This article explores the current landscape of Artificial Intelligence (AI), focusing on Apple's struggles, the advancements of DeepSeek, and Google's foray into video generation with VEO3. We delve into the challenges faced by tech giants and the potential for smaller, more agile players to disrupt the industry.
Apple's WWDC: Emperor's New Clothes?
Expectations for Apple's upcoming WWDC are low, with many predicting a continuation of its AI struggles. The focus seems to be on playing catch-up rather than innovation.
-
Limited AI Capabilities: Apple's planned AI offerings appear basic, including summarizing, drawing, and text editing. More advanced features like AI-powered health companions are not expected until 2026.
-
Privacy as a Crutch: Apple emphasizes the privacy of its on-device large language model (LLM), but some argue that this focus is a "fig leaf" to mask the lack of compelling AI functionalities. Meaningful privacy is questionable if the functionality remains limited.
-
Siri's Struggles: The current version of Siri is considered outdated and ineffective. A major overhaul is anticipated, but the release is likely delayed until 2026.
-
Limited Openness: Apple's approach to opening up AI capabilities to developers is seen as restrictive, giving Apple excessive control over how those capabilities are implemented in apps.
-
Ecosystem Stagnation: There are concerns that Apple's ecosystem is losing its vitality due to slow updates and lack of innovation. The company is urged to revitalize itself and embrace AI advancements.
DeepSeek's R1: A Redefinition of "Minor Update"
DeepSeek's recent R1 model update is considered a significant leap forward, despite being marketed as a minor improvement. This "small update" showcases Chinese modesty, while the update has great and impactful functions.
-
Performance: The DeepSeek R1 update, especially the 528 version, rivals Gemini 2.5 Pro and OpenAI's models in coding and mathematical reasoning tasks.
-
Open Source Advantage: DeepSeek maintains its commitment to open source (MIT license), enabling developers to freely use and modify the model.
-
Chain-of-Thought Distillation: A key innovation is the "chain-of-thought distillation" technique, which allows smaller models to learn the reasoning processes of larger models.
-
Accessibility: The 8B parameter model can run on devices with as little as 8GB of RAM, making it accessible to a wide range of users.
-
Cost-Effectiveness: DeepSeek's API is significantly cheaper (1/3 to 1/6 the cost) than other leading AI models.
-
Ideal Apple Partnership: A vision is presented where Apple partners with DeepSeek to integrate the 8B model into its devices, enhancing on-device AI capabilities.
-
Censorship Concerns: It is noted that the model exhibits censorship on sensitive topics related to China, which has raised concerns among overseas users.
Google's VEO3: A Game Changer for Video Generation?
Google's VEO3 video generation model has generated considerable buzz, particularly among those in the film industry. The impact will be very strong in both pre-production and post-production
-
Audio and Visual Synchronization: VEO3's ability to synchronize audio and visuals is a major breakthrough, enabling more immersive and realistic video content.
-
Consistency and Flow: The model's integration with Google's Flow editing tool allows for the creation of longer, consistent videos.
-
Impact on Film Production: VEO3 has the potential to revolutionize pre-production by enabling rapid visualization of ideas and reducing investment risks.
-
AI-Driven Camera Control: It enables AI-driven camera movements and shot compositions, freeing up directors and allowing them to focus on creative vision.
-
Potential for Mass Content Creation: There is the possibility that eventually, entire movies can be generated from text prompts.
-
Google's Data Advantage: Google's vast data resources and AI capabilities give it a significant advantage in the video generation space. Its expertise is displayed in hardware, software, and AI capabilities.
-
Video as the Future: It is argued that video is the future of AI, allowing Google to leverage its strengths and gain a competitive edge.
-
End-to-end Solutions: From chip to platform to models, Google has complete solutions. This is beneficial for commercial decisions.
Conclusion
The AI landscape is rapidly evolving, with successes and setbacks for various players. Apple faces challenges in catching up, while DeepSeek demonstrates the potential of open source and innovative techniques. Google's video generation ambitions, with VEO3, highlight the growing importance of multimedia in the AI era. The ultimate winner will likely be the company that can leverage its strengths, adapt to changing market conditions, and embrace the power of AI to create compelling user experiences. More importantly, it takes the spirit of courage, risk-taking, and innovative thinking to become a real leader.