Babbitt | Metaverse Daily Must-Read: It is rumored that Apple is building a large-scale AI framework "AJAX", and has developed an internal chat robot "Apple GPT", and the AI competition is heating up again!

Abstract: Early this morning, Apple was rumored to be building its own large language model and developing generative AI tools. It is reported that Apple is building a large-scale language model AI framework "AJAX", and has developed an internal chatbot in the form of ChatGPT for employees, or known as "Apple GPT". It is reported that this work is currently being led by John Giannandrea, Apple's head of machine learning and artificial intelligence, and Craig Federighi, Apple's top software engineering director.

Image source: Generated by Unbounded AI

Hot info:

Hurun released the Metaverse Potential Enterprise List, Alibaba and Huawei became Top 2

According to news from Jinshi on July 20, the Hurun Research Institute released the "2023 Hurun China Metaverse Potential Enterprise List", listing the top 200 Chinese companies with the most development potential in the metaverse field, and 30 "future stars" enterprise". Among them, Alibaba and Huawei were rated as "Top 2 with the Most Potential", and NetEase, Tencent, and Baidu entered the "Top 5 with the Most Potential". There are 57 new companies on the list, represented by HTC, MiHoYo, and United Imaging.

OpenAI increases the number of messages ChatGPT Plus customers can send to GPT-4 to 50 per 3 hours

Babbitt News, on July 20, OpenAI announced that it has doubled the number of messages that ChatGPT Plus customers can send to GPT-4 to 50 every 3 hours. The new restrictions will come into effect next week.

Hefei Artificial Intelligence and Big Data Research Institute officially launched an angel investment fund

According to news from Jinshi on July 20, recently, at the Hefei Artificial Intelligence and Big Data Research Institute’s first anniversary results conference, the Angel Investment Fund of the Hefei Artificial Intelligence and Big Data Research Institute was officially launched. The fund was established by the Hefei Artificial Intelligence and Big Data Research Institute and the Silicon Valley Guidance Fund of the University of Science and Technology. The fund is mainly to match the project cultivation needs of the artificial intelligence research institute, promote scientific and technological innovation and financial capital "flying together", provide early development financial support for the scientific and technological innovation projects incubated by the Hefei artificial intelligence and big data research institute, and play the role of "base + Fund's "incubation + venture capital" model advantage.

Korean technology giant LG launches AI model Exaone 2.0

According to "Kechuangban Daily" citing The Elec, a few days ago, Korean technology giant LG announced its AI model "ExaOne 2.0". Based on ExaOne 2.0, LG announced three AI platforms that it plans to commercialize: ExaOne Universe, ExaOne Discovery, and ExaOne Atelier. It is reported that LG AI Research plans to develop the platform in the chemical, biological, medical, medical, financial and patent fields, and the platform will be deployed to LG's chemical and biological subsidiaries in the fourth quarter of this year. As previously reported, LG released Captioning AI, the first generative AI commercialization service.

Companies like Twitter are asking too much, AI companies like OpenAI and Cohere are turning to synthetic data

According to news from IT House on July 20, Aiden Gomez, CEO of artificial intelligence company Cohere, recently revealed that due to the high price of data collection from companies such as Reddit and Twitter, AI companies including Microsoft, OpenAI and Cohere have used synthetic data to train the AI model.

Gomez said that synthetic data can be applied to many training scenarios, but it has not been fully promoted yet. For example, Gomez said: If a company wants to train a model in advanced mathematics, it can create two artificial intelligence models, play the role of teacher and student, and let them discuss topics such as trigonometry. Humans are mainly responsible for observing, and if they see any mistakes, they can correct them. It is reported that synthetic data (synthetic data) is data artificially generated through computer technology, rather than data generated by real events. However, synthetic data has "availability" and can mathematically or statistically reflect the properties of the original data, so it can be used as a substitute for the original data to train, test and verify large models.

Samsung and "Semiconductor Legend" Jim Keller reached a cooperation to accelerate the development of AI chips

According to IT House citing Business Korea, Samsung Electronics' semiconductor foundry department has started chip research projects with Tenstorrent and Groq, and the design service team within Samsung's foundry department is responsible for these research tasks. Samsung Electronics will work with Tenstorrent and Groq to develop artificial intelligence semiconductor chips for advanced IT equipment, the sources said. If the mission achieves mass production, its chips are expected to be produced on Samsung's sub-5-nn EUV line and packaged in a 2.5D packaging facility. Industry experts speculate that if Samsung successfully completes the joint project with the two companies, it will have a major impact on the foundry market. In 2021, Jim Keller, known as the "Silicon Fairy", a great figure in the semiconductor industry, officially announced that he would join Tenstorrent, an artificial intelligence chip start-up company, as the chief technology officer and president, and become its CEO in January 2023. the company's board of directors.

