Perplexity AI, a company that blends a search engine with generative AI to deliver AI-created content related to user search queries, has been accused of unethical data collection practices. It allegedly scraped content from several websites, including those that expressly disallow it, without proper protocol. The controversy began on June 11th when Forbes claimed that…
A study conducted by researchers from Carnegie Mellon University, Google DeepMind, and MultiOn focuses on the role of synthetic data in enhancing the mathematical reasoning capabilities of large language models (LLMs). Predictions indicate that high-quality internet data necessary for training models could be depleted by 2026. As a result, model-generated or synthetic data is considered…
Artificial Intelligence (AI) innovations continue to pose particular challenges to existing legal frameworks, particularly in the realm of assigning liability due to the lack of discernible intentions, which is traditionally important in establishing liability. This problem is addressed in a new paper from Yale Law School, which suggests employing objective standards in regulating AI.
By viewing…
Last year, MIT President Sally Kornbluth and Provost Cynthia Barnhart encouraged academics to submit papers outlining roadmaps, policy recommendations, and calls to action in the area of generative AI. This generated a strong response, with 75 submissions being made. 27 of these were selected to receive seed funding.
Due to the high interest and quality of…
Large language models (LLMs), instrumental in natural language processing tasks like translation, summarization, and text generation, face challenges in consistently adhering to logical constraints during text generation. This adherence is crucial in sensitive applications where precision and instruction compliance are crucial. Traditional methods for imposing constraints on LLMs, such as the GeLaTo framework, have limitations…
Large language models (LLMs) are central to the field of natural language processing, being utilized in tasks like translation, summarization, and creative text generation. They utilize extensive data to learn patterns and relationships in languages, enabling them to undertake tasks necessitating an understanding of context, syntax, and semantics. However, there's a persistent challenge in ensuring…
Researchers from various international institutions have developed a computational method called RAIN to rapidly identify broadly neutralizing antibodies (bNAbs) against HIV-1. bNAbs can target the virus's envelope proteins to reduce viral loads and stop infection, but the process of discovering them is an arduous one due to the need for B-cell isolation and next-generation sequencing,…
Broadly neutralizing antibodies (bNAbs) play a crucial role in fighting HIV-1, functioning by targeting the virus's envelope proteins which shows promise in reducing viral loads and preventing infection. However, identifying these antibodies is a complex process due to the virus's rapid mutation and evasion from the immune system. Only 255 bNAbs have been discovered, therefore…
The Massachusetts Institute of Technology (MIT) has announced its plan to fund 16 research proposals dedicated to exploring the potential of generative Artificial Intelligence (AI). The funding process began last summer when MIT President Sally Kornbluth and Provost Cynthia Barnhart invited research papers that could provide robust policy guidelines, efficient roadmaps, and calls to action…
Reinforcement learning from human feedback (RLHF) is a technique that encourages artificial intelligence (AI) to generate high rewards by aligning large language models (LLMs) with a reward model based on human preferences. However, it is beset by several challenges, such as the limiting of fine-tuning processes to small datasets, the risk of AI exploiting flaws…
Reinforcement Learning from Human Feedback (RLHF) uses a reward model trained on human preferences to align large language models (LLMs) with the aim of optimizing rewards. Yet, there are issues such as the model becoming too specialized, the potential for the LLM to exploit flaws in the reward model, and a reduction in output variety.…
Machine learning pioneer Hugging Face has launched Transformers version 4.42, a meaningful update to its well-regarded machine-learning library. Significant enhancements include the introduction of several advanced models, improved tool and retrieval-augmented generation support, GGUF fine-tuning, and quantized KV cache incorporation among other enhancements.
The release features the addition of new models like Gemma 2, RT-DETR, InstructBlip,…