OpenAI’s ChatGPT has dominated the field of generative AI since its launch, making a significant impact on various sectors such as teaching, writing, and technology. It even put tech giants like Meta and Google on the defensive. For the past six months, CEOs of major companies rushed to introduce ChatGPT plugins. OpenAI seemed untouchable with its strong competitive advantage, also known as a moat. However, that may be about to change.
The concept of a moat comes from Warren Buffett and refers to a company’s competitive edge. Just like Microsoft bundled its software with Windows, OpenAI had a moat through its widely-used free version of ChatGPT. It became the fastest-downloaded internet app ever, with 100 million users in just two months. This widespread exposure to advanced AI powered by large language models gave OpenAI a significant advantage, especially since it was offered for free.
But now the competition is catching up. Meta recently announced the launch of Llama 2, a new AI model developed in partnership with Microsoft. While they claim it is open source, the open-source community argues otherwise. Companies with over 700 million monthly active users must obtain special permission from Meta to use Llama 2, making it difficult for Meta’s competitors like Google to access the model. This contradicts the principles of open source defined by the Open Source Initiative.
Furthermore, Llama 2 is not as user-friendly as OpenAI’s ChatGPT. It is designed for technical and corporate users who want to leverage the model to power their own applications and services. Nevertheless, Llama 2 presents a powerful tool for companies seeking the advantages of generative AI. Though it may not be strictly open source, it is available at no cost compared to later OpenAI models. Initial evaluations suggest that Llama 2 is on par with ChatGPT.
However, OpenAI faces challenges in terms of quality. Some in the AI community believe that the responses from GPT-4, the advanced AI model behind the premium version of ChatGPT, have become less intelligent. Insider reported that a redesign of the model aimed at faster response times may have compromised the quality of the responses. Researchers from Stanford University and UC Berkeley published a paper supporting this claim, highlighting issues with math problem-solving, code generation, and meaningful responses.
OpenAI needs to address these quality concerns, especially as competitors enter the market and the initial excitement surrounding ChatGPT diminishes. Recent data shows a 9.7% drop in traffic to the ChatGPT website from May to June. OpenAI’s early entry into the generative AI space allowed them to capture the interest of millions before Google, Meta, and others could do the same.
In May, a memo from an anonymous Google engineer surfaced, asserting that open-source models would eventually surpass closed models. The memo, titled We Have No Moat, And Neither Does OpenAI, suggests OpenAI must come to terms with this reality before the release of Llama 3.
As the landscape of generative AI evolves, it remains to be seen whether OpenAI can maintain its position as the frontrunner. The emergence of competing models like Llama 2 challenges OpenAI’s moat, while concerns about the quality of ChatGPT’s responses put its dominance at risk. OpenAI must adapt and improve to stay ahead in this rapidly advancing field.
