AI/ML/DL/NN/LLMs/GenAI/ChatGPT as Make-Believe Projects: from Rule/Rote Learning to Meaningful Learning Machines

The big problem with big tech AI, ML, deep learning, ANNs, LLMs, generative AI, and chatbots is operating "Rote Learning", instead of meaningful learning, associative learning, active learning, or integrative learning.

No rules-based or rote-learning AI, ML, DL, NN, LLMs, GenAI, or ChatGPT, has understanding of the world, being an advanced but unintelligent digital technology.

And it is unintelligible why so many people, from laymen to researchers to politicians, fail to understand that such an AI is nothing more than wishful thinking, which is known to be blind to unintended consequences.

"When we embark on a course of action which is unconsciously driven by wishful thinking, all may seem to go well for a time, in what may be called the 'dream stage'. But because this make-believe can never be reconciled with reality, it leads to a 'frustration stage' as things start to go wrong, prompting a more determined effort to keep the fantasy in being. As reality presses in, it leads to a 'nightmare stage' as everything goes wrong, culminating in an 'explosion into reality', when the fantasy finally falls apart." [What happens when the great fantasies, like wind power or European Union, collide with reality?]

Again, Evidence, Science, Rationality and Reality, All Shows that Meaningful Learning or Understanding is All What REAL or TRUE AI Need to learn REAL patterns, make REAL predictions, and perform REAL-WORLD tasks.

AI/ML as Rote Learning Mnemonic AI Models

All ML/AI modes are RL/AI mnemonic models involving mechanical learning disparaged as stochastic parroting, regurgitation, cramming, or mugging.

The RL/AI machine is programmed to keep a history of calculations and compare new input against its history of inputs and outputs, retrieving the stored output if present. This pattern requires that the RL machine can be modeled as a pure function — always producing same output for same input — formally described as follows:

f( 𝑥1,𝑥2,...,𝑥𝑛) → ( 𝑦1,𝑦2,...,𝑦𝑝) → store (( 𝑥1,𝑥2,...,𝑥𝑛),( 𝑦1,𝑦2,...,𝑦𝑝))

Rote learning memorization techniques as recall of facts is widely used in the learning of elementary/foundational knowledge: the alphabet, phonics reading, the periodic table in chemistry, elementary arithmetic, numbers, multiplication charts, times tables, anatomy in medicine, cases/statutes in law, basic formulae in any science, basic facts, etc.

• RL/AI refers to a form of memorization where its model simply memorizes information without truly understanding its content and context.
• RL/AI system stores and reproduces data or patterns without the ability to generalize and apply knowledge effectively.
• RL/AI learns by memorizing specific examples, as a large dataset of text or images without understanding the content and context.
• RL/AI systems struggle to generalize their knowledge to new, unseen situations. They may perform well on tasks similar to what they've memorized but poorly on tasks outside their narrow scope.
• RL/AI systems typically do not adapt well to changes in data or environment because they lack the ability to reason or adapt their knowledge.
• RL/AI systems are not effective at problem-solving or making decisions based on the information they've memorized.

By training generative models on vast amounts of social media data, such as video or audio, images and text, generative AI can generate the media data tailored to specific user preferences and trends, without any understanding its content or context.

Meaningful Learning AI = True AI = Real AI

Meaningful learning AI (ML/AI) involves understanding how all the pieces of an entire whole, system, concept, information or data fit together. The knowledge gained through meaningful learning could be transferred to new learning situations.

ML AI emphasize the importance of deep understanding over the recalling of facts or statistical memorization.

Meaningful learning refers to the act of higher order thinking and development through intellectual engagement that uses pattern recognition and concept association.

It can include—but is not limited to—critical and creative thinking, inquiry, problem solving, critical discourse, and metacognitive skills.

The concept and theory of meaningful learning is that learned information is completely understood and can now be used to make connections with other previously known knowledge aiding in further understanding.

Since information is stored in a network of connections, it can be accessed from multiple starting points depending on the context of recall.

Meaningful learning is often contrasted with rote learning, a method in which information is memorized sometimes without elements of understanding or relation to other objects or situations.

A real-world example of a concept the learner has learned is an instance of meaningful learning. Utilization of meaningful learning may trigger further learning, as the relation of a concept to a real-world situation may be encouraging to the learner. It may encourage the learner to understand the information presented and will assist with active learning techniques to aid their understanding.

