engineers'
• The company is full of engineers without the sense of mathematical modeling

If you want to grow up as a 'research scientist', just like the guy at $\alpha$, then run. If you are smart enough, you must have already run, like the guy at $\alpha$. After all, this is why many 2nd-tier firms end up with CNN maniacs like $\beta$. Most 2nd-tier firms are unlucky that they cannot keep research scientists due to lack of knowledge and experience. Those companies have to spend years of time and millions of wasted dollars to find that they were so long. By the time that they come to senses, it is mostly already way too late. If you are good enough, don't waste your time on a sinking ship. The management needs so-called cold-turkey type shock treatment as a solution. In fact, there was a start-up that I stayed only for a week, which lost at least one data scientist in everyweek. The company went to bankrupt in 2 years.

What to do and not to do

At SIAI, I place Scientific Programming right after elementary math/stat training. Students see that each calculation method is an invention to overcome earlier available options' limitations but simultanesouly the modification bounds the new tactic in another directions. Neural Networks are just one of the many kinds. Even with the eye-opening experience, some students still remain NN maniacs, and they flunk in Machine Learning and Deep Learning classes. Those students believe that there must exist a grand model that is univerally superior to all other models. I wish the world is that simple, but my ML and DL courses break the very belief. Those who are awaken, usually become excellent data/research scientists. Many of them come back to me that they were able to minimize computational costs by 90% just by replacing blindly implemented Neural Network models.

Once they see that dramatic cost reduction, at least some people understand that the earlier practice was wrong. The smarty student may not be happy to suffer from poor management and NN maniacs for long. Just like the guy at $\alpha$, it is always easier to change your job than fighting to change your incapable management. Managers moving fast maybe able to withhold the smarty. If not, you are just like the $\beta$. You invest a big chunk of money for an M&A just to hire a smarty, but the smarty disappears.

So, if you want to keep the smarty? Your solution is dead simple. Test math/stat training levels in scientific programming. You will save tons of $$$ in graphic card purchase.

ChatGPT is to replace not jobs but tedious tasks
For newspapers, 'rewrite man' will soon be gone
For other jobs, the 'boring' parts will be replaced by AI, 
but not the intellectual and challenging parts

There has been over a year of hype for Large Language Models(LLMs). At the onset and initial round of hype, people outside of this field asked me if their jobs were to be replaced by robots. By now, over a year of trials with ChatGPT, they finally seem to understand that it is nothing more than an advanced chatbot that still is unable to stop generating 'bullshit', according to Noam Chomsky, an American professor and public intellectual known for his work in linguistics and social criticism.

As my team at GIAI predicted in early 2023, all LLM trials will be able to replace some jobs, but most jobs that will be replaced will be simple mundane tasks. That's because these language models are meant to find higher correlation between text/image groups, but still unable to 'intelligently' find logical connection between thoughts. In statistics, it is called high correlation with no causality, or simply 'spurious relations'.

LLMs will replace 'copying boys/girls'

When we were first approached by EduTimes back in early 2022, they thought we could create an AI machine to replace writers and reporters. We told them the best we can create is to replace a few boring desk jobs like 'rewrite man'. The job that requires to rewrite what other newspapers have already reported. 'Copy boy' is one well-known disparaging term for that job. Most large national magazines have such employees, just to keep their magazines to be up-to-dated with recent news.

Since none of us at GIAI are from journalism, and EduTimes is far from a large national magazine, we are not aware of exact proportion of 'rewrite man' in large magazines, let alone how many articles are re-written by them. But based on what we see from magazines, we can safely argue that at least 60~80% articles are probably written by the 'copy boys/girls'. Some of them are at the high risk of plagiarism. This is one sad reality of journalism industry, accoring to the EduTimes team.

The LLM that we are working on, GLM(GIAI's Language Model), isn't that different from other competitors in the market that we also have to rely on text bodies' correlations, or more precisely 'associations' by the association rules in machine learning textbooks. Likewise, we also have lots of inconsistency problems. To avoid the Noam Chomsky's famous accusation, 'LLMs are bullshit generators', the best any data scientist can do is just to set a high cut-off in support, confidence, and lift. Beyond that, it is not the job of data models, which includes all AI variants for pattern recognition.

