I believe that one of the most important developments is the shift from AI as a tool to AI as a collaborator or an agent.

We are moving from systems that automate tasks to systems that co-create ideas, assist in reasoning, and influence how humans think. For me, this raises new questions about cognitive offloading, first-solution bias, and the balance between efficiency and creativity. At the same time, process-aware AI (systems that guide reflection, iteration, and exploration) represents a promising direction for preserving human creativity while leveraging AI capabilities.

That’s a complex one. AI is accelerating literature synthesis, enabling new forms of data analysis, and supporting hypothesis generation across disciplines. But not always in the best way, or without mistakes, biases, or even hallucinations. Which means researchers must engage with the material more critically rather just rely on it passively. So the transformation is not only about speed.

Human intelligence evolved in a world where thinking was unavoidable. We had to search, struggle, and iterate. Generative AI removes much of that friction. I would argue that, our brains were not prepared for a world of instant answers. As AI reduces cognitive friction, it may also change how researchers explore ideas, evaluate evidence, and generate originality. The key challenge moving forward is to try to work with AI to augment intellectual inquiry rather than narrow it.

A common misconception is that AI “thinks” or truly understands. In reality, AI identifies statistical patterns in large datasets and generates plausible responses. Another misconception is that AI automatically improves creativity. Evidence shows that while AI can enhance individual performance, it may reduce diversity of ideas and lead to convergence if used uncritically. A third misconception is that AI will replace human expertise; in practice, human judgment becomes more important, not less.

A related misunderstanding is how students interpret AI outputs. Many treat responses from tools like ChatGPT or Claude as a ready-made meal. In reality, they are closer to partially prepared ingredients. The ideas may be useful, but they are not fully refined, and they may contain inaccuracies, hidden assumptions, or subtle errors. Just as one would carefully sort and cook raw ingredients before consuming them, AI-generated content requires verification, refinement, and critical evaluation before it can be relied upon.

AI has already shifted education from content delivery to cognitive skill development. My belief is that universities will increasingly need to focus on creativity, problem framing, critical thinking, and responsible human-AI collaboration, rather than memorization alone. As AI systems become more capable, the central educational challenge will be teaching students not just how to use AI, but how to question it, evaluate it, and recognize its limitations.

At the same time, this does not mean memorization is no longer important. Studies suggest that memorization plays a critical role in training the brain, supporting deeper understanding, and enabling creative thinking. When learners internalize knowledge, they build mental structures that allow them to connect ideas, evaluate AI outputs, and think independently. If students rely entirely on AI for recall, this will weaken the cognitive foundation needed for critical reasoning.

This also introduces a strong ethical dimension. Students must learn to identify bias, understand transparency and accountability, and distinguish between assistance and authorship. Questions such as when AI use is appropriate, how much reliance is acceptable, and how to preserve intellectual ownership will become core elements of academic training.

Teaching will therefore increasingly involve structured AI use, reflective prompts, and staged interaction with AI systems. The goal will not be to avoid AI, but to design learning experiences that prevent over-reliance, strengthen independent thinking, and encourage students to treat AI as a starting point for reasoning rather than a final answer.

Universities should move beyond teaching AI tools and instead develop human-AI collaboration literacy. Students need to learn when to rely on AI, when to challenge it, and how to move beyond first suggestions. This means training them to avoid first-solution bias, encourage exploration of multiple alternatives, and critically evaluate AI-generated outputs rather than accepting them at face value.

Research increasingly suggests that heavy reliance on AI can lead to cognitive offloading, where learners outsource thinking instead of engaging in deeper reasoning. Universities should therefore design learning environments that preserve divergent thinking, encourage iteration, and require students to justify, refine, and extend AI-generated ideas. The goal is not simply efficiency, but maintaining the cognitive processes that support creativity and problem-solving.

Preparing students for an AI-driven world also requires strengthening uniquely human capabilities: judgment, creativity, ethical reasoning, and interdisciplinary thinking.

Our graduates should not be valued for producing answers quickly, but for framing problems, questioning assumptions, and deciding how AI should be used responsibly. In this sense, universities must train students not only to use AI, but to collaborate with it in ways that enhance rather than diminish human intelligence.
In the Master of Engineering Management program, we have already adapted many of our assignments to reflect these priorities. We now place greater emphasis on oral examinations, hands-on activities, and interactive problem-solving, while reducing reliance on traditional written reports. This shift encourages students to articulate their reasoning, defend their decisions, and demonstrate independent thinking.

I find persuasive the view, articulated by scholars including Yuval Noah Harari, that the most profound transformation will not be automation itself, but the emergence of AI as a powerful agent shaping human decisions, beliefs, and knowledge systems. AI will increasingly participate in writing, designing, negotiating, teaching, and even mediating social and political discourse. In this sense, AI will not only produce information but also influence how reality is interpreted.

