/

This is Just the Beginning.

The year 2024, indelibly etched in human history, marks the dawn of the first year of AI application implementation. A plethora of AI applications are now in full swing, permeating every facet of our daily lives. For instance, generative AI has made groundbreaking advancements in content creation, spanning text, images, audio, and video. Additionally, AI assistants are becoming increasingly intelligent, supporting various industries effectively.

Simultaneously, these applications harness vast amounts of data, collected and analyzed for further training, ultimately enhancing our ability to create more efficient and precise products.

The AI revolution centered around large models mirrors the transformative impact of the industrial revolution, profoundly altering human society. In the cyclical interplay between humans and AI, their relationship is both intimate and distant. Nonetheless, it is undeniable that AI is seamlessly integrating into our social structure, becoming an indispensable part of it.

By 2025, we anticipate an explosion in the deployment of large models, with their commercial value surpassing imagination. The future trajectory of human-AI relations will significantly shape our destiny.

This year, "Business Show" has been meticulously documenting this era of transformation and the strides in AI development. As the year draws to a close, let's reflect and deliberate on how to coexist with AI and chart humanity's course forward. We'll also delve into product applications, aiming to provide valuable insights. Join us as we explore the current state of AI development and strategize on how to embrace, coexist, and evolve alongside AI.

First, let's revisit the remarkable large-model products that have emerged this year. Prominently, ChatGPT stands out. Since its inception, every iteration has garnered global attention and spurred AI research enthusiasm. This year, ChatGPT's GPT-4 model achieved the integration of speech and vision, adept at accepting any combination of text, audio, and image inputs.

However, even top models like ChatGPT encounter ambiguity, misunderstandings, and inaccurate responses. Expectations for its and other large models' evolution center on producing text that mirrors human thinking and expression in terms of logic and coherence.

Data indicates that ChatGPT's 4.0 version has improved dialogue comprehension accuracy by approximately 87% compared to the 3.5 version, yet there's still room for growth.

Next, we have Sora, which amazed the world with its ability to directly convert text into video upon its February 16th release. Naturally, expectations extend beyond this, hoping for higher-quality video generation akin to or even surpassing professional production levels, with enhanced integration of audio, text, and other modalities.

In China, notable models include Doubao, Kimi, ERNIE Bot, ERNIE 3.0 Zeus, and Qwen, among others, which we won't delve into here. Interested readers can explore previous "One Glance at the Macro View" videos for more details. The proliferation of these product applications is intertwined with significant technological breakthroughs by leading companies.

For instance, OpenAI introduced the o1 inference model at the year's outset and the o3 model in December, enhancing chatbots and writing assistants' intelligence and response accuracy. Google's Gemini 2.0 boasts double the speed of its predecessor.

Notably, NVIDIA's GPUs serve as super engines for AI, accelerating AI training and execution. Faster hardware support fosters better AI applications. Moreover, a pivotal trend in 2024 is open-source AI, diversifying AI development and offering more options for developers and enterprises.

After examining key large-model products and companies, let's address the unresolved issues currently facing AI, primarily in three areas: technological application, commercialization, and safety/ethics.

At the technological application level, AI technology is still in its nascent stages. Both fundamental theory and model performance have notable limitations. For example, language models may err in complex semantic understanding, logical reasoning, and long-term contextual dependence, often termed "hallucinations." Furthermore, training and inference processes for large models consume vast computational resources and energy, increasing costs.

Regarding commercialization, due to product and technological constraints, AI commercialization hasn't witnessed a large-scale explosion. During "Business Show" visits to enterprises, we found that some decision-makers fear inaccurate AI advice in critical business decisions, potentially causing losses. Additionally, AI technology's adaptability in complex business scenarios is lacking, making ROI determination challenging, especially in traditional manufacturing where integrating AI into production processes and quality control poses numerous technical and managerial hurdles.

In our research and visits over the past year, most enterprises found it difficult to accurately measure the value AI products or services bring, complicating pricing strategies.

Concerning safety and ethics, AI systems collect and utilize vast amounts of user data, and attack methods are becoming increasingly AI-driven, posing heightened risks. At the ethical level, AI decision-making processes and behavioral outcomes spark numerous discussions. For instance, in an unavoidable collision, should an autonomous vehicle prioritize protecting passengers or pedestrians? Moreover, AI-generated content may contain false, fraudulent, or misleading information.

Regarding the human-AI relationship, while some fear AI will replace human job opportunities, we should explore collaborative strategies, leveraging each other's strengths for human-AI complementarity. I view AI more as a tool than a threat. We should harness it, master collaboration skills, and let AI assist with preliminary tasks while we focus on in-depth analysis, strategic decision-making, and more.

The AI era has arrived, yet it's merely the beginning. Besides embracing this era, we must recognize humans' unique skills that AI lacks.

For instance, complex emotional understanding and empathy. An AI psychological test or consultation feels impersonal. However, the emotional bond between a psychologist and client during face-to-face communication is irreplaceable.

Creativity also sets humans apart. Artists convey emotions through their work, resonating with shared human experiences. While AI can simulate emotional expression, it struggles to genuinely comprehend emotions' essence.

Moreover, humans possess critical thinking and value judgment abilities. In an information-overloaded era, humans discern and judge information. Ethical considerations and humanistic care in professions like investigative journalism and healthcare surpass AI's cold algorithms.

In summary, no matter how advanced AI becomes, it will never master humanity's quest for meaning and inner exploration. "End"