Reporting and perspectives from The Gradient. Headlines and excerpts link to the original articles.
An essay by The Gradient argues that rational humans and rational AIs should not be goal-directed, proposing 'eudaimonic rationality' based on practices instead. It claims this framework could help align AI with safety properties like transparency, helpfulness, harmlessness, and corrigibility.
An essay in The Gradient argues that multimodal AI, which stitches together separate modality modules, will not reach human-level AGI in the near term, and that embodiment and environmental interaction should be treated as primary instead.
An article in The Gradient argues that mathematics remains central to machine learning research even as large-scale, engineering-first methods have produced breakthroughs that outpace existing theory. It says mathematics' role is evolving toward post-hoc explanation and higher-level design guidance, with tools such as intrinsic dimension, curvature and topology now applied to deep learning models.
A Gradient article by Kenneth Li argues that LLM chatbots lack an underlying purpose in multi-round dialogue, and that existing benchmarks measuring single-pass performance may not capture user experience. The piece reviews dialogue-system history and instruction-stability research, and presents the author's Dialogue Action Tokens algorithm trained with TD3+BC, which improved over baselines on Sotopia.
A Gradient essay argues that beneficial AI should be grounded in human wellbeing, that positive visions for an AI-infused society are needed, and that foundation models and their future deployment are critical leverage points.
An article by Richard Dewey and Ciamac Moallemi in The Gradient examines whether large language models can be applied to financial market prediction, weighing obstacles like scarce data and market efficiency against potential uses such as multimodal learning and synthetic data.
The Gradient article surveys research on measuring gender bias in AI models, covering word embeddings, facial recognition, coreference resolution, question answering, and image generation. It also discusses gaps in the research and open questions about how, or whether, such bias should be fixed.
The Gradient explains Mamba, a State Space Model (SSM) positioned as an alternative to Transformers for long-sequence tasks. Its authors, Gu and Dao, report linear scaling in sequence length, fast inference, up to 5x faster operation, and state-of-the-art results across language, audio, and genomics.
An article by Jérémy Cohen in The Gradient examines whether large language models could advance autonomous driving, surveying research areas including perception, planning and generation, and citing models such as Talk2BEV, DriveGPT and Wayve's GAIA-1. The author says it is too early to tell whether LLMs can be trusted in self-driving cars.
Researchers present vec2text, a method that recovers text from its embedding vectors, achieving 92% exact match on 32-token sequences and a BLEU score of 97. The work, published at EMNLP 2023, raises questions about the privacy of embeddings stored in vector databases.
A Gradient article by an author with about 20 years of machine learning experience describes common ways models appear to perform well while failing on real-world data, including misleading data, data leakage and inappropriate metrics. It cites a review by Roberts et al. of Covid prediction models and a Toronto water quality system, and points to the REFORMS checklist.
This article reviews how deep learning is applied to single-cell sequencing data analysis, covering autoencoder architectures such as denoising and variational autoencoders and their use in imputation and denoising tasks.
A consultant describes work on human-in-the-loop machine learning systems for counting fish at hydroelectric dams, where operators subject to FERC rules must produce fish passage data. The account covers a five-step build process, a 95% accuracy target compared with human visual counts, and obstacles including expert bias, environmental conditions, and model drift.
A Gradient article surveys neural algorithmic reasoning, the effort to capture classical computation such as shortest path-finding in deep neural networks. It covers algorithmic alignment theory, graph neural networks for algorithm execution, and applications including mathematics and computer networking.
An essay in The Gradient argues that current AI alignment research is shaped by commercial incentives, making it product development rather than a safeguard against long-term AI harms. It examines alignment definitions, RLHF and Constitutional AI, and the stated goals of OpenAI and Anthropic.