Froth on the Daydream (FOD) – our weekly summary of over 150 AI newsletters. We connect the dots and cut through the froth, bringing you a comprehensive picture of the ever-evolving AI landscape. Stay tuned for clarity amidst surrealism and experimentation. The recent surge in the sophistication of — both proprietary and open-source — presents a paradox of potential and perplexity. These systems, characterized by their remarkable natural language processing capabilities, have propelled us into a new era of technological marvels. Yet, they also bring many challenges, chiefly in the realm of trustworthiness. Large Language Models (LLMs) Last week’s paper, “ ” — a joint work of almost 70 researchers — underscores the multifaceted nature of trustworthiness in LLMs. It highlights how these models, while excelling in tasks like stereotype rejection and natural language inference, still grapple with issues of truthfulness, safety, fairness, and privacy. These findings echo the complexities of ensuring AI that is both effective and ethically sound. TrustLLM: Trustworthiness in LLMs The paper also poses a question: “To what extent can we genuinely trust LLMs?” But can we genuinely trust LLMs? We can’t. Much better would be to adopt the principle of ‘trust, but verify.’ This approach, reminiscent of Cold War-era diplomacy, is increasingly relevant in the digital age, especially with advancements in AI. It suggests a balanced strategy: embracing the utility and potential of these models while stringently scrutinizing their mechanisms and outcomes. When working with LLMs, you can trust your expertise in verifying the work that LLM automates or accelerates for you. But you can’t just genuinely trust it. I even think that, along with the new role of an AI engineer, we should have a new job position for an in-house AI Verifier, akin to a fact-checker in a media publication. The other news from last week ‘complements’ the insights from the paper. Anthropic’s research on AI systems reveals a startling facet of deceptive ‘ ’ within LLMs. The paper studies threat models where AI models could be secretly trained or emerge to behave safely during training but unsafely in deployment. This discovery of hidden, hazardous capabilities within models, capable of evading standard safety protocols, exposes a critical vulnerability. sleeper agents Meanwhile, the nuanced shift in OpenAI’s policy, discreetly , adds another layer to the debate. This move, aligning with the U.S. Defense Department’s stance, prompts a critical examination of the ethical and safety implications of AI in high-stakes domains like defense and intelligence. Here, a question of the trustworthiness of the people who build Large Language Models (LLMs) also raises its head. lifting the prohibition on military applications On a more commercial and, so to speak, physical note, the launch of the , a standalone AI device, exemplifies the rapid integration of AI in consumer technology. Its innovative use of a Large Action Model (LAM)* signals a shift towards more intuitive, seamless interactions between humans and AI-powered devices. However, it also raises concerns about the trustworthiness and security of such pervasive AI integration in everyday life. Rabbit R1 *Many publications mistakenly attribute the coining of LAM to the Rabbit R1 team, when in fact, it was Salesforce Chief Scientist Silvio Savarese who coined it in June 2023 in his blog post “Towards Actionable Generative AI'“. Trust, but verify ;) Adding to the global perspective: In its “ ,” the World Economic Forum identified AI-generated misinformation and disinformation, along with the resultant societal polarization, as more significant threats in its list of top 10 risks for the next two years, surpassing concerns such as climate change, war, and economic instability. Global Risks Report 2024 As we navigate this era of groundbreaking AI advancements, the “trust, but verify” principle remains a beacon. We need to balance the excitement of AI’s potential with rigorous, ongoing scrutiny of its trustworthiness, safety, and ethical implications. The freshest research papers, categorized for your convenience Efficient Model Architectures . Researchers from the University of Warsaw developed MoE-Mamba, integrating Mamba, a State Space Model (SSM), with a Mixture of Experts (MoE) layer. This model outperforms both Mamba and Transformer-MoE in efficiency and performance, achieving equivalent results to Mamba with fewer training steps MoE-Mamba: Efficient Selective State Space Models with MoE →read the paper . Researchers from the University of Cambridge and University College London introduced “Blending,” a method combining smaller AI models to match or exceed the performance of larger models like ChatGPT Blending Is All You Need: Cheaper, Better Alternative to Trillion-Parameters LLM →read the paper . Researchers from the Beijing Academy of AI and Gaoling School of AI developed Activation Beacon, a module enhancing LLMs’ context window length Soaring from 4K to 400K: Extending LLM’s Context with Activation Beacon →read the paper Benchmark and Evaluation . Researchers from MIT CSAIL and Meta AI developed CRUXEval, a benchmark comprising 800 Python functions for evaluating code models’ reasoning and execution skills CRUXEval: A Benchmark for Code Reasoning, Understanding and Execution →read the paper . Researchers from Google Research and Tel Aviv University introduce GRANOLA QA, an evaluation setting for open-domain question answering (QA) that considers multi-granularity answers Narrowing the Knowledge Evaluation Gap: Open-Domain Question Answering with Multi-Granularity Answers →read the paper . Researchers from Carnegie Mellon University introduced TOFU, a benchmark for evaluating unlearning in LLMs using synthetic author profiles TOFU: A Task of Fictitious Unlearning for LLMs →read the paper Attention Mechanisms and Model Efficiency . Researchers from OpenNLPLab developed Lightning Attention-2, an advanced linear attention mechanism for LLMs that efficiently handles unlimited sequence lengths without increased memory usage or decreased speed Lightning Attention-2: A Free Lunch for Handling Unlimited Sequence Lengths in LLMs →read the paper . Researchers from The Hebrew University of Jerusalem and FAIR AI at Meta, redefined decoder-only transformers as a variant of Recurrent Neural Networks (RNNs) called infinite Multi-State RNNs (MSRNNs) Transformers are Multi-State RNNs →read the paper . Researchers from Google Research and Tel Aviv University introduce “Patchscopes,” a new framework for analyzing hidden representations in LLMs Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Models →read the paper Enhancing Model Performance . Researchers from University of California, San Diego and Google introduce “Chain-of-Table,” a framework that enhances table-based reasoning in LLMs for tasks like table-based question answering and fact verification Chain-of-Table: Evolving Tables in the Reasoning Chain for Table Understanding →read the paper . Researchers from Fudan NLP Lab & Fudan Vision and Learning Lab investigate Reinforcement Learning from Human Feedback (RLHF) in LLMs, focusing on improving reward models used for alignment Secrets of RLHF in LLMs Part II: Reward Modeling →read the paper . Researchers introduced DeepSeekMoE, an innovative Mixture-of-Experts (MoE) architecture for LLMs DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models →read the paper Machine Translation and Cross-Lingual Applications . Researchers from Apple introduced “contrastive alignment instructions” (AlignInstruct) to enhance MT in LLMs for unseen, low-resource languages Tuning LLMs with Contrastive Alignment Instructions for Machine Translation (MT) in Unseen, Low-resource Languages →read the paper Efficient Model Inference . Researchers from Intel developed an efficient inference solution for LLMs on Intel GPUs. This solution focuses on reducing latency and increasing throughput for LLMs Efficient LLM Inference Solution on Intel GPU →read the paper In other newsletters Is this what will replace Transformers? Long-Context Retrieval Models with Monarch Mixer · Hazy Research (stanford.edu) If you are interested in reportage from CES — is the one to go to (a very long reportage!) Hardcore software A wonderful overview of Self-Attention, Multi-Head Attention, Cross-Attention, and Causal-Attention in LLMs by Sebastian Raschka 12 techniques to reduce your LLM API bill and launch blazingly fast products by AI Tidbits DPO praise by — a very interesting read Andrew Ng Also published . here