OpenAI’s text-generation AI chatbot ChatGPT has captured global attention. This tool can draft essays, produce code, and more using brief prompts, significantly boosting efficiency. However, there are concerns about its potential misuse. In this article we will list Most Recent ChatGPT Updates.
The reality is, AI tools like these are here to stay. Since its introduction a few months ago, its usage has seen immense growth. Big-name brands are already harnessing its capabilities, utilizing it to create ad and marketing content.
OpenAI is continuously refining and investing in this technology. ChatGPT’s capabilities have recently been enhanced by GPT-4, the newest language model from OpenAI’s research team. Moreover, those who subscribe to ChatGPT can tap into GPT-4‘s more sophisticated and fluid writing capabilities, surpassing the previous model’s performance. Besides, OpenAI has now linked ChatGPT to the internet, offering plugins in an alpha stage to eager users and developers on a waitlist.
Below is a chronological list of ChatGPT’s updates and launches, with the most recent first, to be refreshed periodically.
Timeline of Latest ChatGPT Updates
August 28, 2023
1. OpenAI Introduces ChatGPT Enterprise: This advanced version of ChatGPT is equipped with features that match standard ChatGPT, including composing emails, creating essays, and troubleshooting computer code. What sets it apart is the inclusion of “enterprise-grade” privacy settings, advanced data analysis tools, and options for heightened performance and customization.
2. ChatGPT’s Popularity in America: A recent study by Pew Research indicates that while ChatGPT’s awareness is on the rise, its actual usage is limited. The survey reveals that just around 18% of Americans have tried the AI model, countering the perception of it being ubiquitously used or alarming.
August 22, 2023
3. Fine-Tuning Feature Added to GPT-3.5 Turbo: OpenAI has enriched GPT-3.5 Turbo with the capacity for fine-tuning. This enhancement allows companies accessing GPT-3.5 Turbo via OpenAI’s API to make the model more instruction-compliant. With this feature, they can:
- Set the model to always reply in a particular language.
- Enhance the model’s consistency in response formatting.
- Tailor the output’s “feel,” adjusting aspects like tone to better align with a specific brand or voice.
- Most crucially, it permits OpenAI users to trim down text prompts, leading to faster API interactions and reduced costs.
Also, OpenAI has forged a collaboration with Scale AI to empower companies with the capability to fine-tune GPT-3.5. The nature of this partnership suggests an integration of Scale AI’s expertise to enhance the fine-tuning process for OpenAI’s language model.
Fine-Tuning Cost Breakdown:
- Training Cost: $0.008 for every 1,000 tokens.
- Usage Input Cost: $0.012 for every 1,000 tokens.
- Usage Output Cost: $0.016 for every 1,000 tokens.
This pricing structure offers clarity for companies interested in leveraging the fine-tuning capabilities, helping them budget and strategize accordingly.
August 16, 2023
4. OpenAI’s Acquisition of Global Illumination: Marking its maiden public acquisition in a span of seven years, OpenAI declared its purchase of Global Illumination. This New York-based startup, specializing in the utilization of AI for the creation of creative tools, infrastructure, and digital encounters, is now a part of OpenAI’s dynamic ecosystem.
As mentioned in a concise announcement on OpenAI’s official blog, the whole team from Global Illumination is set to contribute to key OpenAI offerings, prominently including ChatGPT.
August 10, 2023
5. Expansion of ‘Custom Instructions’ to Free ChatGPT Users: Lastly, in a move aiming for wider accessibility, OpenAI publicized its decision to extend the custom instructions feature to encompass users on its free service tier. This added functionality empowers users to lay down specific preferences and stipulations they desire the AI chatbot to take into account during its responses.
This were most recent ChatGPT Updates. You can read more related articles: