link1s.site

Stanford AI project team apologizes for plagiarizing Chinese model

An artificial intelligence (AI) team at Stanford University apologized for plagiarizing a large language model (LLM) from a Chinese AI company, which became a trending topic on the Chinese social media platforms, where it sparked concern among netizens on Tuesday.

We apologize to the authors of MiniCPM [the AI model developed by a Chinese company] for any inconvenience that we caused for not doing the full diligence to verify and peer review the novelty of this work, the multimodal AI model Llama3-V's developers wrote in a post on social platform X.

The apology came after the team from Stanford University announced Llama3-V on May 29, claiming it had comparable performance to GPT4-V and other models with the capability to train for less than $500.

According to media reports, the announcement published by one of the team members quickly received more than 300,000 views.

However, some netizens from X found and listed evidence of how the Llama3-V project code was reformatted and similar to MiniCPM-Llama3-V 2.5, an LLM developed by a Chinese technology company, ModelBest, and Tsinghua University.

Two team members, Aksh Garg and Siddharth Sharma, reposted a netizen's query and apologized on Monday, while claiming that their role was to promote the model on Medium and X (formerly Twitter), and that they had been unable to contact the member who wrote the code for the project.

They looked at recent papers to validate the novelty of the work but had not been informed of or were aware of any of the work by Open Lab for Big Model Base, which was founded by the Natural Language Processing Lab at Tsinghua University and ModelBest, according to their responses. They noted that they have taken all references to Llama3-V down in respect to the original work.

In response, Liu Zhiyuan, chief scientist at ModelBest, spoke out on the Chinese social media platform Zhihu, saying that the Llama3-V team failed to comply with open-source protocols for respecting and honoring the achievements of previous researchers, thus seriously undermining the cornerstone of open-source sharing.

According to a screenshot leaked online, Li Dahai, CEO of ModelBest, also made a post on his WeChat moment, saying that the two models were verified to have highly similarity in terms of providing answers and even the same errors, and that some relevant data had not yet been released to the public.

He said the team hopes that their work will receive more attention and recognition, but not in this way. He also called for an open, cooperative and trusting community environment.

Director of the Stanford Artificial Intelligence Laboratory Christopher Manning also responded to Garg's explanation on Sunday, commenting "How not to own your mistakes!" on X.

As the incident became a trending topic on Sina Weibo, Chinese netizens commented that academic research should be factual, but the incident also proves that the technology development in China is progressing.

