Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain
the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in
Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles
and JavaScript.
The intelligence illusion: why AI isn’t as smart as it is made out to be
Artificial-intelligence models will supposedly take over the world, but AI innovator Luc Julia tells Nature that they’re little more than glorified pocket calculators.
The AI Illusion: Why Machines Aren’t CreativeLuc Julia Wiley (2026)
French-American computer scientist Luc Julia has worked at the interface of artificial intelligence and consumer technologies for more than three decades. Currently chief scientific officer at the car maker Renault Group, he has previously worked at Samsung Electronics, Apple and Hewlett-Packard. He did early work on the natural-language-processing tools that underlie current generative AI models. In his book, The AI Illusion, translated from French, he argues that the hype and fear surrounding the intelligence and creative abilities of AI models are overblown.
What is the ‘AI Illusion’?
The term aims to address a fundamental misunderstanding about AI that has persisted for nearly 70 years, dating back to 1956 when AI research formally began. The term ‘intelligence’ is widely misunderstood, often leading people to anthropomorphize AI tools, attributing human-like qualities to machines. This illusion has been perpetuated by science fiction and media portrayals, which depict AI systems as potentially dangerous or capable of developing human-like emotions and decision-making skills.
In reality, the systems that we call AI are more about processing information than showing intelligence similar to human smartness. The illusion lies in our tendency to overestimate AI’s capabilities and potential threats, rather than understanding it as a collection of sophisticated but narrow tools designed for specific tasks.
Just as a magician uses sleight of hand to create the illusion of magic, the terminology around AI creates the illusion of human-like intelligence. This stems from the dual meaning of the word intelligence, which can refer to both information processing and cognitive smartness. The latter is often projected onto AI, leading to exaggerated expectations and fears. AI, in its current form, operates on algorithms and data, performing tasks with precision but lacking the consciousness and creativity that is inherent in human intelligence. This distinction is essential for understanding the true capabilities and limitations of AI.
Who is being deceived?
The general public, by the technology companies and organizations that benefit from the hype around AI. These companies are in a race to develop the technology and are incentivized to promote the idea of human-like artificial general intelligence to secure funding and market dominance. Members of the scientific community, particularly those who are not directly involved in the race for AI funding, acknowledge the reality that AI is a set of specialized tools, rather than a unified intelligent entity. This distinction is crucial, but it is blurred by commercial interests that amplify the illusion for monetary gain.
The narrative of AI as an impending replacement for human intelligence fuels both fascination and apprehension. It generates excitement and investment, driving technological advancements and economic growth. But it also creates unrealistic expectations and fears, influencing public perception and policy decisions. It is important to recognize that, although AI can augment human capabilities, it is not a sentient entity poised to overtake human roles. Understanding this dynamic is essential for fostering informed discussions about the role of AI in society.
What is intelligence in the context of AI?
Intelligence is a contentious term because it lacks a single, universally accepted definition. In the context of AI, ‘intelligence’ often refers to information processing rather than genuine cognitive ability. A calculator performs calculations faster than a human, which might seem intelligent, but it is merely executing predefined operations. Similarly, AI systems are designed to excel at specific tasks, outperforming humans in those areas, but they lack the creativity and adaptability that is inherent to human intelligence. Philosophers and psychologists offer various perspectives on intelligence, but AI, as it stands, does not have the innate creativity or consciousness that is associated with true intelligence.
The debate over AI’s intelligence highlights the complexity of defining intelligence itself. Human intelligence encompasses a range of cognitive abilities, including reasoning, problem solving and emotional understanding. AI operates in the confines of algorithms and the data on which it is trained, lacking the experiential learning and emotional depth of human cognition.
Yet AI systems are powerful?
Absolutely, AI systems are impressive in their designated functions. The power of AI lies in its ability to process vast amounts of data quickly and accurately. This capability has transformed industries such as health care, finance and transportation.
However, with great power comes great responsibility. The effectiveness of AI models depends on the quality of the data they are trained on and the context in which they are applied. Misuse or misunderstanding of AI can lead to errors, biases and ethical concerns, highlighting the importance of human oversight and regulation. The key to harnessing AI’s potential lies in recognizing it as a tool that complements human abilities rather than as a replacement for human intelligence.
What would a truly intelligent AI look like to you?
It would need to have a form of general intelligence similar to that of humans, capable of continuous and creative thought across various domains. This means the system would need to reflect and act on any subject, innovate spontaneously and create new concepts or solutions independently. AI currently lacks the biological and creative aspects of human intelligence. A system designed to play chess, for example, can defeat human grandmasters but is incapable of understanding or writing a poem.
Why do you argue that AI and machine learning are different?
This distinction is important for understanding the various components and capabilities of modern AI systems.
Enjoying our latest content?
Log in or create an account to continue
Access the most recent journalism from Nature's award-winning team
Explore the latest features & opinion covering groundbreaking research
Facts Only
Luc Julia is a French-American computer scientist and the chief scientific officer at Renault Group.
He previously worked at Samsung Electronics, Apple, and Hewlett-Packard.
Julia contributed to early natural-language-processing tools used in generative AI models.
