Should wrapper startups be indifferent to advancements in AI Agents such as custom GPTs or Microsoft’s Copilot?
The straightforward answer is no. They shouldn’t be indifferent to custom GPTs, or any advancements made by large tech companies as a simple plug-in or feature could immensely affect their business and all the funding they were planning on getting next month as what took them eight months or even more to build was done by tech giants like OpenAI or Google already in the space. However, another question is also how strong are these custom GPTs?
For those unfamiliar with the term “wrapper startups,” they are companies that enhance OpenAI’s existing models, such as ChatGPT, by developing complementary plug-ins. These startups quickly introduce features that might not be implemented by OpenAI directly. For instance, the recent addition of a PDF feature to ChatGPT could pose a challenge to wrapper startups offering similar capabilities, potentially impacting their viability within a short timeframe. In the rapidly evolving AI landscape, being swift and innovative is essential.
This matters primarily to those planning to start a wrapper startup. Unfortunately, this news will impact the founders most in this space. Individuals seeking to save time in their workday will likely use any application that facilitates this goal. With the introduction of custom GPTs, startups may face disruption as anyone with access to ChatGPT-4 can replicate their entire business models. These replicas can be marketed on the GPT Store, raising concerns about the sustainability and longevity of the hype surrounding such AI Agents.
The significant growth in AI startups emphasizes the importance for these startups, whether wrapped around existing businesses or not, to clearly define their Unique Selling Propositions in terms of their offerings. Despite the anticipated substantial contribution of AI to the market, with a projected $15.7 trillion by 2023 according to a PWC study, funding remains a challenge for new entrants.
However, custom GPTs and Microsoft copilot features also don’t seem to be a long-term solution to actionable AI. In the short term, companies may use GPT’s LLMs with APIs to deliver basic tasks, nonetheless moving forward, an agile approach would have to be considered to bypass the current limitations of actionable AI that would cater to improved accuracy and complex action outputs. This encourages a more calculated approach in an environment that can shift overnight.
Agile Loop follows a scalable, practical approach by incorporating LLMs with Large Action Models as it ensures the architecture behind the model cannot be shaken through short-term plug-ins. It comes down to the value that Agile Loop is offering in re-inventing human interaction with computers as the team believes in performing clever actionable tasks by your computer rather than routine plug-ins that custom GPTs or Copilot is offering right now.
When custom GPTs or Copilot don’t cooperate with your long-term automation goals, Agile Loop can be seen building a durable foundation that can not be wiped out by short-term solutions as the researchers have tested 13 different approaches to challenge the current approaches in the market which are using LLMs + HTML +APIs. Agile Loop aims to bridge this gap in the market by building a unique model based on a pre-trained action decision transformer, coupled deep-reinforcement learning, and backed by proprietary human-software interaction data.
