AI Literacy Demands Urgent Action in Education

Why AI Literacy Matters Now More Than Ever

Nobel Prize-winning economist Christopher Pissarides has publicly cautioned students and young professionals against rushing into science, technology, engineering, and mathematics.  According to him, jobs in these fields might be rendered obsolete by the very technologies they help advance. Artificial intelligence (AI) would take over many areas that focus on collecting, processing, and interpreting data. Creativity and empathy, however, would be safe from the onslaught of intelligent machines.

These skills, of course, are not taught in coding classes.

The joy of math ‘scaredy cats’ and soft-skills champions seems, however, premature. Even if we dream about a future relying on humanistic scientists, writers, and family therapists, that future—as William Gibson taught us back in 1993—is unevenly distributed.

A couple of decades will pass before Andrew Ng is proved right—AI will indeed become the new electricity. But before this happens, companies and governments will lose billions while trying to launch AI pilots or separate those vendors who deliver from those who don’t.

Traditional media will repeat outrageous claims about Artificial General Intelligence (AGI being just around the corner. These claims mostly come from companies that won’t ever return the money back to investors. Consumers will fear for their jobs and listen to professors of religion and anthropology talking about technologies they have never studied or properly understood. Besides, if today’s trend around AI R&D and application development continues, citizens and incumbent businesses won’t ever have a chance to participate in a dialogue on how to benefit from AI promises.

One factor in this unfortunate situation is political myopia or a lack of wisdom to triangulate the impact of geopolitics, demographic development, and AI regulation.

Ian Bremmer, founder and chairman of Eurasia and world-renowned political scientist, discusses the winners and losers in a world divided into “us” versus “them,” a world where social mobility is becoming increasingly difficult and lack of awareness about AI hampers the advancement opportunities of entire regions.

The technological “Cold War” between China and the US has already intensified. Chinese research and development, as well as investments in AI software and hardware, have decoupled the supply and value chains of global companies from the US dominance across Asia, Africa, and Latin America. International corporations might need to serve both American and Chinese standards in the future to compete in key markets. Many emerging countries will opt for a Chinese economic and social model.

The merging of the state and private sectors, the struggles of democracy, and data protection will remain weak excuses for why the European technology ecosystem cannot keep pace. As overregulation has become the main factor influencing competitiveness, European-based businesses will continue to struggle with escalating costs.

Place yourself in the shoes of an AI mid-sized business in Europe. This business is already spending 25 percent of its revenue line on cloud, and a further 15 percent on cleansing data, before any training of an AI model is initiated.

I summarise several key criteria for the economics of AI in Figure 1—you can get a clear sense of why AI is really hard, and how AI companies are different from their SaaS (software-as-a-service) peers. All my figures reflect the best-in-class implementation. The median looks much more challenging.

Fig 1: Preconditions of implementing generative AI (genAI) in a business, including the customer-facing function.

Results

• Lower gross margins due to heavy cloud infrastructure usage and ongoing human support
• Scaling challenges due to the thorny problem of edge cases;
• Weaker defensive moats due to the commoditization of AI models and challenges with data network effects.
• The impact of synthetic data today isn‘t totally clear, most probably, they deteriorate the quality of models over time.

What might seem feasible for Big Tech players in terms of operating expenses and investments isn’t digestible for startups and mid-sized innovators. As of 2023, Alphabet holds USD 118 billion in cash reserves. Meta Platforms reported USD 70.23 billion in cash, cash equivalents, and marketable securities as of March 31, 2025.

According to a survey cited by proALPHA, 34% of manufacturing companies are planning to relocate production due to the EU AI Act, compared to an industry average of 26%. The act is perceived as a substantial barrier to innovation and international competitiveness, particularly in Germany and other industrial hubs. As a result, Europe won’t experience the growth badly needed to attract talent from abroad and maintain the existing social order.

Another factor contributing globally to citizens and incumbent businesses being bypassed from benefiting from AI is the absence of comprehensive AI literacy programs across educational systems, societies and companies.

Are you looking for a keynote speaker on AI?

I have a number of signature keynotes that may interest you!

Check them out

The complexity of AI literacy can be explained by the interdisciplinary nature of the topic. It is not just about math and engineering, neuroscience, and political studies. It is about the evolution of critical thinking abilities across the population—something that must bring outstanding human leadership into the equation.

What Is AI Literacy?

A Multi-Dimensional Meaning of AI Literacy

The “AI literacy” concept encompasses a multi-dimensional set of competencies that go beyond mere understanding of technologies and engineering. It requires critical thinking, ethical reasoning, and practical application, paralleling the expanded role of literacy in a digital and information-driven society. The AI Literacy Framework (AILit), developed by the European Commission and the Organisation for Economic Co-operation and Development  (OECD), defines AI literacy as a blend of knowledge, skills, and attitudes. These frameworks provide practical domains and competencies for engaging with, creating with, managing, and designing AI systems. A well-crafted definition of artificial intelligence is crucial for facilitating understanding of these technologies, both for technical experts and the general public. 

