Navigating the AI Landscape: Insights from OpenAI Dev Day
Written on
Chapter 1: The Evolution of OpenAI
While drafting this article, OpenAI was experiencing considerable shifts in its leadership. This led me to ponder whether this discussion would resonate with OpenAI developers and users of ChatGPT. However, I resolved to complete it. Regardless of the future of OpenAI and ChatGPT, I aspire for this piece to serve as a resource for developers and users navigating the new AI landscape.
Every year, I prioritize attending two key developer conferences: Apple's WWDC and OpenAI Dev Day. On November 6th, OpenAI held its inaugural Dev Day, marking a significant step in the progression of AI technologies. We witnessed the introduction of GPTs and the GPT Store, a parallel to the App and App Store but tailored for AI applications.
This article will not provide a detailed recounting of the conference; I highly recommend checking out Sam's keynote at OpenAI Dev Day for that. Instead, it will reflect on the broader implications this conference holds for developers and entrepreneurs. What opportunities does this AI surge present? How can we ensure our innovations do not get eclipsed by both established and emerging technology leaders?
Section 1.1: The AI Revolution
The technological transformation spearheaded by ChatGPT is frequently likened to the moment the iPhone was introduced. Indeed, OpenAI swiftly introduced GPTs along with the GPT Store, reminiscent of the early days of apps and their marketplace.
The launch of GPTs represented a substantial challenge to numerous competitors in the ChatGPT ecosystem. OpenAI has made it remarkably simple for developers—regardless of their coding expertise—to create AI applications using a straightforward formula: Knowledge + ChatGPT = GPTs. Within just a fortnight, tens of thousands of GPTs were generated, and I regularly encounter at least three new GPTs on my X (formerly Twitter) feed daily.
In the initial years following the iPhone's debut, Apple, developers, and users engaged in a period of exploration. Developers began to understand that the types of applications Apple envisioned included those like Uber, TikTok, Snapchat, and Tinder. In contrast, simpler applications, such as flashlights or task lists, no matter how well-executed, risked being supplanted by official offerings. Additionally, attempts to circumvent Apple’s review process or payment systems could lead to the harsh removal of an app from the platform.
These considerations are pivotal for developers and entrepreneurs amid the AI evolution, as they navigate a swiftly transforming technology landscape.
Section 1.2: The Future of GPTs
Returning to the ChatGPT ecosystem, applications that merely enhance the appearance or usability of ChatGPT are likely to fade away in this competitive environment. Notably, Poe launched Poe Bots (akin to GPTs) on October 26th, but without their own LLMs (Large Language Models), their future is uncertain—either they will be outperformed by official apps or be absorbed by ChatGPT.
Conversely, initiatives aiming to employ open-source LLMs like LLaMa from Meta AI, and train them with proprietary data, will face significant hurdles. The Knowledge plus ChatGPT framework provided by GPTs eliminates the necessity of training individual models. Developers can simply upload their data, and the powerful GPT-4 can function as a customized AI, potentially generating revenue.
Subsection 1.2.1: Designing Effective GPTs
While OpenAI has not precisely delineated the boundaries for GPTs, their demonstrations hint at what they expect from developers and the directions they encourage us to pursue. Applications that seamlessly integrate with existing services, address real-world challenges, and enhance quality of life and work efficiency are all highly encouraged by OpenAI. ChatGPTs that are developed without a clear purpose in addressing real-world needs are unlikely to find a market.
Many observe that the core of GPTs is the elimination of the need to retrain a large model; instead, we can create and utilize AI agents. In essence, any repetitive tasks or intermediary roles are likely to be supplanted by GPTs. This encapsulates the nature of AI agents, and we are on the brink of entering a new era defined by them.
Chapter 2: Anticipating Challenges
What obstacles might GPTs or the GPT Store face?
The foremost challenge is identifying the right GPT. With over 10,000 GPTs created in a mere two weeks, organizing, categorizing, and evaluating these options to assist users in finding what they need is a pressing issue. While there are successful models to learn from, the GPT Store resembles a new App Market, more akin to Extension Stores than traditional App or Google Play Stores.
After establishing algorithms for ranking, categorization, and rating, the subsequent challenge is to combat fraud—a persistent issue in many app markets. If left unchecked, this could lead to the downfall of the marketplace.
When launching the GPT Store, OpenAI discussed future revenue-sharing, which raises another question: how can developers be incentivized? The perception that the App Store has a higher quality of apps compared to Google Play is partly due to Apple’s revenue-sharing approach with developers. Although the contentious 70%-30% split exists, the healthy iOS ecosystem, annual developer events, and recognition encourage developers to create for iOS.
Currently, GPT ecosystem developers and users pay a $20 monthly subscription, yet it remains uncertain whether OpenAI will share revenue from these subscriptions with developers or implement a split between paid and free versions for revenue sharing. This decision will have direct implications for both developers and users within the ecosystem.
Other concerns, such as security and intellectual property rights, which were highlighted during the conference, seem to be key focal points for OpenAI, alleviating some of my worries.
In Conclusion
This developer conference is set to undoubtedly propel the evolution of ChatGPT...
Wait, hold on.
As I was finalizing this article, OpenAI underwent a considerable upheaval in its management. The internal structural issues have emerged as significant challenges and risks for both ChatGPT and OpenAI.
As of this article's publication, a satisfactory resolution for all involved has yet to materialize. While Microsoft has announced that Sam and Greg will join its AI subsidiary to continue AI research, this is merely the beginning of the unfolding drama for OpenAI and its users.
As a developer and a ChatGPT user, I hope for a positive resolution to this situation, allowing ChatGPT, OpenAI, and the broader field of AGI to embark on a path of robust and healthy development.
As of the latest news (November 22, 2023): They are back in action.
I hope this will be the final update in the ongoing OpenAI saga.
