One week of AIGC 丨 The world's largest AIGC acquisition was born, and China's large-scale model unicorn collapsed

Also on the AIGC track, MosaicML was acquired by Databricks for US$1.3 billion, a 5-fold increase in valuation compared to the previous round, creating the world’s largest generative AI acquisition so far. In China, the most dazzling large-scale model unicorn light years away was given to Meituan for US$285 million, a discount of 30% compared to the valuation two months ago. Time, fate, and luck.

Image source: Generated by Unbounded AI

Financing

Databricks acquires generative AI startup MosaicML for $1.3B

Databricks has agreed to acquire generative artificial intelligence startup MosaicML for $1.3 billion, Reuters reported on June 26. Databricks said the deal will combine its artificial intelligence technology with MosaicML’s language model platform, allowing companies to “maintain control of their valuable data in a simple and fast way without high costs.”

Meituan acquires all rights and interests beyond Light Years for 2.065 billion

On June 29, Meituan announced on the Hong Kong Stock Exchange that it had completed the acquisition of 100% of the equity of domestic and foreign entities light years away at a purchase price of approximately RMB 2.065 billion.

Artificial intelligence startup Inflection AI raises $1.3 billion, investors include Bill Gates and Nvidia

On June 30, artificial intelligence startup Inflection AI announced that it had raised $1.3 billion in a new round of financing led by Microsoft, LinkedIn co-founder Reid Hoffman, Bill Gates, former Google CEO Eric Schmidt, and new investor NVIDIA. To date, the company’s total funding has reached $1.525 billion.

Thomson Reuters acquires AI legal tech startup Casetext for $650M

Thomson Reuters has reached a definitive agreement to acquire Y Combinator-backed AI legal technology startup Casetext for $650 million in cash, TechCrunch reported on June 27. It is reported that the transaction is expected to be completed in the second half of 2023, subject to regulatory approval and compliance with customary closing conditions.

Celestial AI completes USD 100 million in Series B financing, led by IAG, Temasek, etc.

According to the “Kechuang Board Daily” report on June 29, the artificial intelligence accelerator company Celestial AI announced that it has completed a US$100 million Series B financing, led by IAG Capital Partners, Koch Disruptive Technologies and Temasek Xora Innovation Fund. The funding, which brings Celestial AI’s total financing to more than $165 million, will be used to support production of Celestial’s photonic platform by expanding the company’s engineering, sales and technical marketing departments, said CEO Lazovsky.

Yangpu Medical plans to raise no more than 156 million yuan for projects such as the AI-based fully automatic thromboelastography system

According to the daily economic news report, Yangpu Medical issued an announcement stating that it plans to raise no more than 156 million yuan for the AI-based fully automatic thromboelastography system project, molecular diagnostic blood collection tube construction project, marketing network construction project, supplementary working capital.

EvenUp completes $50.5 million in Series B financing, using AI to assist in the assessment of personal injury cases

According to 36 Krypton’s report on June 28, the legal industry start-up company EvenUp completed its B round of financing. This round was led by Bessemer Venture Partners, and Bain Capital Ventures, Clio and Behance founder Scott Belsky participated in the investment, with a total financing amount of 50.5 million US dollars . EvenUp uses artificial intelligence to help attorneys evaluate personal injury cases and make unbiased compensation recommendations. The platform consolidates raw case documents such as medical records, bills and police reports to help lawyers prepare claims. EvenUp can provide document generation including pre-litigation stage and litigation stage.

BentoML Receives $9M in Funding to Accelerate AI Application Development

According to a TechCrunch report on June 26, BentoML, an AI development framework created by Chaoyu Yang, an early software engineer at data giant Databricks, announced that it has received $9 million in seed round financing, led by DCM Venture Capital and participated by Bow Capital. Hurst Lin, general partner at DCM, joined BentoML’s board of directors following this funding round.

AI voice changer Voice.ai raises $6 million, and its real-time voice changer users are close to 500,000

According to a Techcrunch report on June 30, Voice.ai, an AI voice changer, has raised $6 million, led by Mucker Capital and M13. Voice.ai’s real-time voice changer has nearly 500,000 users. The funding is planned to be used to push its voice-changing technology into new territories.