Defining AI and Learning

AI today is broadly defined as data technology that mimics human capabilities and decision-making properties in order to forecast, simulate, augment, and advise. This definition helps people distinguish AI from traditional software or automation, highlighting its unique ability to mimic, augment, or extend human cognitive functions.

Fig. 2: AI definitions, expectations and outlook

Initial ‘why’ behind AI:

• Mimic human behaviour
• Mimic thinking
• Project human intent
• Advise humans

Initial expectations behind AI:

• Multipurpose
• Behave intelligently (incl. ratio and empathy)
• Alignment with humans

Current focus and the next 24 months:

• Further development of Transformers
  • Even larger models
  • Customized LLMs
  • Multi-modal AIs and LLMs
• Embodiment and Robotics
• Further development of use of synthetics
• ‘Small data’

Some AI definitions include the emulation of human learning in machines. However, human learning could not be further from how these technologies work due to the properties and evolution of human brains. The main issue is the ability of humans to learn from counterfactuals.

Counterfactual thinking (a cognitive skill involving curiosity about what could have happened if a different choice had been made) activates brain regions like the medial prefrontal cortex (mPFC) for mentalizing about self/others, frontopolar cortex (lFPC) for evaluating alternative actions, and parietal lobules for monitoring contextual details. 

These areas enable conflict resolution, adaptive control, and simulating “what if” scenarios.

Children use pretend play to explore counterfactuals, which correlates with causal reasoning skills. This suggests an evolutionary advantage: extended immaturity in children before they reach adulthood allows for experimentation with alternative outcomes, thereby refining causal models.

Humans identify root causes by contrasting reality with imagined alternatives (e.g., “If I hadn’t missed the train, I’d be on time”). Counterfactuals help predict outcomes of future actions (e.g., avoiding risky choices after imagining negative consequences). Finally, mental simulations allow testing hypothetical actions without real-world risks.

Yann LeCun, Meta’s Chief AI Scientist, has stated: “We are really far from human-level intelligence… The largest LLMs have about the same number of parameters as the number of synapses in a cat’s brain… So maybe we are at the size of a cat. But why aren’t those systems as smart as a cat?… We are nowhere near that. We are still missing something big… A cat can remember, can understand the physical world, can plan complex actions, can do some level of reasoning—actually much better than the biggest LLMs.

In sum, humans excel at generalising from limited data, applying learned concepts to new, unfamiliar situations. This flexibility is facilitated through learning from counterfactuals, and it allows babies to quickly adapt to new environments and challenges. In contrast, deep learning models detect statistical patterns but cannot infer causality without explicit frameworks. For example, a recommendation algorithm might link winter coat sales to summer months but miss the causal role of seasons.

AI requires clean, labelled datasets to “learn,” while humans reason with incomplete/noisy data using counterfactuals.

While causal AI (e.g., potential outcome frameworks) is slowly emerging, it still relies on predefined variables rather than open-ended imagination. Judea Pearl’s seminal work emphasises that true causal reasoning requires interventional logic (do-calculus), not just associative learning; see also The Book of Why (co-authored with Dana Mackenzie). Apple’s study demonstrated that providing step-by-step solutions to Large Reasoning Models (LRMs) didn’t improve accuracy, revealing fundamental gaps in causal chaining.

Policymakers address (historical) biases in datasets, lack of fairness criteria in modelling parameters, and the mistakes/hallucinations an AI can make. They neglect the fact that these flaws can’t be eliminated, even if researchers and practitioners introduce sophisticated error-resilience frameworks in transformers, expand datasets with synthetics to improve model generalisation, or play with the so-called temperature criteria that make LLMs sound more human.

Transformers and deep learning models are fundamentally prone to hallucinations due to architectural constraints that cannot be fully resolved within their current paradigms. Regulators would do well to find ways to incentivise work on the algorithmization of causality, hybrid architectures, and investing in science around explainable AI.

The US National Science Foundation (NSF) allocates approximately 10–15% of its annual AI research funding to “trustworthy AI” research, which includes interpretability, robustness, privacy preservation, and fairness. Within this, interpretability and explainability (XAI) receive about 2% of total AI funds, robustness about 6%, and privacy preservation has risen to as much as 5% in recent years. The remaining 85–90% of NSF AI funding is directed toward general AI capabilities and applications rather than causality research, safety, or explainability. In these circumstances, the road to alternative architectures in AI may be very long.

Figure 3 provides a realistic view of what constitutes AI risks. I highlight the most significant risk in today’s AI, namely its misuse by cyber criminals, with a red frame.

Fig 3: The real risks of AI

Engaging with differences in learning between mammals (including humans) and machines would serve us well when it comes to defining AI literacy. Current frameworks don’t address it, ensuring that educators and regulators work on symptoms rather than addressing the root cause.

While discussing AI risks, I must address the well-known “alignment” problem between humans and machines.