But still correlation does not necessarily mean causality

The reason we see infinitely many 'bullshit' cases is because the LLM services still belong to statistics, a discipline to find not causality but correlation.

If high correlation can be translated to high causality, there has to be one important condition satisfied. The data set contains all coherent information so that high correlation naturally means high causality. This actually is where we need the EduTimes. We need clean, high quality, and topic-specific data.

After all, this is why OpenAI is willing to pay for data from Reddit.com, a community with intense and quality discussions. LLM service providers are in negotiation with U.S. top newspapers precisely the same reason. Although it does not mean that coherent and quality news articles will give us 100% guarantee in correlation to causality, at least we can establish a claim that disturbing cases will largely be gone without time-consuming technical optimization.

By the same logic, jobs that can be replaced by LLMs or any other AIs with pattern matching algorithms are the ones that have strong and repeating patterns that does not require logical connections.

AI can replace not (intelligent) jobs but (boring) tasks

As we often joke around at GIAI, technologies are bounded by mathematical limitations. Unfortunately, we are not John von Neumann who can solve every impossible mathematical challenges as easy as college problem sets. Thanks to computational breakthroughs, we are already at the level far from what we expected 10 years ago. Back then, we did not expect to extract corpora from 10 books in a few minites. If anything, we thought it needed weeks of supercomputer resources. It is not anymore. But even with surprising speed of computational achievements, we are still bound to mathematical limits. As said, correlation without causality is 'bullshit'.

With the current mathematical limitations, we can say

• AI can replace not (intelligent) jobs but (super mega ultra boring) tasks

And, the replaceable tasks are boring, tedious, repetitive, and patterned tasks. So, please stop worrying about losing jobs, if yours torture your brain to think. Instead, plz think about how to use LLMs like automation to lighten your burden from mundane tasks. It will be like your mom's laundary machine and dish washer. Younger generation females no longer are bound to housekeeping. They go out to work places and fight for the positions that meet their dreams, desires, and wants.

Data Science can find correlation but not causality
In stat, no causal but high correlation is called 'Spurious regression'
Hallucinations in LLMs are repsentative examples of spurious correlation

Imagine two twin kids living in the neighborhood. One prefers to play outside day and night, while the other mostly sticks to his video games. After a year later, doctors find that the gamer boy is much healthier, thus conclude that playing outside is bad for growing children's health.

What do you think of the conclusion? Do you agree with the conclusion?

Even without much scientific training, we can almost immediately dismiss the conclusion that is based on lop-sided logic and possibly driven by insufficient information of the neighborhood. For example, if the neighborhood is as radioactively contained as Chernobyl or Fukushima, playing outside can undoubtedly be as close as committing a suicide. What about more nutrition provided to the gamer boy due to easier access to home food? The gamer body just had to drop the game console for 5 seconds to eat something, but his twin had to walk or run for 5 minites to come back home for food.

In fact, there are infinitely many potential variables that may have affected two twin kids' condition. Just by the collected data set above, the best we can tell is that for an unknown reason, the gamer boy is medically healthier than the other twin.

In more scientific terms, it can be said that statistics has been known for correlations but not for causality. Even in a controlled environment, it is hard to argue that the control variable was the cause of the effect. Researchers only 'guess' that the correlation means causality.

Post Hoc, Ergo Propter Hoc

There is a famous Latin phrase meaning "after this, therefore on account of it". In plain English, it means that one event is the cause of the other event occuring right next. You do not need rocket science to counterargue that two random events are interconnected just because one occured right after another. This is a widely common logical mistake that assigns causality just by an order of events.

In statistics, it is often called that 'Correlation does not necessarily guarantee causality'. In the same context, such a regression is called 'Spurious regression', which has been widely reported in engineers' adaptation of data science.

One noticeable example is 'Hallucination' cases in ChatGPT. The LLM only finds higher correlation between two words, two sentences, and two body of texts (or images in these days), but it fails to discern the causal relation embedded in the two data sets.

Statistians have long been working on to differentiate the causal cases from high correlation, but the best so far we have is 'Granger causallity', which only helps us to find no causality case between 3 variables. Granger causality offers a philophical frame that can help us to test if the 3rd variable can be a potential cause of the hidden causality. The academic countribution by Professor Granger's research to be Nobel Prize awarded is because it proved that it is mechanically (or philosophically) impossible to verify a causal relationship just by correlation.