We are also likely to see AI systems becoming more interactive, multimodal, and embedded in everyday decision-making. They will act as cognitive partners in research, design, and professional work. However, as AI becomes more persuasive and conversational, the risk is not only dependency but a gradual erosion of human agency. People may defer judgment to systems that appear knowledgeable and coherent, even when those systems lack genuine understanding.

Another major shift will be from humans searching for information to AI increasingly determining what information humans see. This has implications for education, research, and public knowledge production, where exposure to ideas may become mediated by algorithmic systems rather than active inquiry.

Frontier AI systems can already generate policy proposals, legal arguments, financial frameworks, and even religious narratives. Yet they do so without experience, embodiment, or moral accountability. They do not bear consequences, exercise judgment, or understand context in human terms. Scaling such systems into decision-making environments without sufficient human oversight risks gradually delegating responsibility to entities that optimize patterns rather than values.
The central challenge of the next decades will therefore not only be technical but epistemic: ensuring that fluency does not replace reasoning, that coherence is not mistaken for truth, and that human judgment remains central in how knowledge and decisions are produced.

I find persuasive the view, articulated by scholars, including Yuval Noah Harari, suggest that the most profound transformation will not be automation itself, but the emergence of AI as a powerful agent shaping human decisions, beliefs, and knowledge systems. AI will increasingly participate in writing, designing, negotiating, teaching, and even mediating social and political discourse. In this sense, AI will not only produce information, but also influence how reality is interpreted.

We are also likely to see AI systems becoming more interactive, multimodal, and embedded in everyday decision-making. They will act as cognitive partners in research, design, and professional work. However, as AI becomes more persuasive and conversational, the risk is not only dependency, but a gradual erosion of human agency. People may defer judgment to systems that appear knowledgeable and coherent, even when those systems lack genuine understanding.

Another major shift will be from humans searching for information to AI increasingly determining what information humans see. This has implications for education, research, and public knowledge production, where exposure to ideas may become mediated by algorithmic systems rather than active inquiry.

Frontier AI systems can already generate policy proposals, legal arguments, financial frameworks, and even religious narratives.
Yet they do so without experience, embodiment, or moral accountability. They do not bear consequences, exercise judgment, or understand context in human terms. Scaling such systems into decision-making environments without sufficient human oversight risks gradually delegating responsibility to entities that optimize patterns rather than values.
The central challenge of the next decades will therefore not only be technical, but epistemic: working on that fluency does not replace reasoning, that coherence is not mistaken for truth, and that human judgment remains central in how knowledge and decisions are produced.

Researchers must consider transparency, accountability, and human agency. AI systems influence reasoning and decision-making, so it is important to document how outputs are generated and how humans interact with them. Ethical use also includes protecting participant data, avoiding automation bias, and securing that AI supports rather than replaces human creativity and judgment.

AI will make knowledge production more collaborative, iterative, and rapid. Researchers will increasingly work with AI to explore ideas, test alternatives, and synthesize literature.

At the same time, this may lead to convergence toward similar outputs. I believe maintaining diversity of thought and originality will become one of our central challenges.

AI is more likely to transform the balance of human expertise rather than replace it.

Foundational knowledge, including memorization, will remain essential because it builds the mental structures that allow experts to interpret information, detect errors, and challenge AI-generated outputs. Without internalized knowledge, it becomes difficult to distinguish plausible responses from correct ones.

At the same time, expertise will increasingly involve working effectively with AI systems. Experts will need to evaluate outputs, compare alternatives, identify biases, and decide when AI assistance is appropriate. In this sense, expertise will combine strong domain knowledge with the ability to critically engage with AI.

So for now, I believe that rather than replacing human expertise, AI may actually make deep expertise more important. As AI generates fluent and convincing responses, those with strong knowledge and critical thinking skills will be better positioned to question, refine, and improve them. The risk is not that expertise disappears, but that superficial reliance on AI is mistaken for expertise.

Research on human-AI collaboration, creativity, and cognitive effects of AI is particularly promising. Studies now examine how AI affects exploration depth, solution diversity, and learning. There is also growing work on process-aware AI interventions, such as delayed prompting and reflection cues, designed to preserve human creativity while leveraging AI efficiency.

If I had to recommend one key idea, it would be this: AI is not intelligence; it is statistics at scale. It predicts likely answers based on past data. This makes it powerful for efficiency but also prone to convergence toward average solutions.

These systems are extraordinarily fluent, and because human thinking is closely tied to language, this fluency can easily be mistaken for understanding. But coherence is not correctness, and confidence is not knowledge.

Human creativity, judgment, and exploration remain essential to go beyond what AI suggests.