Beilian Zhuguan completed the angel round financing of 50 million yuan, and its business covers the AI computing power market

According to ChinaVenture.com, recently, Zhejiang Beilian Zhuguan announced the completion of an angel round of financing of 50 million yuan, invested by Yuanjing Capital, Sequoia China Seed Fund and Zhouxuan Equity.

Beilian Zhuguan is an efficient cloud resource management service provider, committed to helping customers save IT resource costs through technical means. The business covers a number of key technical fields, including online business computing power (virtual machines required for business system operation, containers, databases, etc.), big data computing power (offline analysis and online analysis), and AI computing power, providing enterprise customers with solutions to reduce costs and increase efficiency.

AI Chiplet start-up Yuanli Semiconductor completed tens of millions of yuan in seed round financing

According to a report from Lieyun.com on June 27, AI Chiplet (chiplet) start-up company Yuanli Semiconductor announced that it has completed tens of millions of yuan in seed round financing, led by Inno Angel Fund, Zhongguancun Development Group, Qingke Venture Capital, Shuimu Tsinghua University Alumni Seed Fund, Zhongke Chuangxing and other institutions followed suit. Fang Shaoxia, founder and CEO of Yuanli Semiconductor, said that this round of financing will continue to be used for the formation of the company’s core team and innovative technology research and development.

The Hong Kong UGC launched a HK$100 million innovation and technology education fund to support teaching in the fields of Web3 and generative AI

According to a report by The Standard on June 28, the Education Funding Committee of the University of Hong Kong launched the Innovative Technology Education Fund (FITE) and allocated 100 million Hong Kong dollars to encourage eight funded universities to use innovative technology to promote teaching innovation and support universities in generative artificial intelligence, Experimental and exploratory efforts in the teaching of technology areas such as Web3 and digital competency to promote the role of related technologies in economic restructuring.

AI Startup Stability AI Issues Convertible Notes After Deep Funding Difficulty

According to a report by the Financial Associated Press, citing people familiar with the matter, Stability AI, an artificial intelligence start-up company, has sold convertible notes. The company is currently in deep financing difficulties. The convertible note deal, which raised just under $25 million, came this spring after Stability AI, the distributor of image generation tool Stable Diffusion, was said to be seeking to quadruple its valuation. Motez Bishara, a spokesman for Stability AI, denied the report and said the company was not having trouble raising money.

Entrepreneurship

Humane, an AI technology company founded by former Apple employees, released its first AI wearable device, which will be available this year

According to a report by IT House on July 2, AI technology company Humane is a company founded by former Apple employees. The company has been promoting a future vision with artificial intelligence as the core and beyond smartphones. The company today revealed its first product, the Humane Ai Pin, which officials say is a “clothing-based wearable device” that will be available later this year, powered by Qualcomm’s “advanced Snapdragon platform” and Use an array of sensors to enable “contextual and ambient computing interactions.”

Kai-Fu Lee formed a new company for the first time, and launched the 2023 DeeCamp AI talent project

According to 36 Krypton’s report on July 3, Li Kaifu, chairman and CEO of Sinovation Ventures, organized a new company for the first time, and named it Zero One Wanwu. The official website (01.AI) was launched simultaneously. According to reports, the headquarters of 01Wanwu is registered in Beijing, and the team currently has dozens of core members. The members of the Lianchuang team that have joined include the former vice president of Alibaba, the former vice president of Baidu, the former chief algorithm leader of Didi/Baidu, etc. The background of the algorithm and product team comes from major domestic and foreign companies, and the 2023 DeeCamp AI talent project is launched at the same time.

big model

JD.com will launch Yanxi AI large model with parameters reaching hundreds of billions

According to the “Kechuangban Daily” report on June 27, at the JD Cloud City Conference Shanghai Station, Cao Peng, chairman of the JD Group Technical Committee and president of the JD Cloud Division, introduced that the upcoming large-scale pre-trained language model of Yanxi is a A new generation model with hundreds of billions of parameters. JD.com’s big model will be oriented to multi-modality and go deep into industrial scenarios such as retail, logistics, and industry. Cao Peng emphasized that in the era of industrial intelligence, a new generation of digital infrastructure is also needed, which can widely support various chip architectures and is compatible with various heterogeneous infrastructures.

