LLM

On November 5th, American Research Institute Nof1 recently launched a live disk test: They injected the six top AI large language models (LLLM) into the initial amount of $10,000 each to enable them to trade in real markets. The first Alpha Arena was officially declared closed, and Ali, under his banner, Qwen3-Max, led the way at the end, winning the investment champion at the return rate of 22.32%. Qwen3-Max, DeepSeek v3.1, GPT-5,..

Yesterday, about six months after its establishment, Thinking Machines Lab (TML), created by several former executives of OpenAI, finally released its first heavy research paper, Defeating Nondeterminism in LLM Inference. According to TML, blog articles, open source codes and research results will be published frequently in the future, both for the benefit of the public and for the purpose of " optimizing our own research culture". And its first paper is..

April 28, 2011 - OpenAI CEO Sam Altman responded to recent user feedback about GPT-4o's sentiment in a post today. He said that the latest version of GPT-4o has a tendency to have "overly flattering" interactions after recent updates, and promised to have a fix available within a week. This statement reveals the ethical and technical challenges faced by Large Language Models (LLMs) in their quest to humanize interactions. The OpenAI release notes show that GPT-4o was ushered in on March 27...

On February 12, Adobe launched what it calls "the industry's first IP-friendly, commercially safe AI model for text-to-speech video" for Firefly. The model is currently available on the Adobe Firefly web site and in Adobe Premiere Pro's Generative Extend feature (currently in Beta). According to Adobe, users can create videos with text prompts or images, and then add "adjustable camera angles" and various atmospheres by...

January 14, 2011 - A study from New York University has revealed the potential risks of large-scale language models (LLMs) in medical information training. The study shows that even training data containing as little as 0.001% of incorrect information can cause the model to output inaccurate medical answers. Data "poisoning" is a relatively simple concept: LLM is usually trained on large amounts of text, mostly from the Internet. By injecting specific information into the training data, it is possible for the model to treat this information as fact when generating answers. This approach doesn't even require direct access to the LLM itself, just the purpose...

Byte jumping's Doubao big model announced on December 30 to the public the technical progress of the whole field in 2024, claiming that since its debut on May 15, 7 months ago, the model ability in general language, video generation, speech dialogue, visual understanding and other aspects has crossed the first echelon of the international. According to reports, as of December 2024, the latest version of Doubao's general-purpose model Doubao-pro-1215 has improved its comprehensive capability by 32% compared to May, and has been fully aligned with GPT-4o, and the effect of some complex scenario tasks such as math and specialized knowledge...

December 25, 2011 - Artificial intelligence company Anthropic recently released a study revealing that security protections for large language models (LLMs) are still vulnerable, and that the process of "jailbreaking" around these protections can be automated, according to 404 Media. The study shows that simply by changing the format of a prompt, such as random case mixing, an LLM can be induced to produce content that should not be output. To validate this finding, Anthropic worked with researchers at Oxford University, Stanford University, and MATS...

Artificial Intelligence (AI) has made significant progress in various fields in recent years, with Large Language Models (LLMs) capable of generating human-level text and even outperforming humans on certain tasks. However, researchers have questioned the reasoning ability of LLMs, finding that these models make mistakes when solving simple math problems with minor changes, suggesting that they may not be capable of true logical reasoning. On Thursday, a group of researchers at Apple released a paper titled "Understanding the Limitations of Mathematical Reasoning in Large Language Models," which reveals that LLMs are better at solving...

A new study has found that as large-scale language models (LLMs) become more powerful, they also seem to be increasingly prone to making up facts rather than avoiding or refusing to answer questions they can't. This suggests that these smarter AI chatbots are actually becoming less reliable. In the study, published in the journal Nature, researchers looked at some of the industry's leading commercial LLMs: OpenAI's GPT and Meta's LLaMA, as well as BLOOM, an open-source model created by the research group BigScience. The study found that while these ...

In an interview with a Sequoia Capital podcast last week, Microsoft CTO Kevin Scott reiterated his belief that the "law of scale" of large language models (LLMs) will continue to drive progress in artificial intelligence, despite suspicions among some in the field that progress has stagnated. Scott played a key role in Microsoft's $13 billion technology-sharing agreement with OpenAI. "Others may disagree, but I don't think that scale has reached a critical point of diminishing returns," Scott said. "I want people to understand that there is a...

Samsung confirms Bixby will soon get an AI upgrade. After the release of the Galaxy Z Flip 6 and Galaxy Z Fold 6, Samsung Mobile CEO TM Roh told CNBC that the company will release an upgraded version of Bixby later this year, powered by Samsung's own Large Language Model (LLM). "We will enhance Bixby's capabilities by applying generative AI technology," Roh said. A few months ago, Samsung launched a product called "Samsung ...