His book, *The AI Illusion*, was published in 2026 and translated from French.
The book argues that AI's intelligence and creativity are overhyped.
AI research formally began in 1956.
Julia describes current AI systems as sophisticated but narrow tools for specific tasks.
He compares AI to glorified pocket calculators, emphasizing their lack of human-like cognition.
The term "intelligence" in AI refers to information processing, not cognitive smartness.
Tech companies benefit from promoting the illusion of human-like AI to secure funding.
AI systems excel in processing vast amounts of data quickly and accurately.
AI lacks consciousness, creativity, and the ability to innovate spontaneously.
Julia distinguishes between AI and machine learning as separate components of modern systems.
Executive Summary
Luc Julia, a French-American computer scientist and chief scientific officer at Renault Group, argues in his book *The AI Illusion* that the hype surrounding artificial intelligence is overblown. He contends that AI systems, despite their impressive capabilities, are fundamentally tools for processing information rather than exhibiting human-like intelligence or creativity. Julia, who has worked at major tech companies like Apple and Samsung, highlights that the term "intelligence" in AI is often misunderstood, leading to anthropomorphism and exaggerated fears or expectations. He emphasizes that AI lacks consciousness, adaptability, and the ability to innovate spontaneously, comparing current models to advanced calculators. While AI excels in specific tasks, such as data processing in healthcare or finance, its limitations stem from reliance on predefined algorithms and training data. Julia critiques the commercial incentives driving the AI hype, noting that tech companies benefit from promoting the illusion of human-like AI to secure funding and market dominance. The debate underscores the need for clearer distinctions between AI's capabilities and human intelligence, as well as the importance of ethical oversight to mitigate biases and misuse.
The discussion reflects broader tensions in public and scientific perceptions of AI. Some experts acknowledge AI as a set of specialized tools, while media and corporate narratives often amplify its potential as a transformative or even threatening force. Julia's perspective aligns with skeptics who caution against overestimating AI's autonomy, arguing that its power lies in augmentation rather than replacement of human cognition. The conversation also touches on philosophical questions about intelligence, creativity, and the ethical responsibilities tied to AI development. As AI continues to evolve, balancing its practical applications with realistic expectations remains a critical challenge for society.
Full Take
Luc Julia’s critique of AI hype offers a valuable counterbalance to the breathless narratives that dominate tech discourse. His argument—that AI is a tool, not a sentient entity—is a necessary corrective to the anthropomorphism that fuels both fear and fascination. The strongest version of his narrative is rooted in a clear-eyed assessment of AI’s capabilities: it processes data with remarkable efficiency but lacks the adaptability, creativity, and consciousness of human intelligence. This distinction is critical, especially as AI’s role in society expands. Julia’s background in industry lends credibility to his claims, and his call for realistic expectations aligns with a growing chorus of experts who warn against overpromising AI’s potential.
Yet the pattern scan reveals subtle tensions. The framing of AI as an "illusion" risks oversimplifying the debate, potentially dismissing legitimate concerns about AI’s societal impact—such as job displacement or algorithmic bias—as mere hype. While Julia rightly critiques the commercial incentives behind AI’s exaggerated portrayal, the narrative could benefit from acknowledging the nuanced ways AI *does* reshape human labor, decision-making, and even creativity (e.g., generative art). The focus on AI’s limitations, while important, might inadvertently downplay the second-order effects of its deployment, such as the concentration of power in tech giants or the erosion of human agency in automated systems.
Root cause: The paradigm driving this narrative is the long-standing tension between technological determinism and human-centric skepticism. Julia’s argument echoes historical debates about automation, where tools are either idolized as revolutionary or dismissed as overhyped. The unstated assumption here is that intelligence is a binary—either human-like or not—when in reality, AI’s impact lies in its ability to augment, distort, or replace specific human functions without replicating the whole. This echoes the Luddite debates of the 19th century, where the fear wasn’t about machines thinking but about their role in reshaping power structures.
Implications: For human agency, Julia’s perspective empowers individuals to see AI as a tool rather than a threat, but it also risks underestimating the systemic changes AI enables. Who benefits? Tech companies and investors who control AI’s development. Who bears the costs? Workers in automated industries, artists whose work is mimicked, and societies grappling with AI-driven misinformation. The second-order consequence is a potential complacency: if AI is "just a calculator," do we scrutinize its ethical deployment less rigorously?
Bridge questions: What if AI’s true danger isn’t sentience but its ability to scale human biases and errors at unprecedented speeds? How might Julia’s framework account for AI systems that, while not "intelligent," still disrupt social and economic systems? What would it take for AI to achieve even a fraction of human-like adaptability, and what safeguards would that require?
Counterstrike scan: A bad actor pushing this narrative might use it to dismiss legitimate concerns about AI regulation, framing critics as alarmists while tech companies operate with minimal oversight. However, Julia’s argument doesn’t align with this playbook; it’s a principled critique of hype, not a denial of AI’s risks. The content is clean—no signs of coordinated evasion or distortion.
Patterns detected: none
Sentinel — Human
The article shows strong signs of human authorship, with natural variability in sentence structure, a distinct voice, and context-specific arguments that are unlikely to be AI-generated.