In the late part of 2024 and much of 2025, the AI scene was dominated by the Chatbot Era. Everything involved text — long prompts, copying answers, repeating things over and over. Phones became tools to translate between our world and systems that couldn’t see it. Using AI often felt like giving directions to someone with their eyes closed. It was frustrating because these models were good with words but struggled with real-world understanding. Describing a photo or a screen took more effort than it should have, turning simple tasks into lengthy explanations. AI was smart with language but lacked perception, which limited what it could do.

The Shift from Language-Centered AI to Action-Oriented Systems

The real breakthrough happened when AI moved away from just focusing on language. Instead, systems started to see and interact with the world around them. No longer just telling you what something is or how to do something, AI began acting on your behalf. It could click, adjust, organize, and respond based on what it actually saw, not just what you described. This change means AI can now perform tasks that require perception, turning it into a more capable partner for real-world actions. This shift is crucial because it bridges the gap between understanding and doing, making AI more useful in everyday life.

Moving beyond text input opens many new possibilities. Instead of explaining everything in detail, users can simply show or point to what they mean. AI systems can interpret visual data and act accordingly. This new approach transforms AI from a passive assistant into an active participant in tasks that involve physical interaction or complex decision-making.

The New Wave of Intelligent Systems: Seeing and Acting

Before this change, models like ChatGPT were limited to recognizing and explaining images or sounds. They could describe what they saw but couldn’t do anything with that information. It was like having a very smart witness who couldn’t touch or change anything. The real progress came with systems designed for reasoning and agency, such as ChatGPT agents and GPT-5. These systems are built to understand problems deeply and take action without waiting for step-by-step instructions. For example, if they see a broken car part, they understand what’s wrong, what tools are needed, and what steps to take to fix it.

Another major development is Google’s Gemini system, which excels at understanding context over time. It remembers what you looked at weeks ago and can bring that information back when needed, without you asking for it explicitly. Alongside these, physical intelligence models, often called “pi,” are trained on data from robots. These models understand depth, weight, and balance, enabling robots and AI systems to interact more naturally with physical objects. This new wave of AI is not just about recognizing things but about understanding and acting based on what it perceives in the real world.

All these advancements point toward an AI future where seeing and acting go hand in hand. Instead of just processing language, AI will be able to perceive its environment and respond intelligently. This makes AI more versatile and practical, opening doors to new applications in automation, robotics, and everyday life. The ability to see the world as humans do is key to creating truly helpful and autonomous AI systems.

Inspired by

• https://justainews.com/blog/beyond-the-chatbot-why-the-future-of-ai-needs-to-see-what-you-see/