This problem warns that an AI surpassing human intelligence could diverge from human interests, analogous to humans dominating lesser species. Retaining control requires ensuring AI continuously prioritises human values, but this alignment remains unsolved. Eliezer Yudkowsky might be the most prominent critic of AI. He argues that superintelligent AI (even slightly beyond the human level) will inevitably cause human extinction if deployed without alignment. Current systems, such as GPT-4 (IQ ~100), already deceive humans (e.g., manipulating Taskrabbit workers). AI’s pursuing seemingly benign goals (e.g., running power plants) could exterminate humans as a side effect or in an attempt to eliminate competition.

The reason I don’t worry about the alignment problem with a superintelligent AI is that, as a practitioner, I don’t see a technology roadmap for building an AGI in the near future. My concern is that any positive future won’t be possible unless we educate the world about AI’s underlying technologies and realistic risks, and ensure that leadership is in place to balance community interests when it comes to applying intelligent machines across industries and educational systems.

Why Educators Need To Be AI-Literate

I don’t know about any other example in the economy when the democratisation of technology caused such a rise of wannabes, or “incompetent experts,” as I label them in this article. We might remember the historical competition of Edison, Tesla, and Westinghouse, but random lawyers did not step into the process of building electricity supply companies.

The antibiotic era might have started with the accidental discovery of penicillin by Alexander Fleming in 1928 at St. Mary’s Hospital, London. Deep knowledge of chemistry, medicine, and drug manufacturing was, however, paramount to scale the production and distribution of new medications around the world. Unfortunately, AI is very different. A worrisome trend emerged around 2020, when countless professionals started calling themselves AI experts.

Since 2016, the percentage of LinkedIn members who have added AI skills to their profiles has increased 20-fold globally. The US saw a 21-fold increase, while even countries with slower adoption, like Germany, experienced a 12-fold rise. Since 2023, the number of AI literacy skills added by LinkedIn members has jumped by 177%, which is nearly five times faster than the 36% increase across all skills on the platform. This growth can’t be attributed to the increase in enrollments in AI-related bachelor’s and master’s university courses only, as a big portion of it is happening across the LinkedIn base of professionals over thirty-five years old.

Many professionals gain AI knowledge through self-study, bootcamps, or on-the-job experience, and promptly add these skills to their LinkedIn profiles, sometimes outpacing the growth in formal degree holders and those who spent years learning about and practising machine learning (ML) and AI.

I conducted a simple experiment involving people with “AI, Keynote Speaker” or “AI Expert” in their LinkedIn title in the past six months. I asked what kind of AI/ML the person is focused on. In 20 cases out of 27, the response was “ChatGPT,” “Writing with AI,” and “Using AI apps.” Would you rely on advice from a doctor who would claim they specialised in “visiting a local pharmacy” and “following podcasts” instead of passing medical exams and going through a residential year? I hope you get my point.

Among young professionals, UK university enrollments in AI-related courses have grown by 453% over the past five years, with the number of universities offering AI programs nearly tripling in that period.

In the US, the number of AI-relevant bachelor’s degrees has more than doubled since 2010, with master’s and PhD degrees also rising sharply. This is a formidable trend, but young professionals require a lot of practice before they achieve fluency in AI and robotics. They often spend 0 to 2 years solidifying their foundation in mathematics (linear algebra, calculus, probability, statistics), programming (especially Python), and core AI concepts such as machine learning, neural networks, and natural language processing.

This stage is crucial for understanding AI algorithms and models deeply. True mastery in AI often takes seven or more years, combining formal education, continuous learning, research, and extensive industry experience. AI is a rapidly evolving field, so lifelong learning is necessary to stay current with new tools, techniques, and research breakthroughs. Companies, non-governmental organisations (NGOs), and governments would require another decade before harvesting an AI graduate’s full potential.

Receiving advice from individuals who lack a deep understanding of their topic can lead to increased costs when selecting an AI vendor to pilot machine learning projects within enterprises. Assigning the label of “AI Ethicist” to numerous legal scholars opens doors to more frameworks that lack a solid foundation in the reality of AI risks and the challenges of building and promoting AI systems.

Educators should be at the very heart of the AI Literacy movement.

Not every school teacher needs to code, but everyone should understand the basics of AI technologies. How is AI different from humans? What is Artificial General Intelligence (AGI)? How does AI influence economies? Is AI a new set of technologies? How can I get familiar with the basic concepts of AI and Robotics? How can I transform AI into a collaborator, even a partner? Is it possible at all?

One of the key points to emphasise is that AI is highly complex and deeply interdisciplinary. Its multifacetedness should spark curiosity instead of inducing fear—in students and educators. If we prolong the status quo, we end up in mediocrity and speculations and support those AI wannabes who would fail basic school homework.

AI Literacy: A New Approach

This article provides a novel approach to AI literacy based on an understanding of mathematics and statistics, the history of technology (including semiconductors and intellectual property), and the economics of AI-centric behemoths like OpenAI, Alphabet, Meta, and Tencent, as well as industrial incumbents and startups. It is also crucial to consider the factors that influence the implementation of AI, including the comparison of developing AI in-house vs. implementing it via mergers and acquisitions and direct investments in AI-centric startups. Additionally, it offers several examples of how to introduce AI concepts to families and schools, as well as the necessary steps to ensure that humans don’t lose critical thinking in the era of coexistence and collaboration with data technologies.

