Jie Tang, representing Zai.org, identified long-horizon tasks, autonomous agent systems, and self-evolving capabilities as critical for achieving Artificial Superintelligence (ASI). The organization aims for full self-training and autonomous smart agents capable of executing long-term tasks to continuously improve models. This strategy, dubbed "Touch High," targets Artificial General Intelligence (AGI) within the next two years.
The focus on autonomous code writing and self-training agents suggests a future where AI systems can independently drive their own development and model improvement.