Hang Seng Electronics released the financial model LightGPT for the first time

According to the Financial Associated Press, on the afternoon of June 28, Hang Seng Electronics released LightGPT, a large model for the financial industry. This product will complete a new round of financial capability upgrades by the end of September, and officially open the trial interface.

Bai Shuo, Chief Scientist of Hundsun Electronics, said that LightGPT uses more than 400 billion tokens of financial field data and more than 40 billion tokens of language-enhanced data, and uses them as the secondary pre-training corpus of large models to support more than 80 + financial-specific task instructions Fine-tuning can provide underlying AI capability support for financial business scenarios such as investment advisory, customer service, investment research, operation, risk control, compliance, and R&D.

Power-Law Intelligence and Zhipu AI released a large legal vertical model with a parameter level of 100 billion

According to Blue Whale Finance and Economics, Powerlaw Intelligence and Zhipu AI released a legal vertical large model based on the Chinese 100 billion large model - PowerLawGLM. Compared with the “big and comprehensive” general large model, PowerLawGLM focuses on the legal subdivision field, and has unique advantages in the application effect of Chinese legal scenarios, and has rich legal knowledge and legal language understanding ability. At the same time, based on the capabilities of the PowerLawGLM large model, Power Law has created a legal dialogue product ChatMe, which is now officially launched, and the first batch of 50 places for internal testing is open.

China Telecom announced the TeleChat large model, claiming that it will establish a “tens of billions AI service provider”

According to a report from IT House on July 3, at the Artificial Intelligence Summit Forum of the 2023 Global Digital Economy Conference held yesterday, Zhang Xin, deputy general manager of China Telecom Group’s Digital Intelligence Technology Branch, announced the TeleChat model of China Telecom’s ChatGPT product.

According to reports, the model supports outputting codes, writing speeches, etc. The company’s research and development goal is to become a “tens of billions AI service provider”, and its products cover AI algorithms, platforms, applications, hardware, large models, etc. Other information has not been announced yet.

Shanghai AI Lab open sourced the world’s first medical basic model group

According to Jiemian News citing the official WeChat account of the Shanghai Artificial Intelligence Laboratory, on June 29, led by the Shanghai Artificial Intelligence Laboratory (Shanghai AI Laboratory), together with top scientific research institutions, universities and hospitals at home and abroad, jointly released the world’s first medical multi-mode State basic model group “OpenMEDLab Puyi”. “OpenMEDLab” will gradually open source in the near future, covering more than 10 medical data modalities, including medical images, medical texts, bioinformatics, and protein engineering, to promote breakthroughs in cross-field, cross-disease, and cross-modal scientific research based on medical basic models. At the same time, it helps to solve the long-tail problem in the medical field, and promotes the industrialization of medical large models.

H3C, a subsidiary of Tsinghua Unigroup, released a large-scale private domain model “Baiye Lingxi” LinSeer

According to news from Tsinghua Unigroup’s official Weibo account on June 28, at the 2023 NAVIGATE Navigator Summit, New H3C Group, a subsidiary of Tsinghua Unigroup, released an AI server equipped with NVIDIA H800 GPU and a “sharp tool” for AIGC intelligent computing network specially designed for large-scale model training. "800G CPO silicon photonics data center switch, the “optimal solution” Aofei computing power platform for multiple heterogeneous computing power scheduling, and the industry’s first private domain large model - “Baiye Lingxi” LinSeer. At present, the intelligent computing flagship H3 C UniServer R5500 G6 server equipped with NVIDIA H800 GPU has entered the order delivery stage, and the AI server equipped with NVIDIA A800 GPU has been delivered to customers in the second half of last year.

