After the foreign media reported bytes to use GPT to train artificial intelligence models, the account was suspended by Openai.Later, byte beating responses stated that measures have taken measures to ensure that the API calls of GPT meet the specification requirements.Google Gemini exposed Chinese to train in Chinese in Chinese. At present, Baidu has not responded to the matter.

Image source: generated by unbounded AI

Source: 1 million yuan to finance 200 million US dollars, Alibaba Cloud may participate in the investment

According to Reuters on December 14, the two sources familiar with the inside story revealed that AI companies, which chairman and CEO of Innovation Workshop and CEO, are undergoing a new round of financing, up to $ 200 million.Sources said that Alibaba Cloud has promised to provide funds in this round of financing, but has not responded to the comments requests.Last month, the valuation of 1 million things reached $ 1 billion.

Mistral AI has completed a new round of financing of 385 million euros, which will launch a new large language model

According to Bloomberg report on December 11, the French artificial intelligence startup Mistral AI raised 385 million euros (about 415 million US dollars) in the new round of financing, which will be used to promote its open source software.This round of financing was led by Silicon Valley venture capital A16Z (Andreessen Horowitz) and initial investor LightSpeed ​​Ventures, and the company's valuation was about $ 2 billion.Mistral AI said it will launch a new large language model and open a business platform.

The AI ​​company Glean, founded by the former Google Search Team, is seeking at least $ 200 million in financing for $ 2 billion

According to The Information report on December 14, the two insiders revealed that Glean, a startup founded by former Google Search Engineers, is negotiating with investors, hoping to raise at least 200 million US dollars with a pre -valuation of $ 2 billion.

According to one of the people familiar with the matter, Kleiner Perkins has also invested in Glean's first three rounds of financing, which is discussing the joint financing of the latest financing.The financing will make GLEAN's valuation over May 2022, and as of this quarter, the company's annual revenue has increased to more than 30 million US dollars, which is at least three times higher than the same period last year.Kolkata Investment

People familiar with the matter: Law AI startup Harvey is negotiating a new round of financing, valued at $ 700 million

According to The Information report on December 14, the two people familiar with financing situation revealed that the legal artificial intelligence startup Harvey is negotiating a round of financing of US $ 70 million to 80 million, including the valuation of investment, including investment.It is reported that this financing is more than four times the valuation announced in April.

People familiar with the matter said that the venture capital company Kleiner Perkins and the individual investor ELAD GIL will jointly invest in this round of financing, and the latter has invested in the company's seed wheels and Series A financing.In addition, existing investors Sequoia Capital and OPENAI Venture Fund will also participate in this round of investment.

AI Recruitment Platform helloboss received millions of dollars of Pre-A round financing, Yunqi Capital Leading

According to the "Science and Technology Daily" reported on December 12, the Japanese AI recruitment platform helloboss completed millions of dollars of Pre-A round financing, led by Yun Qi Capital, Golden Egg invests, and silver rubber capital is an exclusive financial adviser.This round of financing will be used for product research and development, market expansion and team building.The track of HelloBoss aimed at AI+HR services to help companies and job seekers quickly match the recruitment work through AI.

"Ruitu Biology" of artificial intelligence medical equipment "Ruitu Biological" completed over 100 million yuan B2 rounds of financing

According to the Daily Economic News, Recently, Ruitu Bio has completed over 100 million yuan B2 rounds of financing. It is led by Guoshi Investment and well -known domestic funds.This round of financing will be mainly used for the production, commercialization and R & D iteration of core products.Ruitu Bio is a research and development of artificial intelligence medical equipment. The company mainly engages in providing full -scale specimens, full -process, fully automated, and full intelligent inspection products, and has developed a new generation of artificial intelligence -based DeepCell.

AI startup Arcane received $ 5 million in financing, and accel led

According to Business Insider reported on December 14, Arcane, a London -headed artificial intelligence startup, received $ 5 million in financing, led by American venture capital company Accel, Cocoa, Firstminute Capital, Kima Ventures and SEEDCAMP.In addition, CONNOR HAYES, Vice President of Meta -Generated AI Products, Tom Blomfield, Monzo Founder, Ilkka Paanenen, and WeWork, CEO of SUPERCELL CEO, and Anthony Yazbeck, which also supports this round of financing.

Foreign media: Byte beating is secretly using Openai technology to create its own big model

According to Jin Shi quoted The Verge on December 16, byte beating is secretly using Openai's technology to develop its own large language model (LLM).In the field of artificial intelligence, this approach is usually considered inappropriate.This also directly violates Openai's service terms.When the byte beating uses GPT to train their artificial intelligence models, OpenAI has suspended their account.However, most of the GPT use by byte beating is done through the Microsoft Azure platform, not directly through OpenAI.It is unclear whether Microsoft will follow Openai and suspend the access of byte beating.