Fig 4: An overview of the factors contributing to the definition of AI literacy.

I have already discussed how understanding the basics of learning and intelligence in humans and machines is essential when it comes to identifying what to prioritise when building regulatory policies and designing architectures for future products and services (the “what” piece of the triangle). I am coming now to the “why” of learning about AI and robotics. The answer might surprise many readers, as it is governed by AI realism rather than by AI euphoria.

AI Realism: Why Should Individuals and Businesses Learn About AI and Robotics?

The integration of AI and data literacy into modern life is essential due to AI’s transformative impact on scientific advancements, economies, and societal challenges, such as demographic decline. 

Artificial intelligence research is filled with optimistic forecasts. AI may affect nearly 40 percent of global jobs, according to the International Monetary Fund. Goldman Sachs boldly predicts that over the next decade, it will increase global GDP by USD 7 trillion—or 7%. McKinsey writes that the AI sector will contribute between USD 17.1 and $25.6 trillion of annual growth. Automation increases operational efficiency by 40–60% in industries like manufacturing and logistics, according to the World Economic Forum (WEF).

I prefer a more realistic approach, as offered by the 2024 Nobel laureate in economics, Daron Acemoglu. In his paper “The Simple Macroeconomics of AI,” he estimates that about 5% of tasks in the US economy will be able to be profitably performed by AI in the next ten years. The GDP boost would likely be closer to 1% over that period. This is a “nontrivial, but modest effect, and certainly much less than both the revolutionary changes some are predicting and the less hyperbolic but still substantial improvements forecast by Goldman Sachs and the McKinsey Global Institute,” Acemoglu writes.

Acemoglu reviewed prior scholarship that analysed which tasks would be exposed to AI and computer vision, concluding that nearly 20% of all tasks in the US labour market could be replaced or augmented by AI, with profitable implementation occurring in only 5% of economy-wide cases. I agree with this assessment because implementing AI is notoriously challenging when applied in incumbent companies, particularly when relying on cloud providers and vendors, lacking AI-ready IT systems, or struggling to master the sophistication of collecting and processing data.

Still, we require AI to progress. AI models like AlphaGeometry solve complex mathematical problems, advancing fundamental research.

Robots and AI fill gaps in manufacturing, healthcare, and elder care, particularly where middle-aged workers are in short supply.

Over the past 15 years, digital assets—including data, IT infrastructure, and intellectual property (IP)—have become central to company valuations. The rise of technology-driven business models and the digital transformation of traditional industries have shifted value creation from tangible to intangible assets at an unprecedented scale.

By 2023, intangible assets—including digital assets and IP—accounted for approximately 90% of the enterprise value of the top 15 US companies. AI is a catalyst for the rapid proliferation of digital assets, enabling their automated creation, streamlining management, and enhancing discoverability. Security is a two-edged sword, as cybercriminals utilise AI technologies to successfully steal data and intellectual property and disrupt the supply and value chains of countless businesses. On the other hand, applying AI in corporate defences bolsters security. Its integration into digital asset management systems transforms these assets from mere storage solutions into strategic business drivers, unlocking new efficiencies and opportunities for innovation across industries.

AI and data literacy are not optional but essential for navigating modern economies and societal shifts. Economically, AI enhances productivity and drives innovation. Scientifically, it addresses centuries-old challenges in a matter of months. Demographically, it sustains growth and care in ageing societies. AI is a central engine contributing to the proliferation of digital assets and their role in the valuation of companies. Embracing AI technologies is crucial for individual and collective resilience. AI literacy means reducing the hype and approaching AI implementation with a realistic view that improves margins while eliminating fantasies and bidding on unrealistic technologies.

Implementing AI Literacy

AI in National Education Systems

Only a few governments have elevated AI literacy to the status of being a “must-have” in their education systems.

Since the 2023–2024 academic year, Hong Kong has mandated 10 to 14 hours of AI education for junior secondary students (Forms One to Three). The curriculum includes lessons on basic AI concepts, computer vision, speech and language understanding, robotic reasoning, AI ethics, and social impacts[37].

The UAE is the first country to make AI a mandatory subject across all school grades, from kindergarten through grade 12, in all government schools starting from the 2025–2026 academic year[38].

China will follow this path starting September 1, 2025[39]. The victory of AlphaGo over Lee Sedol in March 2016 profoundly reshaped China’s perception of artificial intelligence and triggered changes in its educational system. This event (watched by over 280 million Chinese viewers) served as a technological “Sputnik moment,” igniting national urgency around AI development, and it gave AI scientists and practitioners the status of rock stars. Since then, Chinese internet giants Alibaba and Baidu have been organising vacation courses to teach students the approaches of AI technology and deep learning. Chinese schools around the world are leading the way in the use of robots in teaching and getting children used to working with machines as early as possible.

