Published on November 16th, 2023
Here Are Five Key Takeaways:
- Evolution of Chatbots into Powerful Agents
- Customizable GPT Models
- Diverse Use Cases
- User-Friendly GPT Creation
- Gradual Deployment and Evolution
OpenAI is making significant updates to ChatGPT to enable it to respond to events and information that occurred after September 2021.
These updates are part of a broader set of improvements to their AI program.
During their first annual developer day, OpenAI introduced GPT-4 Turbo, a new large language model designed to address user requests for more up-to-date information.
GPT-4 Turbo updates the chatbot’s knowledge base up to April 2023 and will continue to improve over time.
To enhance ChatGPT’s ability to discuss current topics, a new “retrieval function” is being added, allowing users to upload documents or databases for reference.
Additionally, GPT-4 Turbo can handle up to “128,000 tokens” of input, significantly increasing its capacity. Previously, GPT-4 was limited to text inputs between 8,000 and 32,000 tokens in length.
OpenAI is also lowering the API prices for developers. GPT-4 Turbo is expected to be about 2.75 times cheaper than GPT-4, with costs of $0.01 per 1,000 input tokens and $0.03 per 1,000 output tokens.
Initially, GPT-4 Turbo is being made available to developers as a preview through OpenAI’s API.
It’s unclear whether Turbo will be restricted to ChatGPT Plus users, but a stable production-ready model is expected to arrive in the “coming weeks” via the ChatGPT interface.
These changes represent a substantial step forward in improving ChatGPT’s capabilities and accessibility.
Creating Custom Chatbots
During OpenAI’s developer conference, CEO Sam Altman shared the company’s vision for the future of AI technology, emphasizing the potential for chatbots to evolve into highly capable “agents” that can perform a wide range of tasks on behalf of humans.
Altman highlighted the desire for AI to be smarter, more personalized, and more customizable, where users can simply request tasks from a computer, and it will handle them efficiently.
To realize this vision, OpenAI is focusing on enabling individuals to create custom chatbot models referred to as “GPTs.” These GPTs can go beyond the standard ChatGPT program and be tailored for specific areas or tasks.
For instance, a user could create a GPT specialized in teaching computer science in schools or helping consumers generate marketing materials like logos and posters from a text prompt.
These custom GPTs can be used privately, shared publicly, or exclusively deployed within a company.
OpenAI is developing a GPT builder app to facilitate this customization effort. This initiative builds on the plugin function introduced for ChatGPT in March and aims to offer GPTs on a digital store, sharing revenue with developers.
The “GPT Store” is set to launch in the near future, and users can already start creating their own GPTs.
In a demo, Altman demonstrated that users can build custom GPTs without requiring programming skills.
Users can communicate their requirements to the bot using language, and GPT will do its best to understand and comply with the instructions.
Altman sees GPTs as a stepping stone towards more powerful AI programs that OpenAI is actively working on. These AI programs will gradually become capable of planning and performing increasingly complex actions on behalf of users.
The company believes in the importance of gradual and iterative deployment, encouraging people to start using these AI agents to gain insights into how they will shape the future as they become more capable.
This approach allows users to adapt to and better understand the evolving AI landscape.
Feature Image Source: Jonathan Kemper