What happened at OpenAI’s First Ever Developer Event in San Francisco?
OpenAI’s inaugural developer event, held recently, showcased a flurry of significant announcements. Among the key highlights was the remarkable growth of ChatGPT, which, within just a year of its launch, has garnered a considerable following, with more than 100 million users engaging with the platform on a weekly basis. This achievement solidified ChatGPT as one of the fastest-growing consumer products to reach the 100 million monthly user milestone, an accomplishment realized mere months after its initial release. Additionally, OpenAI disclosed that an impressive cohort of over 2 million developers is actively involved in creating innovative solutions through the platform’s application programming interface (API).
An eagerly anticipated unveiling at the event was the introduction of the highly advanced GPT-4 Turbo. An enhanced iteration of the renowned GPT-4 model, GPT-4 Turbo offers two distinct versions tailored to accommodate varying computational needs. The first version is dedicated to text analysis, while the second version is equipped to handle both text and image comprehension. Notably, the pricing structure for the text-only model entails a cost of $0.01 for every 1,000 input tokens and $0.03 for every 1,000 output tokens. Moreover, the image processing capabilities of GPT-4 Turbo come at a rate of $0.00765 for handling a 1080×1080 pixels image. With an expanded context window of 128,000 tokens, approximately four times the capacity of its predecessor, GPT-4 Turbo provides an extensive knowledge cutoff up until April 2023, surpassing the limitations of GPT-4, which had a knowledge cutoff set at September 2021.
OpenAI’s decision to enable users to create their personalized versions of GPT marks a significant step in facilitating innovative usage scenarios. This novel feature allows users, without necessitating any coding expertise, to build customized bots by simply employing prompts. Furthermore, enterprise clients will have the opportunity to construct internal GPTs using OpenAI’s robust knowledge base. OpenAI further revealed that developers can seamlessly integrate GPTs with various databases and knowledge repositories, including email systems, enabling the seamless importation of external information into the platform.
Anticipating the launch of the GPT Store later this month, OpenAI outlined its plans to provide a platform for users to publish their customized GPTs. Initially, the store will feature creations from “verified builders,” with potential opportunities for creators to be monetarily rewarded for their contributions. This approach aligns with the recent introduction of a creator payment program by Quora’s Poe platform, which incentivizes creators for their valuable contributions. Notably, example GPTs will be accessible to ChatGPT Plus and ChatGPT Enterprise customers starting immediately.
In response to growing market demands, OpenAI unveiled its new Assistants API, aimed at empowering developers to create tailored “agent-like experiences.” Developers can leverage this interface to design specialized agents capable of retrieving external knowledge or executing specific programming functions, catering to a diverse range of use cases, from assisting in coding tasks to facilitating AI-driven vacation planning services.
Moreover, OpenAI launched the DALL-E 3 API, enabling users to leverage the text-to-image capabilities of the DALL-E 3 model through a user-friendly API, complete with integrated moderation tools. Notably, the DALL-E 3 API offers output resolutions ranging from 1024×1024 to 1792×1024, available in multiple formats at a cost of $0.04 per generated image.
Further expanding its suite of offerings, OpenAI introduced a new text-to-speech API, branded as the Audio API, featuring a selection of six distinct preset voices, namely Alloy, Echo, Fable, Onyx, Nova, and Shimer. The API provides seamless access to two generative AI models and is priced at $0.015 per 1,000 input characters, facilitating the transformation of text-based content into high-quality audio outputs.
Addressing concerns regarding intellectual property (IP) protection, OpenAI unveiled its Copyright Shield program, designed to safeguard businesses utilizing OpenAI’s developer platform and ChatGPT Enterprise from potential copyright claims. Under this initiative, OpenAI has committed to covering legal expenses incurred by customers facing IP lawsuits related to content created using the company’s tools. This approach reflects a broader industry trend, with major players such as Microsoft, Cohere, Amazon, and IBM also implementing similar indemnification policies for their respective customers.
Additionally, the event saw OpenAI’s pledge to introduce a program that enables companies to develop customized customer models with support from the company’s research team. OpenAI also announced the elimination of the model picker in ChatGPT, simplifying the user experience and allowing for a more streamlined utilization of the platform’s functionalities. Concurrently, the company launched Whisper large v3, the latest iteration of its open-sourced speech recognition model, catering to evolving market demands. Notably, OpenAI has taken measures to enhance the token limit rate for all paying GPT-4 customers, providing a more optimized and efficient experience for its valued user base.