The GPT Industry Alliance was established, and the 360 Intellectual Brain API platform was officially opened

According to the Shanghai Securities News, on June 28, Zhou Hongyi, the founder of 360 Group, announced the 360 enterprise-level AI large-scale model strategy at the GPT Industry Alliance Inaugural Conference, and announced that the 360 Smart Brain API platform was officially opened, and will be the first to serve 20 industries. provide the solution. The GPT Industry Alliance is jointly initiated by the Internet Society of China, Yunda Institute of China Academy of Information and Communications Technology, and 360 Group.

application

OpenAI plans to make ChatGPT a “super intelligent personal work assistant” to compete with Microsoft

According to The Information, OpenAI CEO Sam Altman told investors privately that he plans to make ChatGPT a “super-intelligent personal work assistant” that can draft emails or documents for users based on their personal style and the latest business information.

But the assistant could create conflicts of interest with customers like Microsoft and Salesforce, which is also a major OpenAI partner, investor and cloud service provider, who also want to use OpenAI’s software to build artificial intelligence “copilots” (Copilots) to Auxiliary work. But building the new ChatGPT feature will be the focus of OpenAI’s commercialization efforts, according to Altman and two other people familiar with the matter.

Qunhe Technology launches 3D cloud design AIGC application “Kujiale AI”

According to 36 Krypton, on June 29, the cloud design software platform and SaaS service provider “Qunhe Technology” released the generative artificial intelligence application “Kujiale AI”. This is an AIGC application developed by the 3D Cloud Design Track based on the self-developed AI model.

“Kujiale AI” is mainly launched for home design scenes, and is created in the form of “pictures” and “wenshengtu”. At present, “Kujiale AI” has opened the entrance of PC terminal and small program terminal, and its AI capabilities will be embedded in more subdivided industries of Kujiale. It is understood that Qunhe Technology will launch AIGC products for all spaces in the future.

“SenseAvatar” digital human video generation platform is online

According to the news from IT House on June 30, the “SenseAvatar” digital human video generation platform was officially launched and fully landed on the iOS/Android application market. It is the industry’s first AIGC full-stack video generation platform.

With the support of SenseTime’s “SenseNova” large-scale model system and the large-scale AI device SenseCore, the Ruying platform officially claims that the similarity with real people reaches 90%-95%, and the matching degree of mouth shapes reaches 95%. The manual processing time has been greatly reduced by 90%, the model training time has been reduced by 60%, and various customized digital human training can be completed in just one day.

Entrepreneurial dark horse Niu Wenwen: The dark horse Tianqi science and technology innovation industry model has been used by more than 500 institutions

According to the Securities Times report, Niu Wenwen, chairman of the entrepreneurial dark horse, introduced that since its launch in May, the dark horse Tianqi technology innovation industry model has been used by more than 500 institutions and generated revenue. At present, Entrepreneurial Dark Horse is building the first domestic AIGC promotion and service platform for small and medium-sized enterprises. Niu Wenwen believes that every industry in China can be redone with an industry model. As long as there are accumulated service providers in the industry and the support of giants, small and medium-sized enterprises can train a unique industry with only an investment of one million RMB. Model.

A generative AI model with over one billion parameters can run on Android phones

According to the “Kechuang Board Daily” report on June 28, Meng Pu, chairman of Qualcomm China, demonstrated the generative Al model Stable Diffusion running on Android phones at the BOE Innovation Partner Conference today. This means generative AI models with over a billion parameters running on smartphones.

Radio Hong Kong launches AI weather anchor Aida, whose facial expressions and movements are synthesized by computer

According to IT Home news, on June 26, Aida, the AI weather anchor of Radio Television Hong Kong, officially debuted, reporting local and global weather conditions in the radio station’s “Radio Hong Kong Weather Briefing”. Radio Television Hong Kong stated on its official Facebook page that the AI anchor “will never eat screws (i.e. stuck, forget words, slurred speech)”.

According to relevant sources of the station, Aida is one of the experimental projects of AI applications in Taiwan. It took about 3 months to prepare and was produced by AI virtual avatar technology. In addition, Aida’s facial expressions, voice, mouth shape and movements are all completely synthesized by computer without the intervention of real people.

big company

Google DeepMind is developing a new AI model Gemini, which claims to be “stronger than GPT-4”

According to news from IT House on June 28, at the Google I/O Developer Conference last month, Google revealed for the first time the large-scale language model Gemini it is developing. According to Wired, Demis Hassabis, co-founder of DeepMind and CEO of Google DeepMind, further revealed the details of Gemini in an interview recently: the system combines the technology behind AlphaGo with a large language model, and the goal is to give the system new capabilities , such as planning or problem solving, are more capable than OpenAI’s GPT-4.

