The story of artificial intelligence (AI) is one of steady progression—from simple rule-based automation in the 1950s to today’s generative models that craft text, images, and full motion video. What began as symbolic “if-then” logic evolved into machine learning, and then into the deep neural networks and large language/video models of the 2020s and beyond. At each stage, capabilities expanded and applications broadened: from chess-playing programs, to speech recognition, to self-driving cars, to creative generation of multimedia content.
In the early years, AI was largely confined to well-defined tasks: playing board games, solving mathematical problems, recognizing patterns in data. With the advent of deep learning in the 2010s, thanks to architectures like convolutional neural networks and large labelled datasets, AI began outperforming humans on image classification, natural language processing and audio tasks. The 2020s ushered in a new phase of “generative AI”—models that don’t just react, but create: text (e.g., chatbots), imagery (AI-art), and eventually motion video.
Into this landscape entered the now-famous viral clip of Will Smith eating spaghetti. This video—generated by AI in early 2023—depicted Smith, quite unrealistically, consuming a plate of spaghetti with odd facial expressions and uncanny movements. It spread across Reddit and social media as a humorous yet unsettling demonstration of how far (and how far from perfect) AI video generation still was. The clip quickly became an informal benchmark in the AI community: if your model could render “Will Smith eating spaghetti” convincingly, you were making real progress.
Fast-forward to 2025, and things look very different. Advanced video generation tools such as Veo 3 (from Google) and its competitors are able to simulate human actions, lip-sync, audio cues and realistic motion far more convincingly. A new “Will Smith eating spaghetti” version by Veo 3 went viral not because of its oddness but because of how close it came to looking real—and yet still gave away subtle signs of synthetic origin (strange sound textures, slightly off eye-movement). What this shows is the evolution: from crude AI video artefacts to high-fidelity generative media.
But the symbolism runs deeper. The spaghetti video, first in 2023 and then the 2025 upgrade, does not merely entertain—it reflects larger shifts in AI’s role. Where once AI was confined to narrow tasks, today it’s entering creative domains, narrative production, media manipulation and synthetic reality. This raises both opportunity and concern. On the one hand, creative industries can use AI to generate content faster, personalize experiences and expand storytelling possibilities. On the other hand, the sharper the synthetic realism becomes, the harder it will be to distinguish what’s “real” from what’s generated—deepfakes, misinformation, manipulated media become more plausible.
Thus, the evolution of AI is not only technical but cultural. The spaghetti video is a micro-example of that: what began as meme-fodder now becomes professional benchmark. It invites reflection on authenticity, media trust, and the ethics of synthetic humans. As models improve, so too must our ability to validate, authenticate and understand the provenance of media content.
Looking ahead, the next frontier may lie in multimodal agents that combine vision, audio, text, sensor data and real-time interaction—AI that doesn’t just generate but interacts, reasons and adapts in dynamic environments. The progression from the early “Will Smith eating spaghetti” meme to a 2025 polished AI version thus maps a broader arc: from novelty to mainstream, from artefact to expectation.
In summary, AI’s evolution—from simple automata to generative agents—can be glimpsed through cultural icons such as the Will Smith spaghetti clip. What began as a bizarre curiosity is now a benchmark of realism and a symbol of AI’s power—and its challenges. As we move into 2025 and beyond, the spaghetti test reminds us: we’re no longer enchanted by the glitchy fake—we’re asking whether we can even tell what’s real anymore.