Weekly AI events | Huawei will release large model products, OpenAI visits exceed 1 billion times

Welcome to the webmaster's home[Weekly AI Events], here we record the hot content in the AI ​​field worthy of attention in the past week to help everyone better understand the dynamics and trends in the field of artificial intelligence Development trend.

Part1 dynamics

[domestic news]

Bytes internal testing conversational AI project

ByteDance is internally testing an AI dialogue product called "Grace", which is currently only used for internal experience testing.

The Artificial Intelligence Law has been included in the legislative work plan

Recently, the General Office of the State Council issued the "State Council 2023 Legislative Work Plan". It shows that the draft artificial intelligence law is ready to be submitted to the Standing Committee of the National People's Congress for review.

Huawei refutes rumors that “Pangu Chat” was named

There are reports that Huawei is expected to release Pangu Chat, a multi-modal 100 billion-level large-scale model product that directly benchmarks ChatGPT at the "Huawei Developer Conference 2023". A senior Huawei Cloud official responded to the news: "Fake." Huawei's official statement did not directly mention the release of the Pangu model. Huawei will not have a name like "Pangu Chat", nor will it "benchmark" industry products.

The first Wenxin Qianfan large model platform was established

Recently, Wuxi Qianfan Racing Technology Co., Ltd. was established. The legal representative is Qiu Chijuan and the registered capital is 120 million yuan. The company's business scope includes Internet of Things application services; artificial intelligence application software development; blockchain technology-related software and services, etc.

Baidu launches code writing assistant Comate

Baidu Intelligent Cloud has released the "Comate" code assistant, a new generation of coding assistance tool based on large models, and has opened for testing. Test results show that among the codes written with the help of “Comate”, nearly 50% of the suggested codes were adopted by developers.

"Wen Xin Yi Yan-Turbo" beta test

Baidu Intelligent Cloud announced at a technical exchange meeting held in Chengdu that its Wen Xinyiyan reasoning performance has improved 10 times since the start of internal testing in March for two months. Currently, enterprises can apply to test the "Wenxin Yiyan-Turbo" high-performance mode on the Wenxin Qianfan large model platform.

Huawei version of ChatGPT will be released on July 7

Huawei is about to launch a multi-modal 100-billion-level large-scale model product, which will be displayed and tested internally at the Huawei Cloud Developer Conference (HDC.Cloud2023) held on July 7 this year, mainly for ToB/G government Enterprise customers.

Xmind mind map acquires AI product ChatMind series software

Xmind Mind Map has successfully acquired the AI ​​product ChatMind series software. Through this acquisition, Xmind will further strengthen the AI ​​features of its products.

iFlytek: Will release iFlytek Spark cognitive large model V1.5

iFlytek has now issued an announcement, announcing that the company will release new progress on the iFlytek Spark cognitive model as scheduled on June 9, and this conference will release the iFlytek Spark cognitive model V1.5. According to reports, the iFlytek Spark Cognitive Big Model V1.5 has achieved breakthroughs in open question and answer, with multiple rounds of dialogue and mathematical capabilities further upgraded, and text generation, language understanding, and logical reasoning capabilities continuing to improve.

Homework Helper Internal Test Education Model

Currently, Zuoyebang is conducting internal testing of a large education model based on the Chinese market. This model covers multiple educational application scenarios such as multi-disciplinary problem solving, Chinese and English composition correction, and multi-language dialogue, and includes services such as tool apps, smart hardware, and books.

360 Intelligent Brain Large Model Application Conference is scheduled for June 13th

360 Intelligent Brain Large Model Application Conference will be held in Beijing at 14:00 on June 13. "360 Intelligent Brain" is a large language model with hundreds of billions of parameters independently developed by 360. In April this year, the implementation of "360 Intelligent Brain" combined with 360 search, smart hardware and other scenarios has gradually matured and is open for internal testing.

[International News]

WordPress introduces artificial intelligence writing assistant

Automattic, the company behind WordPress.com, launched an AI assistant on Tuesday to power the popular website content management system. The assistant can be easily integrated into WordPress.com and all websites powered by Jetpack. When writing articles or pages, users can add an "AI Assistant" module to the content. Users can enter a prompt in natural language, and the AI ​​assistant will start generating text based on this prompt. In addition to generating content ideas, AI assistants can also create structured lists and tables in blog posts.