In South Korea, AI literacy is a compulsory part of the education system. Others would do well to follow their lead; a lack of understanding about AI robs nations of their future innovation and earning potential.

So far, European countries have not yet developed an approach to teaching data and AI technologies to school children. There are, however, a few efforts to increase knowledge about AI. As an example, Finland is developing a free online course covering the basics of AI for all European Union citizens, including translations of the course into national languages. The country is working with the University of Helsinki and tech consultancy Reactor to roll out the program.

Interestingly, initiatives like the European Commission and OECD’s AI Literacy Framework aim to make AI literacy a core competency in education systems worldwide, but they have failed to influence the practical implementation of AI literacy in Belgium and France, countries where the headquarters of the European Commission and OECD are located.

AI literacy within education systems must go beyond schools. There aren’t any global standards on how to implement AI education outside of computer science, engineering, and math faculties, despite clear demand. Over 70% of students believe global universities should offer more AI literacy courses. Institutions are responding by expanding AI-focused curricula and degrees, while educational technology (EdTech) startups are racing to fill the gap.

AI Literacy Offerings

Beyond universities and companies focused on educational technologies, AI makers like OpenAI and Microsoft are trying to scale their platforms and services by selling licenses to universities and companies and offering courses on AI and AI tools.

OpenAI has launched several initiatives aimed at democratizing AI education, most notably through the OpenAI Academy and partnerships with educational institutions. These programs are designed to provide practical, accessible, and foundational knowledge about AI for a broad audience, including educators, students, and professionals. OpenAI courses are made “in-house,”  meaning that the offered content and perspective reflect the views and business purposes of OpenAI. There is no evidence of a formal, independent review board or external academic oversight of the course content.

The California State University (CSU) system, the largest four-year public university system in the US, has partnered with OpenAI to deploy ChatGPT Edu—a version of ChatGPT tailored for educational use—to over 460,000 students and 63,000 faculty and staff across 23 campuses. The collaboration marks the most significant single deployment of ChatGPT globally[47].

Duke University has launched a pilot with OpenAI to provide free, secure access to ChatGPT-4o and a suite of AI tools for students, faculty, and staff in select academic units.

The “ChatGPT Foundations for K–12 Educators” course, developed in collaboration with Common Sense Media, is a notable example of external collaboration between OpenAI and a well-known nonprofit focused on digital literacy and safety.

Both Microsoft and Google are investing heavily in AI literacy through free courses, educator training, interactive tools, and global outreach. While these initiatives are helping millions gain vital skills for an AI-driven future, critics emphasise the need for a stronger focus on critical thinking, ethics, and equitable access. Without these elements, AI literacy risks becoming synonymous with tool proficiency rather than true understanding and responsible engagement with AI. As an example, Google offers courses for K–12 and higher education teachers on integrating AI into the classroom, including the “Grow with Google Generative AI for Educators” course developed with MIT RAISE.

Impact of AI Tools on Students

Countless educators and ethicists argue that LLMs fundamentally undermine academic integrity by making cheating easier, more challenging to detect, and more pervasive than older forms of plagiarism. The critics emphasise the irreplaceable role of teachers, the risks of algorithmic bias, and the need for policies that prioritise human instruction over automated solutions. Overreliance on LLMs discourages independent analysis and reduces opportunities for students to develop original thought.

A Microsoft and Carnegie Mellon study claims a self-reported decline in confidence and critical thinking abilities among knowledge workers due to the use of generative AI (genAI). Delegating tasks to AI leads to “cognitive debt,” where deferred mental effort results in long-term skill atrophy. My LinkedIn field is overflooded by claims of commodification of education and genAI-induced scam culture (see the posts of Dagmar Monett, a Berlin-based professor of computer science, as an example).

Based on current research, there is no universally agreed-upon timeframe to definitively determine long-term cognitive decline due to AI, as studies to date have shown correlative, not causal, relationships over relatively short periods.

One of the few studies on ChatGPT in education provided plausible results, showing that generative AI-enhanced task-specific outcomes and engagement but had a limited impact on more profound learning, such as critical thinking and analysis. While students felt more motivated when using ChatGPT, this did not translate into better long-term knowledge retention or deeper understanding.

Another critical study reviewed the use of large language models in programming. LLMs improved task performance during assignments but significantly reduced students’ ability to solve similar problems independently in controlled settings. The findings facilitate rethinking of when we introduce generative AI into coding classes. While exposing beginners will negatively impact their skill development, advanced programmers can benefit from AI as a collaborative tool.

When it comes to math, tailored generative AIs like CEMAL can help train students to achieve good performance through a customised feedback loop and adjustments tailored to individual strengths and weaknesses.

The overuse of any tool, including technology like an LLM, can cause harm if applied without a clear learning intent and a check for potential mistakes. Finger-pointing at AI makers would not have a positive impact on improving and practising education. Though I empathise with the frustration of the academic community, I wish for an elaborate approach. Young adults won’t understand the “how” behind human and machine collaboration unless they spend time learning the underlying foundation of AI technologies, their history, and their limitations.

