During the pandemic, scanning products achieved growth against the trend. IDC reports that the shipment volume of scanners in China increased by 9.7% year-on-year in 2021. Google's Tesseract OCR engine boasts an almost 100% recognition accuracy, greatly addressing data accuracy issues. Quark Scan King has launched a万能 scanning function that supports editing of text and table content, enhancing user experience. Quark Scan King addresses young users' pain points with automatic document format and layout recognition. Large model technology supports Quark Scan King's handwriting and formula recognition accuracy reaching 99%, leading the industry.

The enthusiasm for HBM continues to surge, with Samsung and SK Hynix increasing HBM production, leading to a significant rise in market attention. The release of NVIDIA's H200 has stimulated HBM demand, and both Samsung and SK Hynix have announced substantial expansion plans. HBM technology addresses the memory wall issue, meeting the demands for large-scale data storage and high computational needs. HBM has clear advantages as an AI accelerator card; however, more effort is needed for its adoption in consumer-grade products.

The resignation of OpenAI's founder as CEO has ignited an internal power struggle. AI entrepreneurs in both China and the U.S. are under pressure to become profitable, with significant differences in the investment markets. The era of large models presents new challenges for entrepreneurs, as the survival space for independent apps shrinks. Investors are focusing on different segments: application layers are gaining attention in China, while the investment market in the U.S. is booming. The differences in AI entrepreneurship between China and the U.S. stem from the maturity of large models and the activity level of the investment markets.

As the capabilities of large models improve, simple prompts may be automatically generated, lowering the design threshold. Prompt engineers will need to handle more complex scenarios and have an in-depth understanding of both the model and the context to design effective prompts. The development of multimodal interactions will raise new prompt requirements for processing combinations of prompts related to images, audio, etc. Professional prompt engineers will tackle more challenging tasks, further widening the gap. High-quality prompts remain important, and prompt engineers will continue to facilitate human-computer interaction.

People often overestimate the short-term impact of technology while underestimating its long-term contributions to social development. Huang Xuedong believes that in the commercialization of AI, there is a need to focus on the integration and interoperability capabilities of models. The future trend will center around large models with joint development of multimodal functionalities.