Skip to content
TheChatArchiveThe Chat Archive
gpt-4o-2024-08-06· OpenAIAi2 WildChat-4.8M (AI2)

I'm wondering if anyone has a view about the best way to use LLMs and what potential…

  1. User

    I'm wondering if anyone has a view about the best way to use LLMs and what potential investors think about it?

    For example, if building a product and pinging the ChatGPT api works well enough, is that all that's needed to satisfy investors? Or do they want more proprietary technology (models trained with own data)? Or more control (e.g. self hosting a llama model)?

    Interested if anyone has a view.

    Many thanks

  2. Assistant

    When considering the best way to utilize Large Language Models (LLMs) like ChatGPT in a product, there are several factors to consider from both a technical and an investment perspective:

    1. Effectiveness and Cost-Effectiveness:

      • If using an existing API like ChatGPT satisfies your product's needs effectively and cost-efficiently, it can be an excellent choice. This approach allows you to focus on building features and improving user experience without the overhead of developing and maintaining your own models.
    2. Scalability and Performance:

      • Investors often care about the scalability of your product. Demonstrating that you can scale with an API solution can be convincing. However, be prepared to address concerns about performance limitations or latencies associated with third-party services.
    3. Proprietary Technology and Differentiation:

      • Creating proprietary technology with your own trained models might attract investors looking for differentiation in the market. It allows you to tailor models specifically to your use case, potentially leading to better performance and unique features.
      • Proprietary technology could also result in intellectual property, which might add value to your company.
    4. Control and Customization:

      • Self-hosting models like LLaMA might appeal to investors who value control over a tech stack and the ability to customize at a deep level. It can also mitigate dependency risks on third-party services.
      • However, consider the significant resources and expertise needed to self-host and maintain these models.
    5. Security and Data Privacy:

      • Investors will be interested in how you handle data privacy and security. Managing user data locally with self-hosted models might offer privacy advantages over using an API.
    6. Technical and Team Capabilities:

      • Your team's capability to effectively manage, train, and maintain models will be scrutinized. If your team lacks deep ML expertise, using a third-party API might be more feasible.
    7. Go-to-Market Strategy and Speed:

      • Using an API allows for rapid development and deployment, which may align better with a fast go-to-market strategy. This could be a strong point for investors focused on capturing market share quickly.
    8. Long-term Strategy and Vision:

      • Align your technical approach with your long-term business vision. Investors will want to see that your technical choices support your strategic goals and can evolve as these goals change.

    Ultimately, the choice depends on your specific product needs, your team's capabilities, your market differentiation strategy, and the preferences of your potential investors. It may be beneficial to incorporate flexibility—starting with APIs for speed and cost, with a plan to develop proprietary technology or move to self-hosted solutions as needed.