ChatGPT Official App Big Update

ChatGPTiOS version application has been updated, adding Siri and shortcut command functions, which can be used in conjunction with Siri and shortcuts. This update also allows the app to run on the iPad, allowing the app to better adapt to the iPad's larger screen. Another new feature is drag-and-drop support, so individual messages can now be dragged and dropped into other apps.

Instagram is building an AI chatbot

Developers have discovered that Instagram is developing a chat system similar to Snapchat’s popular MyAI robot. Instagram users will be able to choose from 30 different personalities for their AI conversation partners, who will answer questions and provide suggestions.

OpenAI has not yet started training the large language model GPT-5

The Microsoft-backed OpenAI startup has pledged not to conduct research and development on GPT-4’s successor GPT-5 “for some time,” leaving many industry executives and academics concerned about the rapid development of large-scale language models led by Sam Altman. Developments have expressed concern, but months have passed and OpenAI has still not started training GPT-5.

OpenAICEO: Due to the special company structure, listing is not considered for the time being

OpenAI CEO Sam Altman said at a conference that the company has no plans to go public in the short term. He explained, I don't want to be sued by... the public markets, Wall Street, etc., so, (the answer is) no, not really interested in that. Our structure is very special. We have this profit cap thing.

iOS17 keyboard automatic correction upgrade newly added transformer language model

iOS17 keyboard automatic correction has received a comprehensive upgrade, and a new transformer language model has been added. This is an advanced on-device machine learning language model for text prediction that continuously improves the experience and accuracy every time users type.

Artificial intelligence has caused nearly 4,000 people to lose their jobs in May

Artificial intelligence caused nearly 4,000 job losses last month, according to Challenger, Gray & Christmas, as interest grows in the rapidly evolving technology's ability to perform advanced organizational tasks and reduce workloads.

OpenAI becomes the top 20 most visited websites in the world

OPENAI website monthly unique visits soared to 1 billion in March. A total of 847 million users visited OpenAI's website in March, a month-on-month surge of more than 54%. If we look at online visits alone, OpenAI has ranked among the 20 most visited websites in the world.

Part2Popular Products

AlphaDev makes a stunning debut, subverting the landscape of human algorithms!

The Alpha family has added a new member: AlphaDev. The foundation of the entire computing ecosystem may be overturned by new algorithms created by AI! AlphaDev can not only speed up sorting algorithms by 70%, but can even be as much as three times faster than humans on some algorithms. The C sorting library has changed for the first time in over a decade. AI optimizes the world's code and reaches a new milestone.

Runway’s AI video editor Gen-2 is open for free trial

New York startup Runway is developing AI-enabled video editing tools such as the Gen-1 AI model, where users can already edit videos using simple text prompts. The new Gen-2 feature is further enhanced with support for creating new videos from text. According to new reports, RunwayGen-2 is now available in browser and smartphone apps for iOS.

HuBo Technology releases and open source self-developed large model TigerBot

TigerBot's performance is close to 96% of OpenAI's models of the same size, and surpasses it in some areas such as inferential question answering. The main functions of TigerBot include content generation, open question and answer, summary summary, code generation, math questions, information extraction, picture generation, etc.

The low-code development framework PromptAppGPT is launched with fully automatic compilation

PromptAppGPT, a low-code development framework based on ChatGPT, is online, supporting low-code prompt (i.e. Prompt) application development, GPT text generation, DALL-E image generation, online code editor compiler runner, automatic user interface generation, and plug-ins Extensions and other functions.

Google StyleDrop: Just give a picture to copy the style of the work

As long as a picture is given as a reference, StyleDrop can learn the style of the picture and help generate works of the same style. StyleDrop captures the complexity of custom styles, including color schemes, shadows, design patterns, and local and global effects.

HairstyleAI: An AI “tony” teacher

With HairstyleAI, you can try various hairstyles and colors without leaving home. All you have to do is upload multiple avatars and HairstyleAI will use its AI technology to generate 40 popular styles, 10 colors and 160 realistic renderings of your new look. Users can see looks in every style and color combination, making it easier than ever to find a new hairstyle that suits their face shape and personality.

