By Lv Naiki
It's been over a year since OpenAI released ChatGPT, but discussions surrounding it are still vibrant.
"From 0 to 1" and "From 1 to 100" are common expressions seen in the media in recent years. The former refers to original innovation, while the latter signifies the realization and widespread application of that innovation across various domains. But what exactly is "1"? The process leading up to and following the release of ChatGPT by OpenAI revolves around the journey "from 0 to 100," with the pivotal point being "1." This process offers three perspectives on understanding "1":
Looking Forward: After OpenAI released GPT-3 in 2020, originally planning to continue developing GPT-4, concerns arose about rival companies potentially releasing AI chatbots before GPT-4. Consequently, OpenAI established a laboratory to collect feedback and improve new models by enhancing GPT-3. Thirteen days later, ChatGPT was born.
Upon completion, OpenAI invited testers to evaluate ChatGPT, but the feedback was disappointing because users didn't know what to chat about with the chatbot. OpenAI attempted to develop specialized chatbots for specific fields but lacked suitable data for training.
Taking a leap of faith, OpenAI publicly released ChatGPT, allowing users to use it freely and let it evolve on its own. Co-founder Greg Brockman admitted, "I didn't know if it would succeed." Subsequently, the sudden popularity of ChatGPT caught the company off guard.
This process illustrates the challenge of aligning supply with demand and a sense of uncertainty. Developers struggled to precisely anticipate user needs or faced deficiencies in data and technology, while users were uncertain about the relevance of cutting-edge technology to their own needs or even what their needs were. The result was a tentative alignment between supply and demand.
Rather than careful planning, developers opted for a "wait and see" approach, while users, especially the numerous "off-plan" users, "suddenly witnessed the ability of AI up close." The explosion in growth over two months demonstrated a strong affinity between supply and demand. ChatGPT isn't just "generative AI" but also "generative" innovation. Generative innovation requires a societal environment with a high tolerance for uncertainty.
In the journey "from 0 to 1," ChatGPT's infancy was marked by chaos, akin to the source of a great river, where streams meander almost aimlessly. Once a particular "stream" gains prominence and vaguely connects with the ocean (societal demand), resources converge onto this stream, turning it into a mainstream. For example, Microsoft's release of AI-enhanced "Bing" saw a tenfold increase in global downloads. Of course, there are instances where reaching the ocean remains uncertain, such as the former metaverse.
Google is a peculiar example in some sense. Google already had a product similar to ChatGPT, but its AI ethics experts advocated for cautious use of text generation technology, leading Google to postpone the release of related products out of ethical concerns.
After OpenAI's pioneering release of ChatGPT, some Google executives were incensed, AI ethics experts were fired, and the former ethical taboos became irrelevant. Under competitive pressure, Google hastily launched Bard on February 7, 2023, only to discover flaws that led to a 7.4% drop in stock price the next day. However, Google now seems bolder and more open, aiming to catch up in products and their ecosystem.
Timing is crucial in navigating the competitive landscape, recognizing and controlling the "technological rhythm" and "business rhythm" duet, with the theme being "product/market fit." Being ahead in business rhythm while lagging in technological rhythm can lead to troughs or even death valleys along the Gartner curve; conversely, missing opportunities allows others to capture the market. Moore's Law exemplifies a perfect duet. With the first step of ChatGPT, OpenAI's latest release, Sora, accurately hits the rhythm's mark.
One of ChatGPT's successful innovative footholds is its universality, making it suitable for various industries and significantly lowering the usage threshold with simpler natural language, making it accessible to everyone. Another foothold is "generation," empowering users to greatly enhance work efficiency. Lowering the threshold expands the breadth of demand, while improving efficiency satisfies the depth of demand.
In summary, looking back "from 0 to 1," "1" represents a new technology's first alignment with demand or awakening a universal human need.
Looking Ahead: Without the subsequent "From 1 to 100," "1" would be like a castle in the air, only admired by oneself, or even stillborn.