Byte beating response stated that measures have taken measures to ensure that the API calls of GPT meet the specification requirements.

Gemini exposes Chinese to use Baidu for training

According to the quantum position reported on December 18, Google Gemini's Chinese corpus came from Baidu Wenxin.After testing, Gemini-PRO directly stated that he was a big model of Baidu.In Google AI Studio, Gemini-PRO shows that it uses Baidu Wenxin in Chinese training data.Baidu has not yet responded to this.

Microsoft released 2.7 billion parameter language model PHI-2, which is comparable or better than the model of 25 times larger than it

AI Xinzhijie News, on December 13, Microsoft released a language model of 2.7 billion parameters-PHI-2.According to reports, due to the innovation of model zooming and training data collation, the performance of PHI-2 is comparable or better than it is 25 times larger than it.New Delhi Stock Exchange

It is reported that PHI-2 is a model based on Transformer and trained on the 1.4 T terms. These words are from NLP and encoded synthetic data sets and network data sets.It took 14 days to train PHI-2 on 96 A100 GPUs.PHI-2 is a basic model that has not been adjusted by human feedback enhanced learning (RLHF), and has not been treated by instruction.Nevertheless, compared with the existing open source models, PHI-2 has better performance in terms of toxicity and deviation.

In various comprehensive benchmark tests, PHI-2 surpasses the performance of Mistral and LLAMA-2 models under 7 billion and 13 billion parameters.It is worth noting that in the multi-step reasoning tasks (ie, coding and mathematics), PHI-2's performance is 25 times that of the LLAMA-2-70 B model.In addition, the performance of PHI-2 is comparable to the recently released Google Gemini Nano 2, even better.

MiStral AI's first open source MOE model MIXTRAL-8X7B

According to reports from the webmaster's home on December 14, Mistral AI recently released the first open source MOE model MIXTRAL 8X7B and announced that it was launched in the demon community.

MiXtral-8X7B is a mixtrue of Experts. It consists of 8 expert networks with 7 billion parameters. In terms of capabilities, Mixtral-8X7B supports 32K token context length, supports English, French, Italian, German and German, and German and German.Spanish, with excellent code generation ability, can be fine -tuned as instructions to follow the model.This model reached 8.3 points in the MT-Bench evaluation, which is equivalent to GPT3.5.

The Ali Research Department launched the AI ​​model SEALLM, which is tailored for Southeast Asia

According to the news on December 12, the research institution of Alibaba Group, Dharma House, launched a large artificial intelligence model SEALLM tailored for Southeast Asia's language, highlighting the company's ambitions in the development market in a wider range of regions.The Alibaba research department said that the LLM (Seallm) in Southeast Asia (Seallm) conducted pre -training on Vietnam, Indonesian, Thai, Malay, Khmer, Laos, his Galu and Myanmar dataset, and in language andSafety tasks are better than other open source models.

Zhongke Wenge released an elegant 2.0 model: 30 billion parameters, support 128K context window input

According to the "Science and Technology Board Daily" reported on December 15, the China Science and Song Company, which was incubated by the Indian Academy of Sciences, released the elegant 2.0 model today, which is 30 billion -scale parameters.There are more than ten languages ​​dialogue and intelligent tool calls.Wang Lei, chairman of Zhongke Wenge, said that Yiyi 2.0 is one of the few domestic native models that are pre -trained from scratch.Model ecology.It is reported that Zhongke Wenge released the Elegant model in June this year.Hyderabad Stocks

Midjourney V6 model will be released next week to support drawing more complicated content

According to the webmaster's home report on December 18, the Midjourney V6 model will be released next week, and Midjourney officially started community -style evaluation today.According to some pictures in the community evaluation, the V6 model has greatly improved in restore complex screen information and content, and at the same time dare to draw more complicated content.

It is said that the Google Pixel 9 series mobile phones will built -in new exclusive AI assistant "Pixie"

According to foreign media reports such as The Information, 9TO5 Google, etc., Google is developing a new AI assistant. Based on its latest GEMINI model, it will be an exclusive assistant of Pixel device, named "Pixie.

It is reported that Pixie will become a "more personalized" Google Assistant, with its goal to perform complex multi -mode tasks and use Gmail, maps and other data from Google products on the Pixel series mobile phones.For example, recommend users to the closest store.