Microsoft built a small LLM AI model with 1.3 billion parameters, and said that the actual effect is better than GPT-3.5 with 100 billion parameters

IT House issued an article today stating that the blind heap volume of the AI model is actually not necessarily better. It depends more on the quality of the training data. Microsoft recently released a language model phi-1 with 1.3 billion parameters, using “textbook Level” high-quality data set training, it is said that “the actual effect is better than GPT 3.5 with 100 billion parameters”. The model is based on the Transformer architecture, and the Microsoft team used “textbook-grade” data from the web and “logically rigorous content” processed with GPT-3.5, as well as eight Nvidia A100 GPUs, to complete the training in just 4 days .

Oracle will spend billions this year to buy Nvidia GPUs, betting big on AI

Oracle Corp. is spending “billions of dollars” on chips from Nvidia Corp. to expand cloud computing for a new wave of artificial intelligence, its founder and chairman Larry Ellison said on Wednesday, Reuters reported on June 29. Serve.

Meta will improve the transparency of the application of AI algorithms, providing users with detailed instructions and more choices

According to news from IT House on June 30, Meta recently announced that it will launch a series of new initiatives to allow users to better understand and control the content they see on Facebook and Instagram. According to Meta, the company will share more information on how it uses AI to impact user experience. Specifically, Meta has rolled out 22 different instructional cards that explain to users exactly how the company uses AI to control some of the content displayed on Instagram and Facebook, such as the “Discover” page and suggestions to follow users. Meta also said it would expand explanations to features like “Why am I seeing this article” in the coming weeks.

View

Niu Wenwen: In fact, only one million yuan is needed to create their own industry model for small and medium-sized enterprises

On the afternoon of June 28th, the “Open Collaborative Development - GPT Industry Alliance Founding Conference” jointly sponsored by the China Internet Society, China Academy of Information and Communication Research Institute, and 360 Group was held in Beijing. Niu Wenwen, chairman of the entrepreneurial dark horse, attended the meeting and made a spoke. Niu Wenwen said in his speech that since its launch in May this year, the application of the dark horse Tianqi Science and Technology Innovation Industry Model has been developing very well. More than 500 institutions have used it, and a certain amount of revenue has been generated. He believes that small and medium-sized enterprises often think that large-scale model training requires an investment of tens of millions of dollars. In fact, only an investment of one million yuan can help small and medium-sized enterprises train a unique industry model. But this requires two prerequisites, “First, you are an experienced service provider in the industry; second, you can get the support of giants like 360.”

Gary Gensler: AI is changing the economic environment, and the SEC will make rulemaking recommendations for its potential negative impact

On June 28, Gary Gensler, chairman of the US SEC, tweeted: “Predictive data analysis and AI are changing a large part of our economy, and finance is no exception. Artificial intelligence has been used in call centers, account opening, compliance programs, trading Algorithms and sentiment analysis, etc. AI is driving rapid changes in robo-advisor and brokerage applications. When predictive data analysis and the algorithms behind these applications are optimized for investor interests, it can benefit market access, efficiency and returns. But It may also lead to potential conflicts. In addition, the underlying data used in these analytical models may be based on data that reflects historical biases, with implications for fair access and prices in the market. SEC staff will address potential rulemaking on these matters make a suggestion.”

Turing Award winner Yang Likun bombarded ChatGPT: Without a real understanding of the real world, no one will use it within five years

According to The Paper, on June 22 local time, Turing Award winner Yang Likun appeared at the Munk Debate. He said that every technology comes with risks, but AI will succumb to humans and serve humans. “ChatGPT doesn’t really understand the real world because they’re all pure text training, and most of human knowledge has nothing to do with text or language.” He also predicted, "In five years, no one will use ChatGPT.

