Unveiling the True Innovations of TikTok and the Pitfalls of Starting a Business Around GPT-3

Unveiling the True Innovations of TikTok and the Pitfalls of Starting a Business Around GPT-3

In the world of AI engineering, there is a well-known challenge called the "cold start problem," which refers to the difficulty of determining who to recommend to new users (TikTokers) when there is limited data available. However, TikTok has fundamentally changed this dynamic. Even if a new user posts uninteresting videos, hundreds of people will still watch them. This is because all TikTok users share the burden of watching new videos. Only TikTokers who have already gained a certain number of fans are allowed to post 1-minute videos. ByteDance, the company behind TikTok, has cleverly solved this problem using a human-in-the-loop system that capitalizes on the short nature of 15-second videos.

The question then arises: is the pursuit of creating content or some form of output inherently related to personal growth, or is it merely a means to expand one's influence? The era of striving to expand influence is gradually coming to an end. Regardless of whether the content is created by an unknown individual, as long as it holds value, it is increasingly becoming the norm for it to reach a wide audience. So, how should non-influencers navigate the future? Time is limited in life, and it is crucial to always anticipate what is important in the next era and grab hold of what truly matters. We are entering an era where the value itself, rather than evaluations or credibility, is of utmost importance.

Now, let's shift our focus to the pitfalls of starting a business around GPT-3, as discussed in an article by Allen Cheng. The ease of creating a good-enough app using GPT-3 lowers the barriers to entry significantly. If GPT-3 is easily adopted by incumbents, the competition will shift away from technology and towards other dimensions such as marketing and distribution. As a result, the profits will primarily benefit the algorithm owners, such as OpenAI and marketing platforms like Google and Facebook. Unfortunately, most products built on GPT-3 will be indistinguishable from one another and lack a meaningful edge.

When Google first emerged, its search results were significantly better than those of other search engines, leaving little reason to use any competitor. Similarly, the better GPT-3 works out of the box, the harder it will be for any company to establish meaningful differentiation. Early demos of GPT-3 already show impressive results. Prior to GPT-3, building an AI system that achieved a level comparable to human standards was extremely challenging and required substantial investment. However, with GPT-3, there are now more competitors who can offer a "good enough" product, compressing the range of competition into a narrower band.

One of the major drawbacks of building on GPT-3 is that any proprietary progress made using GPT-3 will likely be rendered obsolete by subsequent iterations such as GPT-4 and GPT-5. Additionally, the ability to improve beyond the baseline performance of GPT-3 is limited since companies do not own the core technology behind it. Therefore, it becomes increasingly difficult for companies to differentiate themselves.

Furthermore, companies built on GPT-3 face financial challenges. While the beta API is free, access to the API will eventually come at a cost, with each API call requiring payment. As user numbers and usage increase, the expenses incurred by utilizing GPT-3 will also rise. This situation resembles the model of Spotify, where the more users engage with the product, the more they have to pay.

Additionally, the advantage of having more data in the AI field is questionable. If marginal improvements from additional data plateau quickly, then companies with more data may not necessarily have an advantage. Moreover, GPT-3's limited "working memory" further restricts its potential for network building. It forgets characters and events that occurred just a few minutes earlier, making it challenging to create a seamless user experience.

In Clayton Christensen's terminology, GPT-3 can be seen as a sustaining innovation that benefits existing players rather than a disruptive innovation that benefits new entrants. Sustaining innovations provide improved performance or cost savings to existing products, primarily benefiting market leaders who use these innovations to enhance their offerings. On the other hand, disruptive innovations initially appear inferior to existing products and are often dismissed as mere toys. However, it is important to note that within the same industry, an innovation can be disruptive to some incumbents and sustaining to others.

In conclusion, businesses should carefully consider the limitations and challenges associated with building around GPT-3. While it may seem like a promising opportunity, the lack of differentiation, the potential for rapid obsolescence, and the financial implications make it a risky venture. Instead, businesses should focus on enhancing user experience, product design, and customer support to create value that goes beyond the capabilities of GPT-3.

Actionable Advice:

1. Look beyond the technology: Instead of solely relying on the capabilities of a particular AI system like GPT-3, invest in areas such as user experience, product design, and customer support to create a unique value proposition.
2. Prioritize meaningful differentiation: Understand that building a business solely around a widely accessible technology may not provide a long-term competitive advantage. Seek ways to offer something truly unique and valuable to your target audience.
3. Embrace the era of value: Instead of chasing influence or popularity, focus on creating content or products that hold intrinsic value. As the importance of value itself continues to grow, prioritize delivering meaningful experiences and solutions to your customers.