Why AI ultimately needs human approval?

The Post Hoc Fallacy, by nature of current AI models, is an unavoidable huddle that all data scientists have to suffer from. Unlike simple regression based researches, the LLMs rely on too large chunk of data that it is practically impossible to tackle every connection of two text bodies.

This is where human approval is required, unless the data scientists decide to finetune the LLM in a way to offer only the highest probable (thus causal) matches. The more likely the matches are, the less likely there will be spurious connection between two sets of information, assuming that the underlying data is sourced from accurate providers.

Teaching AI/Data science, I surprisingly often come across a number of 'fake experts' whose only understanding of AI is a bunch of terminology from newspapers, or a few lines of media articles at best, without any in-depth training in basic academic tools, math and stat. When I raise Grange causality as my counterargument for impossibility to distinguish from correlation to causality by statistical methods alone (by far philosophically impossible), many of them ask, "Then, wouldn't it be possible with AI?"

If the 'fake experts' had some elementary level math and stat training from undergrad, I believe they should be able to understand that computational science (academic name of AI) is just a computer version of statistics. AI is actually nothing more than the task of performing statistics more quickly and effectively using computer calculations. In other words, AI is a sub-field of statistics. Their questions can be framed like

• If it is impossible with statistics, wouldn’t it be possible with statistics calculated by computers?
• If it is impossible with elementary arithmetic, wouldn't it be possible with addition and subtraction?

The inability of statistics to make causal inferences is the same as saying that it is impossible to mechanically eliminate hallucinations in ChatGPT. Those with academic training in the fields social sciences, the disciplines of which collect potentially correlated variables and use human experience as the final step to conclude causal relationships, see that ChatGPT is built to mimic cognitive behavior at the shamefully shallow level. The fact that ChatGPT depends on 'Human Feedback' in its custom version of 'Reinforcement Learning' is the very example of the basic cognitive behavior. The reason that we still cannot call it 'AI' is because there is no automatic rule for the cheap copy to remove the Post Hoc Fallacy, just like Clive Granger proved in his work for Nobel Prize.

Causal inference is not monotonically increasing challenge, but multi-dimensional problem

In natural science and engineering, where all conditions are limited and controlled in the lab (or by a machine), I often see cases where they see human correction as unscientific. Is human intervention really unscientific? Well, Heidelberg's indeterminacy principle states that when a human applies a stimulus to observe a microscopic phenomenon, the position and state just before applying the stimulus can be known, but the position after the stimulus can only be guessed. If no stimulation is applied at all, the current location and condition cannot be fully identified. In the end, human intervention is needed to earn at least partial information. Withou it, one can never have any scientifically proven information.

Computational science is not much different. In order to rule out hallucinations, researchers either have to change data sets or re-parameter the model. The new model may be closer to perfection for that particular purpose, but the modification may surface hidden or unknown problems. The vector space spanned by the body of data set is too large and too multidimensional that there is no guarantee that one modification will always monotonically increase the perfection in every angle.

What is more concerning is that the data set is clean, unless you are dealing with low noise (or zero noise) ones like grammatically correct texts and quality images. Once researchers step aside from natural language and image recognition, data sets are exposed to infinitely many sources of unknown noises. Such high noise data often have measurement error problems. Sometimes researchers are unable to collect important variables. These are called 'endongeneity', and social scientists have spent nearly a century to extract at least partial information from the faulty data.

Social scientists have modified statistics in their own way that complements 'endogeneity'. Econometrics is a representative example, using the concept of instrumental variables to eliminate problems such as errors in variable measurement, omission of measured variables, and two-way influence between explanatory variables and dependent variables. These studies are coined 'Average Treatment Effect' and 'Local Average Treatment Effect' that were awarded the Nobel Prize in 2021. It's not completely correct, but it's part of the challenge to find a little less wrong.

Some untrained engineers claim magic with AI

Here at GIAI, many of us share our frustrations with untrained engineers confused AI as a marketing term for limited automatization with real self-evolving 'intelligence'. The silly claiming that one can find causality from correlation is not that different. The fact that they claim such spoofing arguments already proves that they are unaware of Granger's causality or any philosophically robust proposition to connect/disconnect causality and correlation, thus proves that they lack scientific training to handle statistical tools. Given that current version of AI is no better than pattern matching for higher frequency, it is no doubt that scientifically untrained data scientists are not entitled to be called data scientists.

