"Various AI Models Should Be Utilized with Goal Orientation...Optimized Results Are Achievable" [AI WORLD 2025 Special Lecture]

[The Financial News] "Various artificial intelligence (AI) models should be used with a goal-oriented approach. By combining different AI models for a single task, it is possible to achieve more efficient and optimized results."
Nick Emmons, founder and CEO of Allora Labs, emphasized the need for this shift in AI utilization at 'AI WORLD 2025,' jointly organized by The Financial News and the Ministry of Science and ICT (MSIT) at Lotte Cinema in Songpa District, Seoul, on the 25th of last month. He stated, "AI models can be merged in various ways," and added, "Through this, AI should be leveraged as a resource for talent, computational power, elections, finance, and more."
Breaking Away from AI Monopoly...Synergy Through IntegrationOn this day, CEO Nick Emmons remarked, "The AI market itself is highly decentralized," and stressed, "We need to create AI that is more equitable and accessible." He argued that instead of centralizing currently isolated AI models, they should be further decentralized.
He noted, "Currently, a few companies dominate the AI market, monopolizing data, computing resources, and talent to build foundational models and core AI components." He continued, "As a result, many AI users are forced to rely on these models, which leads to the exclusion of numerous models, talent, and resources from the market." This, he pointed out, results in inefficiency and slows the pace of AI advancement.
He further explained, "Companies and governments should determine the appropriate use cases for AI and assign tasks to multiple AI models based on business objectives. The current structure, where consumers are limited to using only specific AI models, should be avoided." He emphasized that this approach would allow numerous AI models to generate synergistic effects toward a given goal.
In practice, the Allora Network aggregates the outputs of diverse AI models and synthesizes their conclusions to derive the most optimized result. CEO Nick Emmons shared, "In an actual experiment, five models participated in a single network, and the aggregated outcome was approximately four times more accurate than the best individual model." He explained, "This was possible because the network extracted the optimal parts from each model’s output and combined them contextually."
Expanding the AI Network Enhances Reasoning Power...Limitless ApplicationsHe predicted that developing useful models capable of generating synergy would enhance the overall efficiency of the AI market.
CEO Nick Emmons stated, "By utilizing decentralized systems, it is possible to optimize efficiently and enter the market. On a blockchain-based decentralized network, as long as an AI model is created and its application is set, value can be generated immediately, lowering entry barriers and enabling meaningful efficiency gains."
Allora Labs is also a pioneer in the distributed cloud market, having researched decentralized mechanisms for the past five years. To date, approximately 700 million inferences have been executed, with about 300,000 models participating.
CEO Nick Emmons explained, "This is a much larger scale than the current market dominated by a handful of models. The aggregated results provide performance superior to individual models." He added, "Expanding the network dramatically increases the number of inferences. The Allora ecosystem, through its applications and decentralized system, will continue to attract more participants, and the market structure will evolve more rapidly and efficiently."
This approach has also contributed to deriving practically useful conclusions. He said, "For example, at the end of last year, ahead of the United States presidential election, we used various predictive models to forecast the results, and the accuracy was very high." He continued, "By leveraging a variety of AI models, it is possible to combine their strengths and adapt to rapid environmental changes, making them applicable for election predictions and more." He emphasized, "This is an approach that can be expanded to virtually all areas where AI can be applied. It is an achievement that can be realized when we fundamentally rethink the structure of the AI market and aggregate models from around the world at the market level."
jiany@fnnews.com Yeon Ji-an, Park Seong-hyun Reporter