Dumme: Turn long YouTube videos into short videos in just a few minutes

Dumme is a YCombinator-backed startup that has been using AI to convert longer YouTube videos into shorter ones. The company has hundreds of video creators testing its product and a pre-launch waitlist of more than 20,000.

Part3View

Zhou Hongyi: Large model writing is not about search logic. Questions need to be appropriate and profound

When Zhou Hongyi talked with Feng Lun, chairman of Yufeng Group and founder of Wantong Group, he pointed out that large model writing is not the logic of search, but the rapid organization and formation of an article based on the knowledge he has "read".

Baidu: AI partners are part of the intergenerational change in search Prompt word engineering is a required course

AI partners are part of the intergenerational change in search. Better use of prompt words and AI partners will lead to greater and greater advantages. Prompt word engineering should be a required course for everyone in the AI ​​era. Use large models to create more and gain more.

IMF official warns: Artificial intelligence could "severely disrupt" the job market

Senior IMF officials warned that generative artificial intelligence could cause "significant disruption" to labor markets and called on policymakers to quickly develop rules to govern the new technology.

Apple Software Director: Why I am both afraid of AI and not so afraid of AI

Apple software director Craig Federighi was interviewed by FastCompany about Apple’s security and privacy features. In the process, he also talked about the unexpected and even evil ways that artificial intelligence will affect our privacy. When talking about the threat of AI to privacy, Craig admitted, "There is no doubt that AI-based tools will become better and better at discovering potential security flaws and exploitation paths. The good news is that people with ulterior motives will use AI tools to steal your data." meaning, these tools can also be used by people trying to harden their code and protect their users."

Cook: Apple is paying close attention to ChatGPT

Apple CEO Tim Cook recently said that he uses ChatGPT, an AI chatbot, and is excited about its "unique applications." He added that Apple is paying close attention to ChatGPT.

Part4Depth

Entrepreneurs use AI-generated videos to subvert marketing

The application field of AI vision students has not yet produced obvious value. But there is an early-stage startup company that only received seed round financing, which may have given an enlightening attempt in the application direction of visual generation. It is Tavus. This company uses generative AI technology to allow users to "digitally clone" themselves with a very low operating threshold, and can generate personalized videos in their own image in batches.

The Internet celebrity named by Wang Sicong became an AI girlfriend, earning 720,000 a month?

With only 6 yuan, you can see the friend circles of Internet celebrities who have been commented by Wang Sicong, and you can also chat, make video calls and feel companionship together. Such a "cost-effective" activity appeared on the Internet celebrity "Hanzo Forest" (hereinafter referred to as Hanzo) with 2.626 million fans on Weibo.

Using AI to sell scrap materials, the platform has sold US$300 million in two years since its launch

"Regeneration Doctor" mainly provides two functions: AI identification of mixed waste and a global trading platform, which are realized with the help of multiple products such as SaaS, intelligent hardware, and trading platforms. In 2022, "Doctor Regeneration" will enter the Google Overseas Accelerator. After graduating from the Google Accelerator in 2023, our transaction volume will already be US$300 million.

This college entrance examination, AI candidates are very busy

This year’s college entrance examination candidates have caught up with the AI ​​year. Many people tried to write this year's college entrance examination essay using GPT-4, Claude, Wenxinyiyan, Tongyi Qianwen, and iFlytek Spark. Looking at the writing performance of all AI candidates, at a glance, they all failed in their marks. Quoting scriptures, logical reasoning, and "experiential explanations" are all not a problem, and the overall performance is not bad.

Programmer was eliminated after refusing to use GPT-4 tools

A technical person on Twitter shared an incident he supervised, that is, a programmer with 19 years of coding experience and who could write 100% of the code by hand eventually lost to a programmer with only 4 years of experience, but who was good at using Copilot and GPT- The juniors of 4 were rejected in the interview because they were unwilling to use auxiliary coding tools and only wanted to write controllable code.

AI big models are returning to value investing?

The stock price of AI concept "parted ways", with sharp rises and corrections coexisting. In recent times, the trends of individual stocks have begun to diverge. Some continue to enjoy the carnival, while others have experienced a correction. Is this a precursor to the return of large AI models to value investing? Or is it a brief calm before the market rises further?

The above is the detailed content of Weekly AI events | Huawei will release large model products, OpenAI visits exceed 1 billion times. For more information, please follow other related articles on the PHP Chinese website!