On September 7, OpenAI announced on its official website that it will hold the first global "Developer Conference" - OpenAI DevDay on November 6 in San Francisco, California, USA. (Registration address: https://devday.openai.com/)
At this conference, OpenAI will introduce some new products and technology trends, and you can participate in group meetings hosted by OpenAI technical staff. Both offline and online live broadcast methods are used to facilitate those viewers who cannot reach the scene.
Sam Altman, CEO of OpenAI, said, “We look forward to showcasing OpenAI’s latest research results at the Developer Conference to help developers build new tools.
picture
OpenAI’s first developer conference has the following highlights:
1. Will GPT-5 appear?
Affected by factors such as user privacy and data security, OpenAI has slowed down the development process of large language models. At an event hosted by MIT on April 15 this year, OpenAI interviewed and asked Sam Altman online. The development progress of GPT-5.
Sam said that GPT-5 will not be trained in the short term. Regarding the open letter initiated by Musk and thousands of other people to publicly suspend training that is more powerful than GPT-4, Sam believes that Open AI spent 6 months optimizing its security before officially launching GPT-4.
However, on July 31 this year, OpenAI submitted GPT-5 trademark registration information and signed an AI security agreement with US officials to develop generative AI technology in a transparent, safe and stable method.
Therefore, after receiving official support and investing heavily in the security of products such as ChatGPT, OpenAI will change its development strategy for GPT-5 and show it at the developer conference.
picture
2. When will the multi-modal function of GPT-4 be opened?
Since OpenAI first released the GPT-4 model in March this year, it has demonstrated its multi-modal capabilities, such as asking questions to GPT-4 by sending pictures.
But apart from text, GPT-4 has not opened up more functions. Therefore, many people are concerned about when the multi-modal function will be opened to the public.
3. What products and functions will be launched to improve the security of ChatGPT?
On July 26 this year, OpenAI canceled the AI content detection tool it developed. The reason is that the recognition rate is low and it cannot accurately detect text automatically generated by generative AI such as ChatGPT.
However, in order to improve the security of generative AI, Google released the first "digital watermark" SynthID. The product can be embedded directly into images synthesized by Imagen, Google's text-generating image model. Watermarks are invisible to the human eye and are not affected by damaging operations such as changing colors and adding filters; SynthID can also identify watermarks in pictures.
Will OpenAI follow Google's lead and release similar products to enhance the security of ChatGPT to expand its application scope?
Since OpenAI launched the API, it has been continuously updated to include the most advanced AI models, making it easier for developers to integrate the most advanced AI technology into their projects and products through simple APIs, and enhance user experience. .
Today, more than 2 million developers around the world are using AI models such as GPT-4, GPT-3.5, DALL·E, and Whisper. Users build intelligent assistants suitable for various fields such as finance, law, retail, and insurance.
The material of this article comes from the official website of OpenAI. If there is any infringement, please contact us to delete it.