Although it takes longer than rote memorization information, it is typically retained for a longer period of time.

In all, ML/AI involves

• Problem-Solving Learning involving critical thinking and creativity.
• Inductive learning of general principles from specific examples, foundational in mathematics and science, where hypotheses are formed through observation and experimentation.
• Explanation-based learning comprehending the underlying principles behind phenomena or concepts. Instead of memorizing isolated facts, it learns the rationale and mechanisms to apply learned knowledge in various contexts.
• Discovery learning centering on self-directed exploration and experimentation.
• Analogical learning of making parallels between familiar concepts and new, unfamiliar ones.
• Learning by taking advice from the experiences and insights of experts or mentors.

Understanding is what true AI requires

We have to see differences among Data, Information, Knowledge and Understanding.

Data in their simplest form consist of raw alphanumeric values, or facts and statistics.

Data are observable and measurable properties that are represented in signs or symbols, such as numbers or letters, representing stimuli or signals. Data is raw facts or observations, which are unorganized and unprocessed and therefore have no meaning or value due to lack of context and interpretation.

Information is created when data are processed, organized, or structured to provide context and meaning. Information is essentially processed data, providing answers to "who", "what", "where", and "when" questions.

Knowledge is what we know, application of data and information; answers "how" questions.

Knowledge is the accumulation of facts, information, or skills obtained through education or experience or learning and reasoning.

Understanding is the ability to grasp or comprehend knowledge. Knowledge is the accumulation of information and skills, while understanding involves comprehension and application of that knowledge.

Understanding provides a deeper level of insight than knowledge and can be more difficult to measure or acquire. It answers "why" questions enabling wisdom, which integrates knowledge and understanding for intelligent decisions and interactions.

We have to mention here Bloom's taxonomy, cognitive, knowledge based domain levels, dividing comprehension from knowledge, where "knowledge involves the recall of specifics and universals, the recall of methods and processes, or the recall of a pattern, structure, or setting":

The Data-Information-Knowledge-Understanding-Wisdom-Interaction-Environment chain/hierarchy/framework/continuum, or the Intelligence Pyramid, could be partly presented as a Model Mnemonic of a flow diagram of the data, information, knowledge, understanding, wisdom with feedback loops and control relationships.

Note that the OECD AI Principles define an AI system as "a machine-based system that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments. Different AI systems vary in their levels of autonomy and adaptiveness after deployment".

AI system lifecycle phases involve: i) ‘design, data and models’; which is a context-dependent sequence encompassing planning and design, data collection and processing, as well as model building; ii) ‘verification and validation’; iii) ‘deployment’; and iv) ‘operation and monitoring’.

Today, the European Union, the Council of Europe, the United States, and the United Nations and other jurisdictions use the OECD’s definition of an AI system and lifecycle above in their legislative and regulatory frameworks and guidance

All in all, such a RL AI allows false, meaningless, and nonsensical data from the training data models, such that the principle of GIGO, garbage in, garbage out, should be accounted for its RL mnemonic prompt techniques.

Chatbots as a a cause célèbre

Chatbots are now everywhere, customer services, sales and marketing, education, entertainment, sold through AI-based chatbot platforms, as ChatGPT, BotSailor, ChatPion...try and study the AI chatbot landscape in 2024 with its numerous players:

As computer programs or software applications, chatbots are designed to simulate human conversation and interact with users via text or voice.

They leverage natural language processing (NLP) and machine learning to understand, respond to, and learn from user interactions.

Chatbots operate by the if-then and if-then-else rules or by leveraging RL algorithms:

Rule-based chatbot following a set of predefined rules and patterns

ML--based chatbots, trained on conversations or textual data to rote learn patterns, context and statistical language syntax, missing its grammar, semantics, pragmatics and ontology.

Rules-based, retrieval-based or generative chatbots, using Deep Rote Learning techniques, all has no understanding or real meaningful learning, as much as search engines employing indexes with pattern matching for efficient information storage and retrieval,

All in all, today's AI is not artificial intelligence, but rather "artificial memory" mnemonic systems, trained and developed through rote learning and practice a variety of mnemonic techniques or strategies or algorithms.

Today's big tech AI systems are nothing else but Mnemonic Systems with some Artificial Memory.

True as that.

Resources

AI as a Universal Paradigm [for science and technology, industry and society]

AGI Bible: The World Modelling Machine + AI/ML Models + LLMs + GenAI +...