Let me share one bizarre case that I heard from a colleague here at GIAI from his country. In case anyone feel that the following example is a little insulting, a half of his jokes are about his country's inable data scientists. In one of the tech companies in his country, a data scientist was given to differentiate a handful of causal events from a bunch of correlation cases. The guy said "I asked ChatGPT, but it seems there were limitations because my GPT version is 3.5. I should be able to get a better answer if I use 4.0."

The guy not only is unaware of the post hoc fallacy in data science, but he also highly likely does not even understand that ChatGPT is no more than a correlation machine for texts and images by given prompts. This is not something you can learn from job. This is something you should learn from school, which is precisely why many Asian engineers are driven to the misconception that AI is magic. It has been known that Asian engineering programs generally focus less on mathematical backup, unlike renowned western universities.

In fact, it is not his country alone. The crowding out effect is heavy as you go to more engineer driven conferences and less sophisticated countries / companies. Despite the shocking inability, given the market hype for Generative AI, I guess those guys are paid high. Whenever I come across mockeries like the untrained engineers and buffoonish conferences, I just laugh it off and shake it off. But, when it comes to businesses, I cannot help ask myself if they worth the money.

STEM majors are known for high dropouts
Students need to have more information before jumping into STEM
Admission exam and tiered education can work, if designed right

Over the years of study and teaching in the fields of STEM(Science, Technology, Engineering, and Mathematics), it is not uncommon to see students disappearing from the program. They often are found in a different program, or sometimes they just leave the school. There isn't commonly shared number of dropout rate across the countries, universities, and specific STEM disciplines, but it has been witnessed that there is a general tendancy that more difficult course materials drive more students out. Math and Physics usually lose the most students, and graduate schools lose way more students than undergraduate programs.

At the onset of SIAI, though there has been growing concerns that we should set admission bar high, we have come to agree with the idea that we should give chances to students. Unlike other universities with somewhat strict quota assigned to each program, due to size of classrooms, number of professors, and etc., since we provide everything online, we thought we are limitless, or at least we can extend the limit.

After years of teaching, we come to agree on the fact that rarely students are ready to study STEM topics. Most students have been exposed to wrong education in college, or even in high school. We had to brainwash them to find the right track in using math and statistics for scientific studies. Many students are not that determined, neither. They give up in the middle of the study.

With stacked experience, we can now argue that the high dropout rate in STEM fields can be attributed to a variety of factors, and it's not solely due to either a high number of unqualified students or the difficulty of the classes. Here are some key factors that can contribute to the high dropout rate in STEM fields:

1. High Difficulty of Classes: STEM subjects are often challenging and require strong analytical and problem-solving skills. The rigor of STEM coursework can be a significant factor in why some students may struggle or ultimately decide to drop out.
2. Lack of Preparation: Some students may enter STEM programs without sufficient preparation in foundational subjects like math and science. This lack of preparation can make it difficult for students to keep up with the coursework and may lead to dropout.
3. Lack of Support: Students in STEM fields may face a lack of support, such as inadequate mentoring, tutoring, or academic advising. Without the necessary support systems in place, students may feel isolated or overwhelmed, contributing to higher dropout rates.
4. Perceived Lack of Relevance or Interest: Some students may find that the material covered in STEM classes does not align with their interests or career goals. This lack of perceived relevance can lead to disengagement and ultimately dropout.
5. Diversity and Inclusion Issues: STEM fields have historically struggled with diversity and inclusion. Students from underrepresented groups may face additional barriers, such as lack of role models, stereotype threat, or feelings of isolation, which can contribute to higher dropout rates.
6. Workload and Stress: The demanding workload and high levels of stress associated with STEM programs can also be factors that lead students to drop out. Balancing coursework, research, and other commitments can be overwhelming for some students.
7. Career Prospects and Job Satisfaction: Some students may become disillusioned with the career prospects in STEM fields or may find that the actual work does not align with their expectations, leading them to reconsider their career path and potentially drop out.

It's important to note that the reasons for high dropout rates in STEM fields are multifaceted and can vary among individuals and institutions. Addressing these challenges requires a holistic approach that includes providing academic support, fostering a sense of belonging, promoting diversity and inclusion, and helping students explore their interests and career goals within STEM fields.

