The Massachusetts Institute of Technology (MIT) showcased a brand new robot training model this week that abandons the previous focus on training methods centered around specific datasets, opting instead for an approach that utilizes vast amounts of information similar to that used in training large language models (LLMs). Researchers noted that imitation learning—where agents learn by mimicking individuals performing tasks—can fail when facing minor challenges. These challenges may include varying lighting conditions, different environmental setups, or new obstacles. In such cases, the robots lack sufficient data to adapt to these changes.

This week, MIT showcased a new model for training robots, aimed at addressing the challenges faced by imitation learning when introduced to small obstacles. Researchers pointed out that imitation learning can fail in conditions such as varying lighting, different environments, or new obstacles because robots simply do not have enough data to adapt. The team sought powerful data methods, like those used in models such as GPT-4, to tackle the problem. They introduced a new architecture called the Heterogeneous Pretrained Transformer (HPT), which aggregates information from different sensors and environments.

Meta recently announced that its MobileLLM model is now open to researchers, allowing users to download and use these models for free on the Hugging Face platform, fully integrated with the Transformers library. This initiative aims to promote research and development of large language models (LLMs) on mobile devices, providing a broader range of tools and resources for developers and academia. MobileLLM represents a significant advancement for Meta in the field of mobile AI.

Is AI actually starting to ‘get smarter’?! The latest research from MIT indicates that the internal structure of large language models (LLMs) astonishingly resembles that of the human brain! This study utilized sparse autoencoder techniques to conduct an in-depth analysis of the activation space of LLMs, revealing three hierarchical structural features that are truly remarkable: first, at the microscopic level, researchers discovered the existence of structures akin to ‘crystals’. The surfaces of these ‘crystals’ are composed of parallelograms or trapezoids, similar to familiar lexical analogies such as ‘man:’

Recently, the Meta AI team introduced LongVU, a novel spatio-temporal adaptive compression mechanism aimed at enhancing the language understanding capabilities of long videos. Traditional multimodal large language models (MLLMs) face limitations in context length when processing long videos, and LongVU was created to address this challenge. LongVU operates primarily by filtering duplicate frames and employing inter-frame token compression techniques to efficiently utilize context length, allowing it to reduce video data while preserving visual details.

Large Language Models (LLMs) like the GPT series exhibit astonishing abilities in language understanding, reasoning, and planning, achieving human-level performance in various challenging tasks due to their vast datasets. Most research focuses on enhancing these models further by training them on larger datasets, aiming to develop more powerful foundational models. However, while training stronger foundational models is crucial, researchers believe that empowering models with the capability to continue evolving during the reasoning phase, known as AI self-evolution, is essential for the development of AI.

Recently, Google announced the open-sourcing of its text watermark tool SynthID, aimed at helping developers better identify AI-generated text. This tool is now publicly available through Google's 'Responsible AI Toolkit'. Pushmeet Kohli, Vice President of Research at Google DeepMind, stated that this technology will enable other generative AI developers to detect whether text outputs originate from their own Large Language Models (LLMs), which will help them be more responsible.

At the Honor MagicOS 9.0 launch event and developer conference on October 23, Honor CEO Zhao Ming announced the official release of the latest operating system, MagicOS 9.0. During his speech, Zhao reviewed the evolution of Honor MagicOS, which has positioned itself as an AI OS since its inception, continuously driving technological innovation.

Dow Jones, the parent company of The Wall Street Journal, and the New York Post are suing AI-driven search startup Perplexity for using their news content to train its large language model. The publications from these two news groups accuse Perplexity of copyright infringement by generating answers to people's queries using their articles, thereby diverting traffic from the publications' websites. 'This lawsuit is brought by news publishers seeking compensation as Perplexity openly plans to provide services for free.'

In the coming year of 2025, generative artificial intelligence (GenAI) is expected to reach a critical turning point, where we will gain a clearer understanding of its true potential. As analyses of its various applications across industries deepen, the GenAI that was once shrouded in noise will gradually reveal its true nature. Recently, Hieb noted that businesses investing in GenAI do not necessarily have to choose expensive and complex solutions. In fact, more and more companies are discovering that large language models (LLMs) don't always...

A recent study led by the Technion-Israel Institute of Technology reveals that large language models (LLMs) may be more knowledgeable than they appear. Researchers found that LLMs' internal representations encode information about the correctness of their outputs, allowing them to identify the correct answer even when they ultimately generate an incorrect one. The research team focused on analyzing LLM errors in long text generation, which is more aligned with their real-world application scenarios. They constructed an error detection dataset to compare the model's outputs.

Are you looking for tools to enhance your programming efficiency? I highly recommend trying DevKit! This AI-powered developer assistant has completely transformed my programming experience, and I'm excited to share my genuine thoughts. DevKit (https://www.getdevkit.com/) is not just a tool, but more like a programming partner. It integrates advanced Large Language Models (LLMs) and over 30 convenient mini-tools, covering all aspects of the development process.

In the world of artificial intelligence, the reasoning capabilities of machine learning models, particularly large language models (LLMs), have been a focal point for scientists. Recently, Apple's AI research team published a paper titled 'Understanding the Limitations of Large Language Models in Mathematical Reasoning,' shedding light on these models' limitations when addressing logical problems. In the paper, researchers demonstrate this through a simple mathematical question. They first posed a question about Oliver picking kiwis: as follows: Oliver picked 44 on Friday.

Recently, large language models (LLMs) with ultra-long context windows have become a hot topic of discussion. These models are capable of processing hundreds of thousands or even millions of tokens in a single prompt, opening up many new possibilities for developers. However, how well can these long context LLMs understand and utilize the vast information they receive? To address this question, researchers at Google DeepMind have launched a new benchmark called Michelangelo, aimed at evaluating long context reasoning capabilities. The research findings indicate that...

Recently, Apple's AI research team launched their next-generation family of Multimodal Large Language Models (MLLMs) - MM1.5. This series of models can integrate various data types such as text and images, showcasing new capabilities of AI in understanding complex tasks. Tasks like visual question answering, image generation, and multimodal data interpretation can be better addressed with the help of these models. A major challenge for multimodal models is how to achieve effective interaction between different data types. Previous models often struggled in this aspect.