Global Times

UAE insurance sector continued to grow in Q4-23: CBUAE
The UAE insurance sector continued to grow in Q4-2023, as reflected by increase in the gross written premiums. As of year-end, the number of licensed insurance companies in the UAE remained at 60, according to the Central Bank of the UAE's (CBUAE) Quarterly Economic Review (Q4-2023). The insurance sector comprised 23 traditional national companies, 10 Takaful national and 27 foreign companies, while the number of insurance related professions remained at 491. The review on insurance sector structure and activity showed that the gross written premium increased by 12.7% Y-o-Y in Q4 2023 to AED 53.2 billion, mostly due to an increase in health insurance premiums by 16.5% Y-o-Y and an increase in property and liability insurance premiums by 18.9% Y-o-Y, while the insurance of persons and fund accumulation premiums decreased by 12.4% Y-o-Y, resulting primarily from decrease in individual life premiums. Gross paid claims of all types of insurance plans increased by 12.8% Y-o-Y to AED 31.1 billion at the end of 2023. This was mainly driven by the increase in claims paid in health insurance by 16.9% Y-o-Y and increase in paid claims in property and liability insurance by 10.9% Y-o-Y, partially offset by the decline in claims paid in insurance of persons and fund accumulation by 2.8% Y-o-Y. The total technical provisions of all types of insurance increased by 8.4% Y-o-Y to AED 74.4 billion in Q4 2023 compared to AED68.6 billion in Q4 2022. The volume of invested assets in the insurance sector amounted to AED 76 billion (60.4% of total assets) in Q4 2023 compared to AED 71.4 billion (59.4% of total assets) in Q4 2022. The retention ratio of written insurance premiums for all types of insurance was 52.9 % (AED 28.1 billion) in Q4 2023, compared to 54.9% (AED 25.9 billion) at the end of 2022. The UAE insurance sector remained well capitalized in terms of early warning ratios and risk assessment. Own funds to minimum capital requirement ratio increased to 335.7% in Q4 2023, compared to 309.3% at the end of 2022, due to an increase in own funds eligible to meet the minimum capital requirements. Also, own funds to solvency capital requirement ratio rose to 221% in Q4 2023 compared to 208.5% in Q4 2022, due to an increase in own funds eligible to meet solvency capital requirements. Finally, own funds to minimum guarantee fund ratio reached to 316.3% at the end of 2023 down from 314.6% a year earlier, due to higher eligible funds to meet minimum guarantee funds. In terms of profitability, the net total profit to net written premiums increased to 6.5% in Q4 2023, compared to 2.9% at the end of 2022. The return on average assets increased to 0.3% in Q4 2023 compared to the 0.1% at the of the previous year.
Stanford AI project team apologizes for plagiarizing Chinese model
An artificial intelligence (AI) team at Stanford University apologized for plagiarizing a large language model (LLM) from a Chinese AI company, which became a trending topic on the Chinese social media platforms, where it sparked concern among netizens on Tuesday. We apologize to the authors of MiniCPM [the AI model developed by a Chinese company] for any inconvenience that we caused for not doing the full diligence to verify and peer review the novelty of this work, the multimodal AI model Llama3-V's developers wrote in a post on social platform X. The apology came after the team from Stanford University announced Llama3-V on May 29, claiming it had comparable performance to GPT4-V and other models with the capability to train for less than $500. According to media reports, the announcement published by one of the team members quickly received more than 300,000 views. However, some netizens from X found and listed evidence of how the Llama3-V project code was reformatted and similar to MiniCPM-Llama3-V 2.5, an LLM developed by a Chinese technology company, ModelBest, and Tsinghua University. Two team members, Aksh Garg and Siddharth Sharma, reposted a netizen's query and apologized on Monday, while claiming that their role was to promote the model on Medium and X (formerly Twitter), and that they had been unable to contact the member who wrote the code for the project. They looked at recent papers to validate the novelty of the work but had not been informed of or were aware of any of the work by Open Lab for Big Model Base, which was founded by the Natural Language Processing Lab at Tsinghua University and ModelBest, according to their responses. They noted that they have taken all references to Llama3-V down in respect to the original work. In response, Liu Zhiyuan, chief scientist at ModelBest, spoke out on the Chinese social media platform Zhihu, saying that the Llama3-V team failed to comply with open-source protocols for respecting and honoring the achievements of previous researchers, thus seriously undermining the cornerstone of open-source sharing. According to a screenshot leaked online, Li Dahai, CEO of ModelBest, also made a post on his WeChat moment, saying that the two models were verified to have highly similarity in terms of providing answers and even the same errors, and that some relevant data had not yet been released to the public. He said the team hopes that their work will receive more attention and recognition, but not in this way. He also called for an open, cooperative and trusting community environment. Director of the Stanford Artificial Intelligence Laboratory Christopher Manning also responded to Garg's explanation on Sunday, commenting "How not to own your mistakes!" on X. As the incident became a trending topic on Sina Weibo, Chinese netizens commented that academic research should be factual, but the incident also proves that the technology development in China is progressing. Global Times
Insurers fret over militant attacks, AI hacks at Paris Olympics
LONDON, July 5 (Reuters) - Insurers are nervous that militant attacks or AI-generated fake images could derail the Paris Olympics, risking event cancellations and millions of dollars in claims. Insurers faced losses after the 2020 Tokyo Olympics were postponed for a year due to the COVID-19 pandemic. Since then, wars in Ukraine and Gaza and a spate of elections this year, including in France, have driven up fears of politically-motivated violence at high-profile global events. The Olympics take place in Paris from July 26-Aug 11 and the Paralympics from Aug 28-Sept 8. German insurer Allianz (ALVG.DE), opens new tab is insurance partner for the Games. Other insurers, such as the Lloyd's of London (SOLYD.UL) market, are also providing cover. "We are all aware of the geopolitical situation the world is in," said Eike Buergel, head of Allianz's Olympic and Paralympic programme. "We are convinced that the IOC (International Olympic Committee), Paris 2024 and the national organising committees, together with the French authorities, are taking the right measures when it comes to challenges on the ground."
Amid rising regional tensions, the US announced that it will hold another Rim of the Pacific military exercise