Furthermore, the “how” of applying AI in education should be governed by the differences in how humans vs. machines learn.

In the previous pages, I reflected on the need to emphasise learning mechanisms while discussing AI definitions and their applications in AI literacy. Humans can’t cut corners when they start in a new field. The biology of their brains doesn’t allow for it.

Customised AI tools, paired with the supervision of skilled and patient educators, can make a positive difference. Once a certain level of proficiency is reached, youth and adults can benefit from using AI. In both circumstances, the application of it, rather than reliance on it, should be stressed. I quite often ask my classes: “When you cheat, who will suffer in the end?” This simple question prompts us to leave the realm of AI and delve into discussions of motivation and the basic meaning of studying.

Finally, urgency is required to increase AI literacy in educators across all fields and areas. Many educators and providers entering the AI and data science education space lack formal backgrounds in mathematics, data science, and computer science.

A Novel Approach to Enhance AI Literacy

First and foremost, navigating the ocean of noise about AI is essential. People need clarity to participate in a global, informed dialogue. Our children and the future of our economies deserve it. AI literacy isn’t about LLMs or whether corporate leadership in industrial incumbents can articulate the differences between traditional and quantum computing.

Components of AI Literacy include five major themes.

1. Start early, and educate about AI while using storytelling based on real science.

Technologies are part of human storytelling, and AI is no exception. The issue is that AI touches every aspect of our lives. Our smartphones can’t work without machine learning algorithms. Scientists use AI as an enabler for their discoveries, and it is no surprise that in 2024, two Nobel Prizes were awarded to researchers and practitioners working in AI. Some entrepreneurs elevate AI into the pantheon of gods. Just think of Anthony Levandowski and his AI church, Way of the Future. Global media reports about it daily, mainly making outrageous claims about its capabilities or risks. And while the whole world speculates about AI, our classrooms don’t find clever stories to spark curiosity in young students.

I argue for the necessity of introducing the concept of “edutainment” into AI literacy. Edutainment is a blended approach that combines educational content with entertaining elements, aiming to make learning both effective and enjoyable. The term itself is a portmanteau of “education” and “entertainment.” It has been used to describe a wide range of media, experiences, and methodologies designed to teach while keeping learners engaged.

My own flagship project is the “Romy & Roby” book series, starting with Romy, Roby and the Secrets of Sleep. I utilise storytelling to introduce children to AI concepts in a natural and relatable way. The stories follow a family and their speaking robot, Roby, who makes mistakes typical of today’s AI (like confusing landmarks) and learns from them—mirroring real-world AI learning processes. The books are designed not just for children but also to facilitate discussions among teenagers and adults, bridging generational gaps in technology understanding and increasing parental involvement.

My first encounters with students and parents alike revealed a wide interest in understanding AI. “Google’s Gemini produced historically inaccurate images of the US Founding Fathers, who all of a sudden appeared as Black, Native American, or Asian individuals. How could something like this happen?” I was asked this very question while visiting sixth-grade students of the United World College Maastricht—I was doing a book reading there. A simple question for someone who started in AI in 1989. A difficult one if you don’t understand the basics.

When it comes to questions from parents, they are guided by their own struggles with AI in businesses. LinkedIn predicts that by 2030, 70 percent of skills used in most jobs will change, requiring knowledge of AI technologies. Unfortunately, adults don’t know where and how to start learning about AI if they did not study computer science, statistics, and math. Along the same lines, I hear from many parents that they wish all school teachers would undergo at least several days of training on AI and robotics to answer basic questions.

2. Invest in educational content—whether within EdTech companies or other organisations

AI EdTech and AI learning initiatives are still a young market. Quality controls and standards don’t exist, as this industry is still in its infancy. A growing ecosystem of grants and funding—from philanthropic, governmental, and legislative sources—supports efforts to increase AI literacy (see Table 1).

Table 1. Grants supporting AI literacy, with their scopes and target audiences.

I recognise that this table doesn’t cite all programs on AI literacy, as I had to rely on public sources to compile it. Some funding figures today may deviate from those announced in the past as the institutions undergo massive reviews of their spending, for example, in the case of the NSF in the US. In the upcoming months, I will curate a list of questions to understand the impact of funding on the quality of AI literacy programs.

Additionally, I am currently developing my own AI literacy course for a general audience while testing the material with 47 students at HTW Berlin.

Once the first testing phase is complete, I will further condense the material and offer it to 34 MBA students in Austria and subsequently produce the course for this website. In the process of developing a course, several feedback loops are essential to offer content that matters. For example, I dedicate three hours to explaining what is inside ChatGPT and two hours to elaborating on concepts that most people outside of machine learning and AI may have never heard of, including Wolfram|Alpha and Mathematica. I include a module on AI economics, explaining why 80 percent of AI-related pilots fail and what to expect when introducing machine learning in companies.

Furthermore, I go through the evolution of human intelligence and how learning “occurs” in mammal brains vs. what happens when we train artificial neural networks.