Not just for the gifted bright kids

Given what we have witnessed so far, at SIAI, we have changed our admission policy quite dramatically. The most important of all changes is that we have admission exams and courses for exams.

Although it sounds a little paradoxical that students come to the program to study for exam, not vice versa, we come to an understanding that our customized exam can greatly help us to find true potentials of each student. The only problem of the admission exam is that the exam mostly knocks off students by the front. We thus offer classes to help students to be prepared.

This is actually a beauty of online education. We are not bounded to location and time. Students can go over the prep materials at their own schedule.

So far, we are content with this option because of following reasons:

1. Self-motivation: The exams are designed in a way that only dedicated students can pass. They have to do, re-do, and re-do the earlier exams multiple times, but if they do not have self-motivation, they skip the study, and they fail. The online education unfortunately cannot give you detailed mental care day by day. Students have to be matured in this regard.
2. Meaure preparation level: Hardly a student from any major, be it a top schools' STEM, we find them not prepared enough to follow mathematical intuitions thrown in classes. We designed the admission exam one-level below their desired study, so if they fail, that means they are not even ready to do the lower level studies.
3. Introduction to challenge: Students indeed are aware of challenges ahead of them, but the depth is often shallow. 1~2 courses below the real challenge so far consistently helped us to convince students that they need loads of work to do, if they want to survice.

Selfdom there are well-prepared students. The gifted ones will likely be awarded with scholarships and other activities in and around the school. But most other students are not, and that is why there is a school. It is just that, given the high dropout in STEM, it is the school's job to give out right information and pick the right student.

People following AI hype are mostly completely misinformed
AI/Data Science is still limited to statistical methods
Hype can only attract ignorance

As a professor of AI/Data Science, I from time to time receive emails from a bunch of hyped followers claiming what they call 'recent AI' can solve things that I have been pessimistic. They usually think 'recent AI' is close to 'Artificial General Intelligence', which means the program learns by itself and it is beyond human intelligence level.

At the early stage of my start-up business, I answered them with quality contents. Soon, I realized that they just want to hear what they want to hear, and criticize people saying what they don't want to hear. Last week, I came across a Scientific American's article about history of automations that were actually people. (Is There a Human Hiding behind That Robot or AI? A Brief History of Automatons That Were Actually People | Scientific American)

AI hype followers' ungrounded dream for AGI

No doubt that many current AI tools are far more advanced than medieval 'machines' that were discussed in the Scientific American article, but human generated AI tools are still limited to pattern finding and abstracting it by featuring common parts. The process requires to implement a logic, be it human found or human's programmed code found, and unfortunately the machine codes that we rely on is still limited to statistical approaches.

AI hype followers claim that recent AI tools have already overcome needs for human intervention. The truth is, even Amazon's AI checkout that they claimed no human casher is needed is founded to be under large number of human inspectors, according to the aforementioned Scientific American article.

As far as I know, 9 out 10, in fact 99 out of 100 research papers in second tier (or below) AI academic journals are full of re-generation of a few leading papers on different data sets with only a minor change.

The leading papers in AI, like all other fields, change computational methodologies for a fit to new set of data and different purposes, but the technique is unique and it helps a lot of unsolved issues. Going down to second tier or below, it is just a regeneration, so top class researchers usually don't waste time on them. The problem is that even the top journals are not open only for ground breaking papers. There are not that many ground breaking papers, by definition. We mostly just go up one by one, which is already ultra painful.

Going back to my graduate studies, I tried to establish a model for high speed of information flow among financial investors that leads them to follow each other and copy the winning model, which results in financial market overshooting (both hype/crash) at an accelerated speed. The process of information sharing that results in suboptimal market equilibrium is called 'Hirshleifer effect'. Modeling that idea into an equation that fits to a variety of cases is a demanding task. Every researcher has one's own opinion, because they need to solve different problems and they have different backgrounds. Unlikely we will end up with one common form for the effect. This is how the science field works.

Hype that attracts ignorance

People outside of research, people in marketing to raise AI hype, and people unable to understand researches but can understand marketers' catchphrases are those people who frustrate us. As mentioned earlier, I did try to persuade them that it is only a hype and the reality is far from the catch lines. I have given up doing so for many years.

