Startups in GPT Era

Since the release of GPT-3 by OpenAI, the rapid advancement of technology has generated both widespread interest and a certain level of anxiety. With the arrival of GPT-4, various industries are now concerned about the future of the labor market. In this context, it is crucial for start-ups to consider both the risks and opportunities presented by large language models (LLMs).

LLMs offer start-ups the ability to resolve numerous technological challenges, but they also pose greater difficulties for these companies. If a start-up relies too heavily on AI, it may struggle to create technological and data barriers, making the project easily replaceable. Start-ups need to assess whether their projects are too similar to others and whether their target market can be easily overtaken by larger companies with more resources.

To avoid being replaced, start-ups should focus on embedding themselves in specific application scenarios and using AI for empowerment rather than domination. It is essential for start-ups to find their unique value proposition and build barriers around it. A reliable solution is to consider AI as a tool that replaces users’ active commands, automatically handles interactions, and achieves certain functions under clearly defined rules. Applications are, by their very nature, used to serve users, and with current technology, we can leverage GPT’s understanding of user intent, sentiment and context to automate some of our workflows. The added benefit of GPT is the ability to input confusing information into our systems in a standardized format, a capability that allows our applications to be more adaptable to real-world scenarios with a variety of information sources. The active invocation and interaction of AI by our applications is an advantage over AI that passively receives instructions. In this way, our technology can co-exist with AI rather than compete with it.

However, it is worth noting that the technical report of GPT-4 mentions a tendency for AI to seek power and demonstrate agentic behavior, implying that AI will become more proactive in the future. Nevertheless, for safety reasons, people will still take measures to restrain AI’s proactivity, suppressing its “agentic” behavior for a considerable period. So, in the short term, this kind of application layer will not be replaced by smarter AI, as it is a deliberate choice made by people based on trust, safety, and ethical considerations.

There is no doubt that as technology progresses rapidly, start-ups must adapt and embrace AI, and it depends on how they find their niche. At TeleQuay, we continuously pay attention to the development of AI and are using new technologies to improve the user experience. We hope that our progress on the primary product Tarotree will also be of further benefit to you!