The company released several updates, including a new Assistants API, new modalities, custom GPTs, a new store for models, and new pricing policies at its first developer conference in San Francisco.
“These announcements take direct aim at the open source community as a whole with OpenAI striving to replicate much of the functionality found within frameworks and libraries like LangChain and even point solutions including vector databases, such as Meta’s Faiss and ChromaDB, for example,” said Bradley Shimmin, chief analyst at Omdia.
“By lowering its prices and by building the ability to create functions, using multi-modalities (code, images, etc.), and building in embeddings functionality — all as a native aspect of its new GPT’s user experience — OpenAI is trying to position itself as a viable alternative to build-it-yourself, open source development efforts,” Shimmin added.
Shimmin likened OpenAI’s new updates to Google Cloud, which has introduced a new developer experience via its Vertex AI offering. The moves by OpenAI and Google are likely to create competition for model and tool builders working in the open source community, Shimmin said.
Software vendors to feel the heat
Several open source software firms may face the heat and even threaten the survival of a few due to the new OpenAI updates, analysts predict.
“The new Assistants API could enable developers to build more sophisticated AI applications with less effort, which might attract users from other platforms,” said Jim Kaskade, CEO of AI-driven chatbot maker, Conversica.
The new API, according to OpenAI, is expected to provide new capabilities including a Code Interpreter, Retrieval Augmented Generation (RAG), and function calling to handle “heavy lifting” that would previously require developer expertise in order to build AI-driven applications.
The Assistants API, specifically, may cause revenue losses for open source companies including LangChain, LLamaIndex, and ChromaDB, according to Andy Thurai, principal analyst at Constellation Research.
“For organizations that want to standardize on OpenAI, the more their platform offers, the less organizations will need other frameworks such as Langchain and LlamaIndex. The new updates allow developers to create their applications within a single framework,” said David Menninger, executive director at Ventana Research.
However, he pointed out that until the new features, such as the new API, are made generally available, enterprises will continue to put applications into production by relying on existing open source frameworks.
The companies that were offering specific workflow, orchestration, and other tools for RAG might suffer as well along with vector database companies including ChromaDB and Pinecone that may see an impact on their revenue, according to Thurai.
“If GPT-4 Turbo’s RAG feature significantly improves the efficiency and accuracy of information retrieval and can be integrated into AI applications easily, it could pose a competitive threat to Pinecone’s vector search products,” Kaskade said, adding that the company may have to start offering value-added services that OpenAI doesn’t have to mitigate revenue loss.
On the other hand, Ventana Research’s Menninger believes that while capabilities such as RAG might reduce market opportunities for vector database companies, the market diversity in the segment will see “a rising tide lifting all boats.”
The new updates could also act as a catalyst for innovation in the market, according to Omdia’s Shimmin.
GPT-4 Turbo and multimodal features to create ripples
OpenAI’s DevDay event also saw the launch of a new GPT-4 Turbo model, which is not only faster and more efficient, but also cheaper than existing offerings.
“GPT-4 Turbo is more capable and has knowledge of world events up to April 2023. It has a 128k context window so it can fit the equivalent of more than 300 pages of text in a single prompt,” the company wrote in a blog post, adding that the model has been optimized to improve performance and reduce cost for input and output tokens compared to GPT-4.
The new model with its advanced features along with the reduced pricing will make it extremely difficult for a lot of firms, especially startups, to compete with OpenAI in the generative AI domain.
“Many generative AI small firms and startups were struggling with outreach and awareness. Unless they repurpose themselves to solve specific business problems, they can’t compete against these giants,” Thurai said.
Businesses that emerged earlier this year with value proposition that tantamount to a simple wrapper on the existing ChatGPT APIs will suffer more as they never had a long shelf life, according to Forrester analyst Rowan Curran.
“The new announcements from OpenAI will force them to either adapt and provide some differentiated value-adding features or to expire,” Curran said.
GPT-4 Turbo along with its reduced pricing overcomes major limitations of ChatGPT, including no support for multimodal capabilities. The Turbo model introduces two versions — one just for text, which is cheaper and the other for both text and images.
The text-only model is priced at $0.01/1,000 input tokens and $0.03/1,000 output tokens compared to GPT-4 32K model’s input cost that stood at $0.06. GPT-4 Turbo will cost $0.00765 for processing a 1080×1080 pixels image, the company said.
LLM providers including Cohere and Anthropic will also face stiff competition due to the updates introduced by OpenAI.
“The DevDay announcements and improvements could kill Anthropic and Cohere if they don’t innovate faster. As of now, GPT-4 Turbo is miles ahead of the competition,” Thurai said, adding that enterprises can expect to see some updates from both model providers soon.
GPT-4 Turbo is expected to lower the entry barrier for developers who want to build their AI applications but are resource constrained, Forrester’s Curran said.
The new capabilities also makes OpenAI more appealing for large businesses, who have so far hesitated from adopting some of OpenAI’s LLMs due to their high costs, and the availability of alternative open source based LLMs.
New revenue stream for enterprises
Despite the impact of OpenAI’s announcements on rivals and open source software firms, experts and analysts believe that these very updates could help in generating new revenue streams for enterprises.
“There are quite a few new applications that can leverage these updates, including customized AI solutions for industry sectors, advanced chatbots, voice-enabled applications, advanced data analysis and reporting, AI-powered games and entertainment, programming and code generation, integration services, and voice and text analysis,” Conversica’s Kaskade said.
The Assistants API, according to OpenAI, can have several use cases, ranging from a natural language-based data analysis app to a coding assistant and an AI-powered vacation planner. Other use cases include a voice-controlled DJ and a smart visual canvas.
Copyright © 2023 IDG Communications, Inc.