How will multimodal AI change interaction?

Watching all this AI news today, I suddenly realized an interesting phenomenon: the way we talk to machines is qualitatively changing. Remember the awkward experience of talking to a voice assistant for the first time? Back then, AI was more like a step-by-step question-answering machine, whereas today's multimodal AIs are already actively understanding, and even anticipating, our needs. For example, Google is testing the Agentspace voice conversation feature, so one can't help but imagine whether in the future it is really possible to discuss work programs with AI like a colleague.

At the heart of this change is the fact that AI is no longer limited to a single mode of information processing. Like the MiroThinker model shows the ability of "deep interaction", it can simultaneously call the search engine, code execution environment and other tools, in the context of 256K length to maintain coherent thinking. This reminds me of yesterday's debugging code experience, if there was an AI assistant that could understand the logic of the code and search for solutions to the errors, I might not have stayed up until three in the morning.

From command execution to active collaboration

What's fascinating about multimodal AI is that it's starting to break down the mechanical interaction of "one question, one answer". In research, for example, the ability of AI scientists like Kosmos to autonomously read thousands of papers, run data analyses, and even formulate new hypotheses has completely changed the relationship between researchers and the tools of knowledge. Imagine, when AI can simultaneously process text, image, voice information, but also call a variety of professional tools, our interaction with technology from "order" to "co-creation".

But this change also brings new challenges. Like that controversial AI-generated anime video, how do we define the boundaries of human-computer collaboration as AI becomes more and more creative? Should we treat it as an advanced tool, or a partner with some autonomy? This question may be more worth pondering than the technology itself.

A silent revolution in interaction

Interestingly, the most profound changes often happen without realizing it. When Grok can generate near-perfect anime videos in seconds, and when Gemini Enterprise can coordinate multiple intelligences to complete complex workflows, the way we interact with AI is actually being reconfigured. This reconfiguration is not just technical, it's cognitive - we're starting to get used to working with a digital partner that can see, hear, speak, and think.

To be honest, I sometimes wonder if this kind of all-encompassing human-computer interaction will make us lazier. But on second thought, when repetitive tasks are handed over to AI, we might be able to focus more on the things that really need human qualities: creativity, emotion, and strategic thinking. Just as the camera didn't eliminate painting, but rather spawned a new art form, multimodal AI may lead to unexpected modes of collaboration.

Looking at the AI data center that GMI is building in Taiwan, which is capable of processing nearly 2 million tokens per second, I suddenly realized: the infrastructure for this interaction revolution is already in place. What's going to happen next may be faster and deeper than we think. I just wonder if we're ready for when AI truly becomes a ubiquitous partner in our work life.