Tim Hourigan·1d agoCan AI Really Reason? The More Important Enterprise Question Is What It Is Reasoning FromWhy Increasingly Capable AI Needs Explicit Sponsor Intent, Authoritative Organizational Intelligence, and Governed Context to Reason…
Arturo R Montesinos·Oct 30When the Question Is in Lojban, Does the AI Still Understand? (3/3)Three AIs, two constructed languages, and 27 experiments reveal how understanding, reasoning, and expression can fail in different ways.
Arturo R Montesinos·Oct 29Does Language Change the Way AI Thinks? (2/3)We Made Three AIs Speak Four Languages. One of Them Broke.A response icon1A response icon1
Joshua Idunnu Paul·Oct 29What Are AI Reasoning Models? How They Work and Whether They Really ThinkReasoning models spend more computation on hard problems. That can improve answers, but it doesn’t make them infallible.A response icon1A response icon1
Shirley Peng·Oct 26Beyond Hallucination: What AI Still Gets WrongI’ve been using the word hallucination for a long time to describe AI’s limitations.
InCodeToDeploybyArXiv In-depth Analysis·Oct 24Everyone Thought RL Taught Models to Say “Wait.” It Was Already in the Data.HKUST, Peking University, and Huawei Noah’s Ark trace the word through open training histories — and show zero-RL almost never coins it.A response icon1A response icon1
DataNorth AI·Oct 20AI effort levels: Choosing low, medium, high or max in LLM’sEvery major AI model now lets you choose how much thinking happens before an answer. OpenAI and xAI call the setting reasoning effort…A response icon3A response icon3
Rob Manson·Oct 5Don’t be fooled — LLMs really do reasonI disagree with the title of Thore Graepel’s article, but the more interesting debate is about his unstated assumption that “the model is…
InData Science CollectivebyHayanan·Oct 4ARGH, Why Is This So Hard?Deep inside their private thoughts, the most powerful AI models whisper “Wait,” groan “Hmm,” and cry “ARGH.” It looks like…A response icon1A response icon1
