news.volyx.in

Don't build AI products the way everyone else is doing it (builder.io)

565 points by tortilla · 1016 days ago · 248 comments on HN

Article summary

The article argues that building AI products by simply wrapping existing models, such as ChatGPT, is not a sustainable or differentiated approach. Instead, it suggests creating a custom toolchain by combining specialized models with normal code to build unique and valuable AI products. This approach allows for more control, customization, and cost-effectiveness. The article also highlights the importance of identifying the right problems to solve with AI and using it as an optimization rather than the core business model.

Main themes

  • AI product development
  • custom toolchains
  • business models
  • data management
  • AI optimization
  • problem-solving with AI
  • fine-tuning models
  • resource constraints
  • industry applications
  • innovation and disruption

What commenters say

  • Some commenters agree that building a viable business with AI requires a solid foundation and a clear understanding of the problems to be solved, rather than just relying on AI as a solution.
  • Others argue that fine-tuning models is not always the best approach, especially for businesses with limited resources and unstructured data.
  • There is a debate about the value of using AI as a core business model versus using it as a tool to improve an existing business.
  • Some commenters believe that having a viable business before integrating AI is crucial, as it allows for a more stable foundation and a clearer understanding of how AI can be used to improve the business.
  • Others think that AI can be used to create new business models and opportunities, even if it's not a traditional or established industry.
  • The importance of data management and evaluation is highlighted, with some arguing that it's a significant challenge for businesses to build and maintain their own AI toolchains.
  • There is a discussion about the role of AI in solving real-world problems, with some arguing that it's essential to identify the right problems to solve and use AI as an optimization, rather than trying to create new problems to solve with AI.