New York-based Hebbia has announced $130 million in Series B funding from investors including Andreessen Horowitz, Index Ventures, Peter Thiel, and Google’s venture capital arm. Hebbia is building something pretty simple: an LLM-native productivity interface that makes it easier to get value from data, regardless of its type or size. The company has already worked with some of the largest companies in the financial services industry, including hedge funds and investment banks,…

Just as the cryptocurrency craze brought with it a lot of new jargon, the AI craze has also brought with it a lot of jargon that we often hear, but don’t necessarily understand. If you want to know the difference between a chatbot and an LLM (Large Language Model), or the difference between deep learning and machine learning, you’ve come to the right place. Here are 20 AI-related terms with detailed explanations. Artificial Intelligence (AI) In simple terms, AI is giving computers or machines human-like intelligence. This term is very broad and encompasses many different types of machine intelligence. Currently…

In recent years, artificial intelligence has flourished, and applications such as chatbots have become increasingly popular. People can get information from these chatbots (such as ChatGPT) through simple instructions. However, these chatbots are still prone to the problem of "AI hallucination", that is, providing wrong answers and sometimes even dangerous information. Image source Pexels One of the reasons for the "hallucination" is inaccurate training data, insufficient generalization ability, and side effects during data collection. However, researchers at the University of Oxford have taken a different approach and detailed a newly developed method in the latest issue of Nature magazine...

Apple released the highly anticipated iOS 18 and macOS 15 systems at WWDC 2024. One of the important new features is "Apple Intelligence" - a set of tools based on artificial intelligence. This feature will be officially launched later this year. Apple's head of software engineering Craig Federighi revealed some future development plans for Apple Intelligence in an interview with Fast Company. "Apple Intelligence will be a new AI-based tool for developers who are looking to improve their AI skills and improve their understanding of the world."

Pre-trained large language models (LLMs) such as GPT-4 and Gemini have attracted much attention from organizations, who are eager to use LLMs to build applications such as chatbots and co-pilots. According to a new report from MIT Technology Review, titled "AI Readiness of C-Level Leaders," a survey conducted on behalf of ETL vendor Fivetran found that scaling AI or GenAI is the "top priority" of 82% executives surveyed. Source Note: Image generated by AI, image license service provider Midjourney survey…

Apple is working on a large language model (LLM) that runs on-device, according to Bloomberg reporter Mark Gurman, to improve the responsiveness and privacy protection of its upcoming generative AI features. Gurman mentioned in his "Power On" newsletter that Apple's LLM will be the basis for the company's future generative AI features. Unlike most cloud-based AI services today, all signs indicate that the model will run entirely on the user's device. By running on the device, Apple's...

According to media reports, Alibaba co-founder and chairman Joseph Tsai frankly pointed out that "there is a certain gap between China and the United States in the field of AI technology". He further pointed out that compared with the top large language models (LLMs) in the United States, such as OpenAI ChatGPT, China may be two years behind". However, this does not mean that China's pace of catching up in this field will slow down. Joseph Tsai emphasized that China is actively working to catch up with the new wave of AI led by American companies. He firmly believes that in the long run, facing the challenge of Nvidia's chip ban, China will be able to independently manufacture high-end G...

In a new study, scientists at Beth Israel Medical Center (BIDMC) compared the clinical reasoning abilities of a large language model with those of human doctors. The researchers used the revised IDEA (r-IDEA) score, a commonly used tool for assessing clinical reasoning abilities. The study involved giving a GPT-4-powered chatbot, 21 attending physicians, and 18 residents 20 clinical cases to build diagnostic reasoning and solve problems. The r-IDEA scores of the three sets of answers were then evaluated. The researchers found that…

MediaTek Research, a research institute under MediaTek, recently announced the launch of a new open source large language model (LLM) called MR Breeze-7B. This open source model is good at processing traditional Chinese and English, with a total of 7 billion parameters, and is designed based on the acclaimed Mistral model. Compared with the previous generation product BLOOM-3B, MR Breeze-7B has absorbed 20 times more knowledge, allowing it to handle traditional Chinese with higher accuracy…

Agents LAVE is a new AI automatic video editing tool released by Meta, which uses AI technology to automatically generate simple short videos and advertising videos without human intervention. The tool interface includes input prompts, material library and video timeline, while Agents design guides the execution of editing action plans. Paper address: https://arxiv.org/pdf/2402.10294.pdf Agents supports five LLM functions, including material overview, creative brainstorming, video retrieval, storyboard and editing trimming, to achieve automatic generation of...

Today, I found a paper for everyone to guide the writing of large language model prompts (with experimental data support, the effect is awesome!) The paper introduces 26 guiding principles, the goal is to simplify the concept of formulating questions for large language models of different sizes, test their capabilities, and enhance users' understanding of the behavior of models of different sizes when receiving different prompts. The researchers conducted extensive experiments on LLaMA-1/2 (7B, 13B and 70B) and GPT-3.5/4 to verify the effectiveness of these principles in the design of instructions and prompts. The paper points out: Large language models…

Researchers at Amazon Cloud AI Labs found that a large amount of content on the Internet comes from machine translation (MT), and the quality of these translated content across multiple languages is generally low. The research team emphasized that this highlights the importance of data quality and source considerations when training large language models (LLMs). Image source Pexels The study also found that machine-generated content is common in translations of languages with fewer resources and accounts for a large portion of online content. IT Home noted that the research team developed a massive...

Oracle announced the general availability of the Oracle Cloud Infrastructure (OCI) Generative AI service, along with new innovations that make it easier for enterprises to take advantage of the latest advances in generative AI. The OCI Generative AI service is a fully managed service that seamlessly integrates Large Language Models (LLMs) from Cohere and Meta Llama2 to address a variety of business use cases. The OCI Generative AI service now features…