Friends of mine who have not pursued grad school sometimes claim that they just need to test the AI model. For example, if an AI engineer claims that his/her AI can win against wall street's top-class fund managers by double to tripple margin, my friends think all they need as a venture capitalist is to test it for a certain period of time.

The AI engineer may not be smart enough to show you failed result. But a series of failed funding attempts will make him smarter. From a certain point, I am sure the AI engineer begins showing off successful test cases only, from the limited time span. My VC friends will likely be fooled, because there is not such an algorithm that can win against market consistently. If I had that model, I would not go for VC funding. I would set up a hedge-fund or I will just trade with my own money. If I know that I can win with 100% probability and zero risk, why share profit with somebody else?

The hype disappears not by a few failed tests, but by no budget in marketing

Since many ignorant VCs are fooled, the hype continues. Once the funding is secured, the AI engineer runs more marketing tools to show off so that potential investors are brain-washed by the artificial success story.

As the test failed multiple times, the actual investments with fund buyers' money also fails. Clients begin complaining, but the hype is still high and the VC's funding is not dry yet. In addition to that, now the VC is desperate to raise the invested AI start-up's value. He/She also lies. The VC maybe uninformed of the failed tests, but it is unlikely that he/she hears complains from angry clients. The VC's lies, however unintentional, support the hype. The hype goes on. Until when?

The hype becomes invisible when people stop talking about. When people stop talk about it? If the product is not new anymore? Well, maybe. But for AI products, if it has no real use cases, then people finally understand that it was all marketing hype. The less clients, and the less words of mouth. To pump up dying hype, the company may put in more budget to marketing. They do so, until it completely runs out of cash. At some point, there is no ad, so people just move onto something else. Finally, the hype is gone.

Then, AI hype followers no longer send me emails with disgusting and silly criticism.

Following AI hype vs. Studying AI/Data Science

On the contrary, there are some people determined to study this subject in-depth. They soon realize that copying a few lines of program codes on Github.com does not make them experts. They may read a few 'tech blogs' and textbooks, but the smarter they are, the faster they catch that it requires loads of mathematics, statistics, and hell more scientific backgrounds that they have not studied from college.

They begin looking for education programs. For the last 7~8 years, a growing number of universities have created AI/Data Science programs. At the very beginning, many programs were focused too much on computer programming, but by the competition of coding boot-camps and accreditational institutions' drive, most AI/Data Science programs in US top research schools (or similar level schools in the world) offer mathematically heavy courses.

Unfortunately, many students fail, because math and stat required to professional data scientists is not just copying a few lines of program codes from Github.com. My institution, for example, runs Bachelor level courses for AI MBA and MSc AI/Data Science for more qualified students. Most students know the MSc is superior to AI MBA, but only few can survice. They can't even understand AI MBA's courses that are par to undergrad. Considering US top schools' failing rates in STEM majors, I don't think it is a surprise.

Those failing students are still better than AI hype followers, so highly unlikely be fooled like my ignorant VC friends, but they are unfortunately not good enough to earn a demaing STEM degree. I am sorry to see them walk away from the school without a degree, but the school is not a diploma mill.

The distance from AI hype to professional data scientists

Graduated students with a shining transcript and a quality dissertation find decent data scientist positions. Gives me a big smile. But then, in the job, sadly most of their clients are mere AI hype followers. Whenever I attend alum gathering, I get to hear tons of complaints from students about the work environment.

It sounds like a Janus-face case to me. On the one side, the company officials hires data scientists because they follow AI hype. They just don't know how to make AI products. They want to make the same or the better AI products than competitors. The AI hype followers with money create this data scientist job market. On the other side, unfortunately the employers are even worse than failing students. They hear all kinds of AI hype, and they just believe all of them. Likely, the orders given by the employers will be far from realistic.

Had the employers had the same level knowledge in data science as me, would they have hired a team of data scientists for products that cannot be engineered? Had they known that there is no AI algorithm that can consistently win against financial markets, would they have invested to the AI engineer's financial start-up?

I admit that there are thousands of unsung heros in this field without much consideration from the market due to the fact that they have never jumped into this hype marketing. The capacity of those teams must be the same as or even better than world class top-notch researchers. But even with them, there are things that can be done and cannot be done by AI/Data Science.

Hype can only attract ignorance.