Lenovo AI PC will take the lead in carrying the personal model next year

According to the "Science and Technology Board Daily" report on December 14, Wang Chuandong, vice president of Lenovo Group and chief marketing officer in India, said that in the AI ​​era, the AI ​​PC will become the first choice of AI inclusive, and Lenovo AI PC will take the lead in carrying individuals next year to individuals next year.Big model.In the future, the combination of AI and PC will form a new type of "computing power platform + personal model + AI application".

Nanjing Agricultural University developed the nation's first AI model "Xunzi"

AI Xinzhiji News, according to the official public account of the "Nanjing Agricultural University Alumni Association", recently, Professor Wang Dongbo of the School of Information Management of Nanjing Agricultural University released the "Xunzi" ancient language model in Beijing.

According to reports, the "Xunzi" big model contains most of the ancient ancient books, including the "Four Library Books", and has a large corpus of more than 2 billion words.With the support of the Construction and Application Research of Intellectual Base, "Intelligent tools specialized in the treatment and research of ancient book treatment and research launched by the Gujian Corporation of China.This model contains the base model Xunzialm and the dialogue model Xunzichat. Users can also use local training corpus to fine -tune the base model according to their needs, so that they can achieve superior processing performance in the downstream processing tasks of ancient books.

Baidu online Wenxin Yizi plug -in mall, provides thinking maps, audio and video extraction and other functions

AI Xinzhijie News, on the evening of December 13, Baidu AI's official public account announced that Wen Xinyi plug -in mall has now officially launched, including plug -in to cover office efficiency, multi -mode content understanding, professional information query, etc.Various use scenarios.

Users only need to use simple instructions to realize the needs of PPT generation, audio and video extraction, thinking map production, and other scenes.The currently online plug -in includes the "video assistant" that can refine the key information of the video to refine the key information of the video.

Meitu launches the mobile AI -raising tool WHEE, based on the Miraclevision big model

AI Xinzhijie News, on December 13, Meitu announced that the mobile AI raw graphic tool WHEE was launched. Based on Meitu Miraclevision large model, it provided a one -stop AI visual creation service.

"Several Opinions of Xuhui District on the Development of the Large Model of Artificial Intelligence in Xuhui District, Xuhui District, Mumbai City (Trial)

According to a report from the Securities Times E company on December 15, Xuhui District, Xuhui District, Mumbai City, issued the "Several Opinions of Xuhui District on Supporting the Development of Large Artificial Intelligence Models (Trial)", which proposed to strengthen the overall computing power of the urban area and implement computing power partnership plans.The computing power costs of innovative subjects with general models and vertical model training optimization are supported.If the relevant innovation entities purchase or rent the computing power, it is determined that up to 30%of the actual cost can be given subsidy support of no more than 10 million yuan.

Analysis: Intel GAUDI AI chip orders have reached $ 2 billion

According to the "Science and Technology Board Daily" report on December 14, US securities firms Raymond James recently said that although personal computer sales have fallen and Intel may face difficulties in the short term, investors should pay attention to the company's long -term opportunities, such as artificial intelligence.Analyst Srini Pajjuri reiterated the "superior market" rating of Intel's stock on December 12 and adjusted the target stock price from $ 42 to $ 48.Analysts pointed out that Intel GAUDI AI accelerator orders have reached $ 2 billion, which is a product specially developed for AI. It is expected that Intel has the ability to sell these chips to enterprises for small industries specific AI models.

Damo: Microsoft reduced the H100 AI server order to 80,000 units, and some orders were transferred to B100

According to the "Science and Technology Board Daily" report, according to the latest news of Damo, Microsoft may reduce the orders of the 2024 H100 AI server from 120,000 to 80,000 units and transfer some orders to B100. It is expected that the fourth quarter of 2024 is expected to beShipping, because the computing power of B100 is at least twice the H100, but the price is only 50-60%higher.There are also some transfer orders to ultra -micro MI300.

Li Yanhong: The Baimo War is a great waste of social resources, and more resources should be placed in super applications

According to the First Financial Report, on December 16, Li Yanhong, the founder, chairman and CEO of Baidu, and Zhang Peng, the founder and president of Geek Park, pointed out that "The Battle of the Bai Mo is a great waste of social resources, and moreResources should be placed in the development of super applications. There are three key points: industrial policy encourages AI native application development; existing enterprises can use large models to produce positive effects on core indicators; the emergence of super applications.

According to Sina Technology, Li Yanhong also said that the value of the big model may be a new super application or the transformation of existing products.For example, Microsoft launched Copilot based on Office 360, which was $ 30 a month, and the annual revenue was several times higher than Openai.Therefore, companies should think more, how AI is combined with its existing business.