While producing the second or third internal combustion engine after the first one is undoubtedly important, more crucial is applying the internal combustion engine to various fields, such as transforming horse-drawn carriages into automobiles.
Once the breakthrough from "from 0 to 1" to "from 1 to 100" is made and further aligned with demand, it may lead the way, even in every step. Extending and expanding ChatGPT to meet demand refers to pushing from a universal platform to various application scenarios. In just two months after its launch, users tested and explored ChatGPT's abilities from different angles, including coding, homework, essays, speeches, and generating various texts such as event planning, advertising copy, movie scripts, or providing advice on home decoration, programming debugging, and life planning. More importantly, it extends from a universal platform to various vertical industries. This is the transition from "from 0 to 1" to "from 1 to 100." "From 1 to 100" defends and nurtures "1," forming new industries and fostering new "from 0 to 1" innovations. Since the advent of ChatGPT, a multitude of similar products have sprung up like mushrooms in a short period. Nearly all major tech companies have to some extent engaged in the development of large models. As of October 2023, there were 238 AI large models domestically, a significant increase from 78 in May. Professionals compared answers generated by five different large language models to the same question, seeking biases in categories such as race, body shape, and political ideology. This was done to help developers assess the potential harm of large language models and make informed decisions.
Competition on the demand side involves both 2C (consumer-facing) and 2B (enterprise-facing) sectors. Initially focused on 2C, the need to delve deeper into a broader demand side and make more profound changes to productivity inevitably leads to involvement in 2B, transitioning from large models' general platforms to various vertical industries. Companies across industries select superior large models and choose secondary large models closely tailored to their own scenarios within the companies that own the large models, thus gradually influencing a wider range of individuals. Issues identified and data obtained in both 2C and 2B sectors can feed back into large models.
At the intersection of platforms and verticals, there is an intense battle for dominance. Following the ousting of founder and CEO Ottman by the board ("palace coup"), OpenAI's over 100 clients have approached their competitor Antropic expressing their willingness to switch. An insider revealed, "The company (OpenAI) can no longer avoid commercialization under the pretext of idealistic research. Now, customer demands are pressing." It's evident that startups that emerge successfully, if they fail to take the step from "0 to 100" thereafter, may lose their position as "1" in facilitating others' journey from "0 to 100."
Generally, "0 to 1" is initiated by a single entity, whereas "1 to 100" is propelled by multiple entities including the initiating one, pushing towards completion. With the emergence of ChatGPT, OpenAI is the initiator of "0 to 1", but mishandling the "palace coup" may jeopardize its position as the main entity in "1 to 100." While the industrial significance of "1 to 100" may be affected, its temporal and spatial influence is limited, and other entities will take its place.
In summary, extending and expanding from "1 to 100," as well as strengthening the coupling between "1" and "100," in turn reinforces the significance of "1." "1" becomes a critical point and turning point in the process from "0 to 100."
On one end lies the "primordial time" near "0," with a broad vision and the burden of humanity's ideals and future. On the other end, the closer to "100," the deeper one delves into various scenes, satisfying the diverse current needs of various people. The investment-output ratio of suppliers (including capital) and the functional price ratio of demanders become dominant or even the sole options, with a growing commercial atmosphere, drifting further away from the original intention. The right to speak gradually shifts from the 2% elite to the 98% masses, from lofty ideals in the clouds to mundane concerns of everyday life. For the 2% elite, perhaps the philosopher Gödel's words can be quoted: "The meaning of the world lies in the fact that it goes against the will, and in the effort to overcome the separation of things and desires."
The future direction of humanity is jointly determined by the grand ideals of the 2% elite in the primordial time and the worldly affairs of the 98% masses. The fact that the latter slightly outweighs the former determines the fate of humanity.
"1" is no longer as important in the view from "0 to 100." (The author is a professor at Southeast University.)
"Chinese Science Newspaper" (April 3, 2024, 3rd edition, Field)