Netease Smart Enterprise and Dianhun Network established the "Game Artificial Intelligence Joint Laboratory"

According to a report from the Financial Associated Press, Netease Smart Enterprise and Dianhun Network signed a strategic cooperation agreement on July 19, announcing the establishment of a "Game Artificial Intelligence Joint Laboratory", which will focus on AI competitive robots, anti-cheat, content security, and customer service. , testing, BI, recommendation, user research, voice-driven emoticons and other aspects of in-depth research cooperation, and will also conduct exploratory research and practice around AI capabilities such as intelligent NPC, AI painting, intelligent matching, AI face pinching, and art animation.

Lenovo introduces Baidu's "Wen Xin Yi Ge", AI generated pictures can customize the appearance of notebooks

According to Titanium Media on July 20, Lenovo announced that it has reached an in-depth cooperation with Baidu in the field of AIGC and launched the first ICT AI customization service. Lenovo's official website private customization business has fully introduced Baidu Wenxin Yige, and consumers can customize the appearance of laptops through AIGC-themed painting activities on the official website. It is reported that the private customization business is a special service launched by Lenovo's official website in 2022. Through "high configuration customization + appearance customization", it can meet the individual needs of trendy consumers. Previously, sales of 618 private customization services accounted for more than 20% of Lenovo's official website sales.

Report: In the first half of 2023, the download volume of the AI application market increased by 114% year-on-year, exceeding 300 million times

According to IT House's report on July 20, SensorTower released the "2023 AI Application Market Insights" report today, which shows that in the first half of 2023, the global AI application market will show a trend of rapid growth. It is reported that in 2023, the download volume of AI applications will increase by 114% year-on-year, exceeding 300 million times, surpassing the annual level of 2022. Meanwhile, AI in-app purchase revenue grew 175% year-over-year to nearly $400 million.

The report also pointed out that the United States is the market with the highest AI in-app purchase revenue, contributing 55% of the total revenue of the world's top AI applications. The European market also performed well, accounting for 20% of the total revenue of top AI applications. Among them, "chat robot" and "image generation" are the most popular AI tracks. The download volume of AI "chat robot" application accounts for 49% of the total downloads among the top 100 AI applications in the world, while the AI "image generation" application downloads accounted for 31%.

The ChatGPT API-based chat robot application "Chat with Ask AI" has performed well in the AI "chat robot" track, with more than 25 million downloads and revenue of more than 16 million US dollars (about 114 million yuan). In the AI "image generation" track, the application of DALL-E and Stable Diffusion models has driven the market to thrive. "Lensa AI" has been popular among users for its avatar generation function, and its downloads have exceeded 100 million times.

TSMC: Nearly 50% CAGR comes from AI servers and AI processors

According to news from Jinshi on July 20, TSMC stated that nearly 50% of the compound annual growth rate comes from AI servers and AI processors. This shows the potential and growth rate of the AI market.

Investment and Financing:

Meituan exclusively invested in Zhipu Huazhang, with a post-investment valuation of nearly US$500 million

According to 36 Krypton’s report on July 20, the AI large-scale model company Zhipu Huazhang (hereinafter referred to as “Zhipu”) has completed the B-2 round of financing a few months ago, with an amount of hundreds of millions of RMB, and was exclusively invested by Meituan Zhantou. The post-investment valuation is nearly 500 million US dollars. It is understood that Zhipu’s subsequent rounds of financing have also been completed one after another. The above financing information has been officially confirmed by Zhipu.

Metaverse recommended reading:

"Who is holding back China's ChatGPT? "

With 500 employees, the company's overall valuation is close to 30 billion US dollars. This is OpenAI; with thousands of people, the company's valuation/market value is at best 1-2 billion US dollars. This is a number of AI "small giants" in China. In recent years, the gap between the Chinese and American artificial intelligence circles in the field of large-scale models is not a matter of awareness, starting point, and ability, but a matter of roads and methods.

"Apple doesn't talk about AI, but AI is everywhere, and the large model "AJAX" appeared for the first time"

Early this morning, Apple was rumored to be building its own large language model and developing generative AI tools. It is reported that Apple is building a large-scale language model AI framework "AJAX", and has developed an internal chatbot in the form of ChatGPT for employees, or known as "Apple GPT". It is reported that this work is currently being led by John Giannandrea, Apple's head of machine learning and artificial intelligence, and Craig Federighi, Apple's top software engineering director.

"Meta teamed up with Microsoft to challenge the large-scale model pattern, the latest Llama 2 is free and open source, and can be directly used for commercial use"

In the early morning of July 19th, Beijing time, Meta (formerly Facebook, NASDAQ: META), a technology giant with a market value of nearly 800 billion US dollars, released the latest AI large language model Llama 2 series, including three parameter versions of 7 billion, 13 billion and 70 billion. The information and starting codes are all open source and free for commercial use. It also includes a version of the Llama 2-CHAT model fine-tuned for conversational chat scenarios.