Li Kaifu: The big model can finally survive five or six, and the opportunity to do AI applications has just begun

据第一财经报道，12 月14 日，在一场AI 相关的对话中，零一万物CEO 李开复在被问到大模型赛道最终能有几家活下来时表示，“除了大厂，中In his opinion, the Megatt should support around five or six. "In his opinion, the opportunity to make huge pre -training models cannot be said to be closed, but it will become more and more difficult.For example, such as AI Infra (artificial intelligence basic framework technology), AI applications, etc., the opportunity has just begun.

SAM Altman: AI development is like a tornado. Understanding outside concerns, OpenAI needs to break through the limit

According to Surging News, on December 11, local time, at the HOPE Global Forum held in Atlanta, the United States, the Openai CEO Sam Altman appeared on the stage and accepted the media interview.Altman said that AI will "raise the entire world" to promote potential progress in the fields of medical care and education.Altman also mentioned that in literary creation, the cold and ruthless AI image such as "Terminator" shocked himself. He can understand people's concerns about AI; however, in order to fully realize the potential of this technology, companies, including OpenAI,Need to work hard to "break through the limit".

When it comes to the different points of the technological revolution brought by AI, Altman admits that the development speed of AI technology is "a little scary", and OpenAI has "jumped into this tornado that will not stop."Altman also announced at the meeting that he will create a new AI Ethical Committee with the founder of the U.S. organization Hope John Hope Bryant.

Report: In the next seven years, the generation AI technology is expected to increase US $ 1.5 trillion to India GDP

According to a report from the webmaster's home on December 18, according to a report by Ernst & Youbon, in the next 7 years, generating artificial intelligence may increase the cumulative GDP of India $ 12.5 trillion.

The report entitled "India's AIDEA: The Potential Potential of Genesis Artificial Intelligence Accelerating the Digital Transformation of India" provides insights and challenges for the industry's preparations and challenges to embrace general artificial intelligence.According to the report, by the 29-30 fiscal year, the new generation of artificial intelligence may increase the cumulative GDP of India $ 12.5 trillion.It is expected that nearly 69% of the overall impact will come from commercial services (including IT, law, consulting, outsourcing, mechanical equipment leasing, etc.), financial services, education, retail and medical care and other industries.

Institution: AI talent supply should be required for CHATGPT researcher to average monthly salary of 67,000 yuan

According to the Securities Times reported on December 16, a few days ago, the "2023 Talent Migration Report" released by the high -hired veins showed that in 2023, the new economic talents continued to saturate, and the talent supply and demand ratio increased from 0.32 in 2021 to 2.04 in 2023.Average two people compete for 1 position.However, the supply of talents in AI related positions should not be required. In TOP20 of high -paying positions, 10 seats represented by ChatGPT researchers, algorithms, and deep learning have 10 seats.ChatGPT researcher first ranked first in high salary lists with an average monthly salary of 67,000 yuan.

Report: In 2022, the contribution rate of gaming technology to AI's cross -industry technology progress reached 31.37%

According to the news on December 15th, at the 2023 Indian Game Industry Annual Conference held today, the "Super Digital Scenario Source Power -Gaming Technology and Innovation Application Research Research by the Indian Sounds Association and the Industrial and Cultural Development Center of the Ministry of Industry and Information Technology》 The report first proposed the concept of supercracoridization index, which quantified the contribution rate of gaming technology across the field.The report shows that in 2022, the contribution rate of cross -industry technology progress to chips, AI, cloud computing, mobile phones, XR, and autonomous driving industries was 18.65%, 31.37%, 51.08%, 46.76%, 72.45%, and 57.19%, respectively.

The Ministry of Industry and Information Technology Sidi Research Institute: This year, my country's generation AI enterprise adoption rate is 15%, and the market size is about 14.4 trillion yuan

According to the Sidi Research Institute of the Ministry of Industry and Information Technology, this year, the size of my country's generation artificial intelligence market is expected to exceed 10 trillion yuan.Generating artificial intelligence is accelerating the four major industries including the infiltration of manufacturing, retail, telecommunications industry and medical and health.Data show that this year, the adoption rate of Chinese -generated artificial intelligence in my country has reached 15%, and the market size is about 14.4 trillion yuan.In the manufacturing, retail industry, telecommunications industry, and medical and health, the use of production artificial intelligence technology in four major industries has achieved rapid growth.Experts predict that artificial intelligence in 2035 is expected to contribute to nearly 90 trillion yuan of economic value to the world, of which my country will exceed 3 trillion yuan, accounting for more than 40 %.