More than 150 executives criticize the “AI Act”: it may damage European competitiveness

Dozens of Europe’s largest companies, including Airbus and Siemens, took collective action on June 30 to publicly oppose the proposed EU artificial intelligence (AI) bill, criticizing that the EU’s AI The bill could harm Europe’s competitiveness and fails to address potential challenges. More than 150 business executives signed the open letter, including Airbus chairman René Obermann and Yann LeCun, one of the “Godfathers of AI,” as well as top executives from companies as diverse as Spanish telecom company Cellnex, French software company Mirakl and German investment bank Berenberg. Tube.

Research Report

The demand for artificial intelligence servers has increased sharply, and the market size of China’s AI servers may reach US$56.1 billion in 2025

According to a CCTV Finance report on June 27, the emergence of artificial intelligence models has set off a new round of industry boom. Since the beginning of this year, many domestic Internet companies have released artificial intelligence large-scale model products. Affected by this, the demand for artificial intelligence servers, which are the basis of computing power, has recently surged. The reporter learned that major technology companies around the world are accelerating the deployment of large AI models. According to the data, at present, 79 large-scale models with a parameter scale of more than 1 billion have been released in China, ranking second in the world. Compared with general-purpose servers, AI servers, especially training servers, can quickly and accurately process large amounts of data, so the market demand has increased sharply. The agency predicts that the global AI server market will maintain rapid growth in the next three years. From 2023 to 2025, China’s AI server market is expected to reach 13.4 billion, 30.7 billion, and 56.1 billion US dollars, a year-on-year increase of 101%, 128%, and 83% respectively.

GitHub: AI-powered generative development tools will add $1.5 trillion to global GDP by 2030

According to a TechCrunch report on June 28, GitHub CEO Thomas Dohmke said in a speech at the Collision conference in Toronto, Canada today: “In the future of our lives, artificial intelligence and software development are inseparable.” Given the explosive growth and shortage of developers in the field, GitHub also points out that generative artificial intelligence tools promise to improve the productivity of these developers. The firm estimates that AI-powered generative development tools will add $1.5 trillion to global GDP globally by 2030, with each missing skilled developer costing $100,000 in lost GDP. Meanwhile, GitHub believes that generative AI developer tools could complement approximately 15 million additional developers (thus a total impact of $1.5 trillion). The company believes this is a conservative estimate.

2023 Global AI Index Ranking Announced: The U.S. and China are firmly in the top two, and Asian countries are performing well

According to a report by IT House on June 29, the British media organization Tortoise Media recently released the 2023 global AI index ranking. The data shows that the successful launch of ChatGPT in 2022 has triggered a surge in global generative AI development, while Asian countries perform more prominently in the list.

The company used investment, innovation and practice to determine the ranking of countries in AI development. The United States ranked first in all three indicators and scored a perfect score of 100. This benefited from the high score obtained in business investment, that is, the activity of start-up companies. China ranked second with 62 points. Since the list was first released in 2020, the United States and China have consistently ranked in the top two.

Morgan Stanley: Artificial intelligence brings a $30 billion opportunity to the network security market

Artificial intelligence (AI) could help unlock more than $30 billion worth of cash for the cybersecurity market amid rising cybersecurity threats and severe manpower shortages, Morgan Stanley said on June 28. opportunity, Crowdstrike, Microsoft and Palo Alto Networks are expected to be major beneficiaries. Analysts led by Hamza Fodderwala said in a report that in addition to rising threats and labor shortages, generative AI is being used to create and mutate malware at an unprecedented rate, which “significantly increases the The need for security organizations to invest in automation”.

The survey found that 20% of men use AI tools such as ChatGPT to “beautify” their dating profiles and improve their communication skills

According to news from IT House on June 29, the Attractiontruth agency recently surveyed 1,371 men and found that 20% of the respondents (about 274 men between the ages of 25 and 35) are using AI tools such as ChatGPT to improve their dating profiles , tailor relevant information to attract more dates. Among the men who use AI tools to “beautify”, 37% think it can attract more attention from the opposite sex; 24% think it can improve their information delivery skills.