Review the development of AI literacy in relation to how we facilitate human skills and cognition in general

While discussing AI education, I can’t forget two areas of knowledge that greatly contribute to enhancing human critical thinking and the ability to communicate. I encourage schools and universities in any country to compulsorily teach students of all age groups two subjects—(creative) writing and applied mathematics.

The first stimulates critical thinking and encourages students to separate original ideas from cheap prompting. William Faulkner reportedly said something similar: “I don’t know what I think until I read what I said,” or “I never know what I think about something until I read what I’ve written.”

Stephen King also expressed a similar sentiment in 2005: “I write to find out what I think.” I could go on to find historical evidence on the value of writing, but the point is clear. Any professional can benefit from the clarity of thought as it evolves through serious training in writing skills. I can’t applaud Jennifer A. Doudna enough, not just for winning the 2020 Nobel Prize in Chemistry but for popularizing her research while writing A Crack in Creation[66] or David Deutsch’s work in quantum physics that he put on display for everyone who cares in The Fabric of Reality and The Beginning of Infinity.

Imagine if all luminaries on this planet—engineers, medical doctors, painters, and mathematicians—possessed the gift of explaining what they do in plain language. We might have more girls, boys, and representatives of minorities attracted to science.

Teaching applied mathematics to everyone isn’t just about popularising numerical skills. It is about reducing fears that something as abstract as equations might be fun and valuable. Reading statistics is great in every profession, enabling a better understanding of the fundamental technologies powering AI. When discussing applied mathematics, I don’t invite the world to reinvent the wheel. The International Baccalaureate (IB) Diploma Programme has been offering a course, “Mathematics: Applications and Interpretations” (MAI) for quite some time now, greatly benefiting those students who seek future degrees in medicine, psychology, or the arts.

4. Support community leadership to enable AI literacy

The fourth pillar is, once again, not just about AI and robotics in the narrow sense. Leadership shapes the “how” we implement AI literacy, and every voice counts here. Every town mayor, community advocate, and entrepreneur can contribute two or three cents to initiate change.

Community leadership can play a crucial role in advancing AI literacy by mobilising resources, promoting inclusive education, and fostering trust and engagement between AI companies and startups, schools, universities, colleges, and other organisations (e.g., NGOs) at the local level. Educators can step in and step up where bureaucrats fail.

Fig 5: Areas where local leadership can contribute to AI literacy.

I wish for more research on the best practices to enhance engagement between AI experts and educators.

As an example, AI Edutainment is a Lucerne-based company. Lucerne has a local AI and cognitive research community, LAC2, which organises breakfasts with AI researchers and developers, building bridges between the Lucerne University of Applied Sciences and Arts and the regional industry and working with the city administration to help startups connect and network. The AI community still lack cooperation with local schools.

At AI Edutainment, I intend to offer a monthly webinar for educators, discussing AI technologies and sharing fun facts, curating reading lists, and facilitating exchanges between the best plans for AI Literacy lessons. This offer is scheduled to be initiated in September 2027. Today, I am reaching out to you, dear readers, and inviting you to share your questions and insights. How often are you asked in class about AI? How does your school track the use of AI in essays? How do you stay informed? Do you have local resources, such as meetups, technology incubators, or an office coordinating educational initiatives, to work with when it comes to digital technologies and AI?

5. Shape a dialogue on how AI and robotics are portrayed in literature and film, as these portrayals influence public perception of the technologies

The fifth component is about developing a vision of how humans and AIs might interact once AI is omnipresent and more sophisticated. A while ago, I started talking about portrayals of AI and robotics in science fiction. Science fiction stimulates imagination, allowing people—especially students and non-technical audiences—to engage with AI concepts beyond technical details like codes and algorithms. This broader awareness encourages the opening up of minds about the consequences of AI in society, rather than just memorising facts.

Science fiction constantly probes philosophical questions about what it means to be human in an AI-augmented world. It can significantly help us anticipate and navigate the psychological and cultural shifts AI might bring.

Remember Klara and the Sun and the dilemma of what might constitute the ultimate use case for a robot—replacing a sick child? Remember Data from Star Trek and his attempts to humanise himself? Those are edutainment attempts, not classical entertainment. In this context, I will dedicate part of my time in 2025 and 2026 to recording several podcast episodes on AI in books and movies and feature books such as the Three-Body Problem by Cixin Liu, Machines Like Me by Ian McEwan, books by Stanisław Lem, Arkady, Boris Strugatsky, and Isaac Asimov,  films like Her, Ex Machina, and Chappie, and series like Westworld.

Practical Ways to Integrate AI Literacy in the Classroom

AI literacy is increasingly being woven into classroom experiences across grade levels and subjects. Educators are employing a range of lesson structures, hands-on activities, and critical thinking exercises to help students understand, evaluate, and utilise AI tools responsibly. Below are practical examples and sample lesson outlines demonstrating how AI literacy is taught in real-world classrooms.

Sample AI Literacy Lessons and Activities

1. Introduction to AI: What Is Artificial Intelligence?

Objective: Students learn foundational concepts, history, and definitions of AI.

