Lao Huang said at the GTC keynote speech that "we are in the iPhone moment of AI", and ChatGPT came with its "APP Store" the next day.

ChatGPT launched a plug-in function, which can obtain real-time information online

On March 24 (Friday), OpenAI announced that ChatGPT launched a plug-in function, allowing ChatGPT to use various tools, network and perform calculations. This move was hailed as the "Apple's App Store" moment by the outside world. The first batch of plug-ins were developed by partners invited by OpenAI, covering areas such as hotel flight reservations, food delivery services, online shopping, and automation platforms.

It is worth mentioning that OpenAI's own web browsing plug-ins and code execution plug-ins have allowed ChatGPT to "live in the moment". Among them, the web browsing plug-in allows ChatGPT to break through the time limit. In the past, ChatGPT could only obtain knowledge before September 2021, but with the blessing of this plug-in, ChatGPT will be able to extract data from the Internet through Bing’s search API, and Answer various questions.

The code execution plugin provides ChatGPT with a Python interpreter that can run in disk space. Users can upload files to ChatGPT, and then download the processing results. With the help of GPT-4's excellent natural language understanding, "upload a video, let ChatGPT cut out the highlight moment, and then send the clip to us." This kind of small task that everyone wants ChatGPT to do in daily life, I believe it will be possible soon. Achieved.

👉Click to go to the "ChatGPT" data page

Google Bard open beta, currently only supports English

On March 21 (Tuesday), Google announced that its AI chatbot Bard has started testing in the UK and the US. Currently Bard only supports English and does not have encoding capabilities, so responses about codes are not supported.

The fact that Google Bard was born to support real-time networking makes it not outdated in this "AI armament war". However, this feature also made Google fall into the controversy of embezzling creator content. Some webmasters and media criticized: Google Bard does not provide reference links like Bing in conversation content, and often does not indicate specific sources. The practice violated their intellectual property rights, and also reduced the traffic and exposure of creators. 👉Click to go to the "Bard" information page

Adobe releases Firefly AI tool, plans to integrate into PS and other applications

On March 21 (Tuesday), Adobe released the Firefly generative creative AI tool. The first two products launched, one can create images based on text prompts, and there are hundreds of styles to adjust.

Another tool can create various effects for English letters based on the text prompts people give.

Currently Firefly can only be used on Adobe's website, but Adobe has plans to integrate it into professional software such as Photoshop and Illustrator. At the same time, Adobe has long been aware of copyright issues in the AI era, and stated that it uses its own gallery, public domain and authorized works to train Firefly, so that creative workers can use Firefly without "worry". 👉Click to go to the "Firefly" data page

Github Copilot X released, you can write code with your mouth

On March 22 (Wednesday), Microsoft's GitHub released a new version of the programming aid Copilot X, which is connected to GPT-4 and adds chat and voice functions, allowing users to learn how to complete specific tasks through natural language. encoding.

This update includes a number of tools, among which GitHub Copilot Chat, which is embedded in the chat box in VS Code and Visual Studio, is a big killer. In addition to real-time question and answer communication, this artificial intelligence can also realize code content recognition and error message display , voice communication and other operations to help developers look at the code, find bugs, and write tests. And Copilot Voice, which can command AI to write code through dialogue, makes the future of "I give ideas, AI programming" gradually come true to a certain extent. The arrival of GPT-4 is expected to free coders from the busy programming work. 👉Click to go to the "GitHub Copilot X" data page

Bing connected to DALL·E, lowering the threshold for AI image generation

On March 21 (Tuesday), Microsoft announced that the Bing preview version is connected to OpenAI's DALL·E model, adding the function of AI-generated images. Users can enter context such as an image description, location or activity, choose an art style, and Bing Image Creator will generate the desired image for it. Currently, Bing Image Creator only supports input in English to generate images, but Microsoft said that more language support will be added in the future. 👉Click to go to the "Bing" profile page

NVIDIA launches a dedicated GPU for ChatGPT, which improves inference speed by 10 times

On March 22 (Wednesday), NVIDIA officially held the GTC2023 conference and released the "nuclear bomb" GPU H100 NVL specially designed for ChatGPT and other large language models. The GPU is HGX A100, and compared to HGX A100 for GPT-3 processing, a standard server equipped with 4 pairs of H100 and dual-GPU NVLink is 10 times faster, which can reduce the processing cost of large language models by an order of magnitude .

In addition, NVIDIA, TSMC, ASML, and Synopsys announced that they have completed a new AI-accelerated computing lithography technology cuLitho, which can increase the computing lithography of next-generation chips by more than 40 times, laying the foundation for the production of 2nm and more advanced chips. TSMC will begin production qualification of cuLitho in June this year.

Lao Huang said at the GTC keynote speech that "ChatGPT, Stable Diffusion, DALL-E and Midjourney have awakened the world's awareness of generative AI" and "We are in the iPhone moment of AI". Regardless of whether we have time to digest it, this AI trend is waiting for us ahead of the times at an unimaginable speed, and we can only embrace it and face it. Because it's there.