OpenAI Enhances ChatGPT Capabilities with GPT-4 Turbo Launch
Shari Lynn Kramer / 07 Nov 2023
OpenAI is taking steps to enhance the capabilities of its digital chatbot, ChatGPT, in highly significant ways. People have begun to notice that the AI's database tends to fall behind current affairs since it only knows up until September 2021. But now OpenAI has plans to ensure it stays more up-to-date. During their inaugural developer conference, OpenAI's CEO, Sam Altman, unveiled GPT-4 Turbo. This new model upgrades the AI's knowledge base to include information up to April 2023, a massive advancement from its previous limitations. Altman assured that they don't intend to let the information lag this far behind again and will be updating it more regularly.
OpenAI's initiative extends beyond just updating the knowledge base. To align ChatGPT with current affairs even more, they are integrating a new element called the “retrieval function”. This functionality will allow users to feed documents or databases into the chatbot, refining its understanding. Another significant enhancement associated with GPT-4 Turbo relates to the chatbot's response mechanism. Its scale of input will accommodate up to 128,000 tokens, translating to around 300 pages of text. Earlier versions like GPT-4 could only handle inputs between 8,000 and 32,000 tokens in size.
Developers have additional reasons to be excited about GPT-4 Turbo. OpenAI has significant price reductions for app development using its API. The new model will be cheaper by approximately 2.75 times than its GPT-4 predecessor. The company's current pricing model will be $0.01 for 1,000 input tokens and $0.03 for 1,000 output tokens. It's unclear if Turbo will be restricted to ChatGPT Plus users, but the company anticipates a fully functional model in the forthcoming weeks.
During the developers conference, Altman forecasted an engaging future for the company’s technology. His vision entails chatbots evolving into intelligent “agents” capable of carrying out any task humans require. OpenAI's focus now is on facilitating anyone, even those with no programming skills, to develop customized chatbot models or GPTs.
These could be designed to serve specific areas or tasks and can be used privately, publicly shared, or solely employed by a company. A GPT could be executed to instruct school courses like computer science or help consumers create marketing graphics. The customized bot functionality will be accessible via a new GPT builder application. OpenAI is planning to host these GPTs on an online store, with revenue-sharing arrangements with the developers who create them. Development in this direction marks the groundwork for more potent AI programs that OpenAI already has in the pipeline.
According to Altman, the evolution of GPTs over time will pave the way for intelligent assistants, precursors to agents that will eventually be capable of complex actions and planning on your behalf. Altman emphasized the significance of gradual, incremental deployment and encouraged people to start building and engaging with these agents. This exposure, he believes, will prepare us for when AI becomes an even more integral part of human life.