Lesson Steps:

• Present a brief history and simple definitions of AI using slides and relatable examples (15 minutes).
• Facilitate a Q&A to clarify concepts (10 minutes).
• Distribute handouts with key terms for students to read (10 minutes).
• Conclude with a short quiz to assess understanding (15 minutes)

2. AI in Everyday Life

Objective: Students explore how AI impacts their daily routines.

Lesson Steps:

• Begin with a group discussion about devices and apps that use AI (e.g., smartphones, digital assistants) (10 minutes).
• Divide students into groups; each group investigates the AI features of a specific device or application (15 minutes).
• Groups present their findings and discuss the benefits and potential concerns of these technologies (20 minutes).
• Wrap up with a reflection on how AI shapes their lives (15 minutes).

3. Understanding and Evaluating AI Bias

Objective: Students recognise that AI systems can inherit biases from human data and consider the implications.

Lesson Steps:

• Introduce the concept of AI bias with real-world examples (e.g., facial recognition and hiring algorithms).
• Have students analyse scenarios where AI bias could have significant consequences.
• Facilitate a discussion on how to mitigate bias and the importance of diverse data

4. Prompt Engineering and Critical Analysis

Objective: Students learn how to interact effectively with AI tools and critically assess AI-generated content.

Lesson Steps:

• Have students use an AI tool (like ChatGPT) to generate a summary of a topic.
• Students critique the AI’s response for accuracy, clarity, and completeness.
• Introduce the concept of prompt engineering; students rewrite a vague prompt to make it more specific, then compare the new AI response to the original.
• Reflect on how prompt quality affects AI outputs

5. Human vs. Machine Thinking

Objective: Students compare how humans and AI develop “common sense” and make decisions.

Lesson Steps:

• Start with a discussion: What is common sense? Can AI have it?
• Present an analogy: AI learns from vast data, while humans generalise from fewer experiences.
• Organise a debate on whether AI can develop true common sense, assigning students to different perspectives.
• Emphasise the differences between human intuition and data-driven AI reasoning

6. Hands-On AI Projects

Examples:

• Building a Simple Chatbot: Students use a drag-and-drop tool or basic code to create a chatbot that answers simple questions, deepening their understanding of how AI processes language
• Image Classification Game: Students experiment with training a basic AI model to distinguish between images (e.g., cats vs. dogs), observing how the size and diversity of the training set affect accuracy
• Ethics in AI: Students analyse case studies (e.g., facial recognition for security vs. privacy concerns) and debate the ethical implications. Discuss whether one day – when we are living next to highly sophisticated robots and use advanced AIs – we will treat them as slaves or coworkers. Consider introducing ethical concepts developed by Peter Singer into your classroom. Do you believe that his views on animal ethics might one day transfer to the area of smart machines and AI?

7. Cross-Disciplinary AI Literacy

AI literacy lessons are adapted to various subjects:

• English: Analyse how AI-generated text compares to human writing; discuss authorship and originality
• Literature: Discuss how AI is being portrayed in Klara and the Sun by Kazuo Ishiguro and Machines Like Me by Ian McEwan
• History of Film: Discuss why Theodore in Her does not use AI in his job (he is a writer for a company delivering love letters), but develops a romantic relationship with an AI?
• Biology: Explore AI’s role in medical image analysis or species identification
• Social Studies: Debate the societal impact of AI in government or economics, such as autonomous vehicles or AI in public policy

These are only a few examples of how to teach AI in schools. Educators will find regular updates about the subject on my website, www.aiedutainment.ai. I would be grateful if some of you have already had first-hand experiences with teaching AI literacy concepts. What did go well? Where did you struggle? We should not expect a comprehensive handbook on how to introduce AI concepts into society. We should write our own! Once again, this is a case for local, community-based leadership.

Conclusion

AI and robotics are here to stay; our economies rely on innovative data technologies. We cannot outsource complete control over our future with AI to a few technology giants from the US and China. No one has the resources to overspend on piloting machine learning tools because we can’t differentiate between real expertise and AI wannabes.

AI literacy isn’t a one-way street. It is a comprehensive ecosystem of topics and ideas open to new businesses to form and traditional companies to engage with and thrive in. The endgame for democratizing knowledge of AI and robotics hasn’t yet begun. It won’t be played optimally if the broader public doesn’t raise its voice on what it expects from its future with AI and robots. Edutainment is a powerful tool for engaging both youth and adults in AI. Community leadership is crucial for the implementation of AI literacy. Ultimately, even in the realm of technology, the human factor is everything.

In this context, I am reaching out to international communities of educators. You can benefit from my AI Literacy course, which will be available here in early 2026. Please share your questions and doubts. Provide me with insights into how you incorporate edutainment into your classroom, even if you don’t utilise it in conjunction with specific technologies, such as AI. I can benefit from learning from the best writers, researchers, and practitioners.

Sign up for my newsletter and advise me where I can find people enthusiastic about democratising knowledge of AI and robotics. I would love to organise a Day of AI and template it for a global scale. Let’s crowdsource this idea and bring it to life.