In a significant development shaking up the global AI landscape, Baidu’s founder has openly criticized the growing obsession with text-only AI models, stressing the need for a broader, multimodal approach. The remarks come at a time when artificial intelligence is evolving rapidly and expectations are higher than ever.
Speaking at a recent technology summit, the Baidu founder emphasized that relying solely on text-based large language models (LLMs) like DeepSeek AI limits the full potential of what artificial intelligence can achieve. According to him, true intelligence should be able to seamlessly integrate text, images, audio, and even video — enabling machines to understand and interact with the world in a way that's far closer to human cognition.
"Focusing exclusively on text inputs is like trying to understand the world with one eye closed," he said, drawing applause from the audience filled with AI experts, entrepreneurs, and tech enthusiasts.
The Rise and Challenges of Text-Only Models
Text-based AI models, like OpenAI's GPT series and DeepSeek AI, have undoubtedly revolutionized fields from customer service to creative writing. Yet, critics are beginning to argue that they only scratch the surface of AI's true capabilities. Without the ability to process and reason across multiple modalities, these models remain fundamentally limited in real-world applications, especially where visual and audio understanding are crucial.
Why Multimodal AI is the Future
Multimodal AI combines various types of input — such as text, images, video, and speech — allowing machines to create a richer, more nuanced understanding of context. For instance, an AI doctor that can analyze patient records (text), X-ray images (visuals), and spoken patient concerns (audio) will be far superior to one that can only read reports.
Baidu’s own R&D efforts have been moving in this direction. They are investing heavily in creating models that not only chat but can also see, hear, and interact much like a human would. Multimodal models could redefine areas like self-driving cars, virtual assistants, medical diagnostics, and creative industries.
DeepSeek AI’s Situation
DeepSeek AI, once considered a formidable rival in the text-based LLM market, is now facing declining demand. As businesses and researchers pivot towards more versatile and context-aware models, DeepSeek’s heavy reliance on text-only architecture has started to look like a major shortcoming.
Baidu’s founder’s comments seemed aimed not just at industry peers but also subtly at DeepSeek’s strategic direction. Insiders suggest that DeepSeek is now scrambling to integrate multimodal capabilities into their systems to stay competitive.
Industry Reactions
The tech community has largely echoed Baidu's views. Prominent AI researchers, including some from Google DeepMind and Meta AI, have shared studies showing that multimodal systems outperform unimodal (text-only) models on tasks requiring complex reasoning.
Elon Musk also chimed in on social media, noting, "If AI is to truly reflect human intelligence, it must perceive the world in more than just words."
Other companies like OpenAI, Anthropic, and Mistral are already pivoting towards multimodal platforms, rolling out updates that allow their AI models to analyze images, interpret sounds, and even watch videos.
What It Means for the Future
The push towards multimodal AI marks a paradigm shift. In the next few years, we can expect digital assistants that not only understand your words but recognize emotions in your voice, interpret gestures, and even assist visually.
Education, healthcare, entertainment, and autonomous driving are among the industries that stand to benefit the most. Imagine AI tutors who can read student expressions to adapt teaching styles or medical bots that combine MRI scans and patient interviews for better diagnoses.
Challenges Ahead
While the future looks promising, building truly multimodal AI is fraught with challenges. Integrating different types of data requires more complex architectures, massive training data, and higher computational costs. Ethical concerns also intensify, as collecting visual and audio data raises serious privacy issues.
Nevertheless, Baidu’s message is clear: the age of single-skill AI is coming to an end. The future belongs to systems that can see, hear, speak, and reason holistically.
Baidu’s founder has thrown down the gauntlet, challenging the tech world to move beyond text-only AI and embrace a truly multimodal future. As companies scramble to adapt, one thing is certain — the next generation of artificial intelligence will be smarter, more human-like, and infinitely more capable.