TOKYO - A video archive from a recent AI conference has been released, offering a look into the evolving landscape of artificial intelligence through candid remarks from leading engineers and entrepreneurs.
The event, tied to the Horiemon AI initiative, drew strong interest from both professionals and the general public, reflecting the growing attention surrounding AI development in Japan. One of the featured speakers, engineer Ano, explained that while he had long worked in fields like natural language processing and neural networks, he only began identifying himself as an “AI engineer” around the time of the Tochichisen campaign. He acknowledged that the term now serves as a kind of marketing label and encouraged others in the industry to adopt it, especially given the accelerating visibility and relevance of AI in daily life.
The discussion centered on the dizzying pace of innovation, with panelists highlighting how global giants like OpenAI and Anthropic are locked in a high-stakes race supported by massive funding from governments and large investors. Despite these resources, open-source models are catching up quickly. What was once thought to be a one-year gap between closed and open developments has narrowed to as little as five months, shrinking further as AI systems begin updating and training themselves. This collapse in the development cycle has raised doubts about the long-term value of expensive proprietary efforts, as similar tools emerge quickly at far lower cost through open collaboration.
Speakers also addressed the growing legal and ethical tensions surrounding generative AI. Although companies like OpenAI initially adhered strictly to compliance and copyright standards, new tools now allow for the creation of images in the style of well-known studios such as Ghibli, blurring the boundaries of artistic ownership. While some developers attempt to observe legal norms, others have adopted an aggressive posture—releasing tools that mimic protected aesthetics without restraint. Questions were raised about whether styles or “artistic flavors” can or should be protected, with several participants noting that if such protections were formalized, they could end up restricting not only machines but also human creators working in similar styles.
These concerns extended beyond visual art to include food and music. The discussion referenced well-known chefs who protested imitation of their signature dishes online, despite the fact that recipes themselves are not protected under copyright law. Similarly, an Indian programmer reportedly attempted to register every conceivable short melody as open source, effectively blocking others from copyrighting them in the future. Other developers have published libraries of common musical progressions, challenging traditional licensing models and contributing to a broader redefinition of creative ownership in the age of AI.
The conversation also touched on the geopolitical dimensions of AI development. Panelists speculated that the industry could eventually consolidate into a battle between one U.S. and one Chinese firm, especially as private capital becomes exhausted and national security concerns drive state investment. In this scenario, AI tools could become commoditized worldwide, with Japan and other nations relying on paid access to foreign platforms. Meanwhile, China’s open-source strategy may provide Japanese firms with near-term benefits, but long-term consequences remain uncertain. Some expressed concern that AI models built with Chinese ideological filters could proliferate across the Global South, subtly reshaping global narratives by omitting or altering politically sensitive content such as the Tiananmen Square protests.
Yet despite these risks, speakers noted that language and behavior models can still be fine-tuned. Japanese companies like CyberAgent were able to adjust publicly released Chinese models within weeks, enabling them to answer questions that had been deliberately censored. The panel concluded that while current AI tools are largely reactive—able to recall or suppress information—they will eventually evolve into systems capable of making decisions and performing tasks autonomously. As AI transitions from knowledge to labor, its adoption will depend largely on cost. Panelists pointed to labor-intensive sectors like food factories, where automation remains limited not due to technology, but because human labor is still cheaper. Once that cost balance tips, they argued, the shift to AI-powered robots in areas such as caregiving and household tasks will accelerate rapidly.
Source: 堀江貴文 ホリエモン