The U.S. Navy's Pacific Fleet announced on Wednesday (May 22) that the 2024 Rim of the Pacific Exercise (RIMPAC 2024) is expected to take place on June 26, with 29 countries participating in and around the Hawaiian Islands, a larger lineup than the previous exercise in 2022. The Philippines, which has had multiple maritime conflicts with China recently, and Japan, which has tense diplomatic relations with China, will send troops to participate. China has been excluded from participating in the international military exercise since 2018, and its aggressive actions and reactions are causing tensions in the Pacific region to continue to rise. The biennial Rim of the Pacific military exercise is the world's largest international maritime exercise. The U.S. Navy said that the exercise will last until August 2, and it is expected to involve 29 countries, 40 surface ships, 3 submarines, 14 countries' army forces, more than 150 aircraft and more than 25,000 personnel. The U.S. Navy said that the theme of the 29th RIMPAC 2024 is "Partners: Integrated and Ready", emphasizing inclusiveness as the core, promoting multinational cooperation and trust, and using military interoperability to achieve their respective national goals to strengthen integrated and ready alliance partners. Its goal is to "enhance collective strength and promote a free and open Indo-Pacific region" through joint training and operations. The 29 countries participating in the exercise this year include Australia, Belgium, Brazil, Brunei, Canada, Chile, Colombia, Denmark, Ecuador, France, Germany, India, Indonesia, Israel, Italy, Japan, Malaysia, Mexico, the Netherlands, New Zealand, Peru, South Korea, the Philippines, Singapore, Sri Lanka, Thailand, Tonga, the United Kingdom and the United States. Compared with the 28th RIMPAC held in 2022, which involved 26 countries, 38 surface ships, 4 submarines, 9 countries' army forces, more than 170 aircraft, and about 25,000 officers and soldiers, the number of countries, ships and army forces participating in this exercise has increased. The countries participating in this year's RIMPAC military exercise include all members of the Quadrilateral Security Dialogue (QUAD) between the United States, Japan, India and Australia, and the Australia-UK-US Trilateral Security Partnership (AUKUS), as in the previous exercise. In addition, countries surrounding the South China Sea and the South Pacific island nation of Tonga are also participating. Many analysts believe that the military exercise itself is sending a message to China: China's expansion in the Western Pacific region will be blocked and defeated. The United States invited China to participate in the RIMPAC military exercise twice in 2014 and 2016. In 2018, due to China's expansion in the South China Sea, the United States withdrew its invitation to China. In addition, despite Taiwan's repeated willingness to participate, Taiwan is still not included in the 29 countries participating in this year's RIMPAC military exercise. Analysts pointed out that the US-led RIMPAC military exercise is intended to unite allies to militarily intimidate China. If Taiwan is invited to join, it will be too provocative to China, which will not only aggravate the tension between the United States and China, but also embarrass some allies. The U.S. Navy said the commander of the U.S. Third Fleet will serve as the commander of the joint task force for the exercise, while Chilean Navy Commodore Alberto Guerrero will serve as deputy commander of the joint task force, which is a first in the history of the RIMPAC military exercise. In addition, Japan Maritime Self-Defense Force Rear Admiral Kazushi Yokota will also serve as deputy commander. Other key leaders of the multinational force exercise include Canadian Commodore Kristjan Monaghan, who will command the maritime forces, and Australian Air Force Commodore Louise Desjardins, who will command the air forces. According to the U.S. Stars and Stripes, Vice Admiral Michael Boyle is currently the commander of the U.S. Third Fleet. Vice Admiral John Wade has been nominated to replace Boyle. The U.S. Navy press release said the exercise will enhance the ability of international joint forces to "deter and defeat aggression by major powers in all domains and conflict levels," but did not provide specific information on which exercises will be held this summer. Previous RIMPAC training exercises have included sinking ships at sea with missiles, amphibious landings and the first landing of a Marine Corps Osprey aircraft on an Australian ship.
Carlsberg to buy Britvic for $4.2 billion
Carlsberg to buy Britvic for 1,315p per share Carlsberg will also buy out Marston's from brewing joint venture Danish brewer plans to create integrated beverage business in UK Shares in Carlsberg, Britvic, Marston's all rise July 8 (Reuters) - Carlsberg (CARLb.CO), opens new tab has agreed to buy British soft drinks maker Britvic (BVIC.L), opens new tab for 3.3 billion pounds ($4.23 billion), a move the Danish brewer said would forge a UK beverage "powerhouse" and that sent both companies' shares higher. Carlsberg clinched the takeover with a sweetened bid of 1,315 pence per share - comprising cash and a special dividend of 25 pence a share - after the British company rejected 1,250 pence per share last month. The acquisition will create value for shareholders, contribute to growth and forge a combined beer and soft drink company that is unique in the UK, CEO Jacob Aarup-Andersen told investors on a conference call. "With this transaction we are creating a UK powerhouse," he said. He brushed off concerns from some analysts about integration risks, saying Carlsberg has a strong track record of running beer and soft drink businesses in several markets. Soft drinks already make up 16% of Carlsberg's volumes. COST SAVINGS As drinkers in some markets ditch beer for spirits or cut back on drinking altogether, brewers have looked to broaden their portfolio into new categories like hard seltzer, canned cocktails and cider, as well as zero-alcohol brews. Britvic sells non-alcoholic drinks in Britain, Ireland, Brazil and other international markets such as France, the Middle East and Asia. Carlsberg said the deal will deliver a number of benefits, including cost and efficiency savings worth 100 million pounds ($128 million) over five years as it takes advantage of common procurement, production and distribution networks. It will also see Carlsberg take over Britvic's bottling agreement with PepsiCo (PEP.O), opens new tab. Carlsberg already bottles PepsiCo drinks in several markets and there is scope to add more geographies in future, Aarup-Andersen said. arlsberg halted share buy backs on Monday as a result of the deal. Chief financial officer Ulrica Fearn said these would resume once Carlsberg reaches its revised target for net debt of 2.5 times EBITDA, from 3.5 times currently - a goal it expects to meet in 2027. "Whilst this represents a shift in the strategy away from organic top- and bottom-line growth and consistent returns to shareholders, we view it as a relatively low risk transaction with attractive financials," Jefferies analysts said in a note. Carlsberg also said on Monday it will buy out UK pub group Marston's (MARS.L), opens new tab from a joint venture for 206 million pounds. That will give it full ownership of the newly formed Carlsberg Britvic after the deal. ($1 = 0.7805 pounds) Get the latest news and expert analysis about the state of the global economy with Reuters Econ World. Sign up here. Reporting by Stine Jacobsen, Yadarisa Shabong and Emma Rumney Editing by Sherry Jacob-Phillips, Rashmi Aich, David Goodman and David Evans