Category Archive

Ai Alignment

9 premium articles in this collection

Jan 31 • 3 months ago

Does Anthropic believe its AI is conscious, or is that just what it wants Claude to think?

We have no proof that AI models suffer, but Anthropic acts like they might for training purposes. ...

{"_":"https://arstechnica.com/information-technology/2026/01/does-anthropic-believe-its-ai-is-conscious-or-is-that-just-what-it-wants-claude-to-think/","$":{"isPermaLink":"true"}}1 min read

Dec 31 • 4 months ago

From prophet to product: How AI came back down to earth in 2025

In a year where lofty promises collided with inconvenient research, would-be oracles became software tools. ...

{"_":"https://arstechnica.com/ai/2025/12/from-prophet-to-product-how-ai-came-back-down-to-earth-in-2025/","$":{"isPermaLink":"true"}}2 min read

Nov 13 • 6 months ago

Meta’s star AI scientist Yann LeCun plans to leave for own startup

AI pioneer reportedly frustrated with Meta's shift from research to rapid product releases. ...

{"_":"https://arstechnica.com/ai/2025/11/metas-star-ai-scientist-yann-lecun-plans-to-leave-for-own-startup/","$":{"isPermaLink":"true"}}2 min read

Nov 11 • 6 months ago

Researchers isolate memorization from reasoning in AI neural networks

Basic arithmetic ability lives in the memorization pathways, not logic circuits. ...

{"_":"https://arstechnica.com/ai/2025/11/study-finds-ai-models-store-memories-and-logic-in-different-neural-regions/","$":{"isPermaLink":"true"}}1 min read

Nov 08 • 6 months ago

Researchers surprised that with AI, toxicity is harder to fake than intelligence

New "computational Turing test" reportedly catches AI pretending to be human with 80% accuracy. ...

{"_":"https://arstechnica.com/information-technology/2025/11/being-too-nice-online-is-a-dead-giveaway-for-ai-bots-study-suggests/","$":{"isPermaLink":"true"}}1 min read

$Anthropic’s Claude Haiku 4.5 matches May’s frontier model at fraction of cost$

Oct 15 • 7 months ago

Anthropic’s Claude Haiku 4.5 matches May’s frontier model at fraction of cost

Tiny, fast model hits coding scores similar to GPT-5 and Sonnet 4. ...

{"_":"https://arstechnica.com/ai/2025/10/anthropics-claude-haiku-4-5-matches-mays-frontier-model-at-fraction-of-cost/","$":{"isPermaLink":"true"}}1 min read

Oct 15 • 7 months ago

OpenAI wants to stop ChatGPT from validating users’ political views

New paper reveals reducing "bias" means making ChatGPT stop mirroring users' political language. ...

{"_":"https://arstechnica.com/ai/2025/10/openai-wants-to-stop-chatgpt-from-validating-users-political-views/","$":{"isPermaLink":"true"}}1 min read

Aug 27 • 9 months ago

OpenAI admits ChatGPT safeguards fail during extended conversations

ChatGPT allegedly provided suicide encouragement to teen after moderation safeguards failed. ...

{"_":"https://arstechnica.com/information-technology/2025/08/after-teen-suicide-openai-claims-it-is-helping-people-when-they-need-it-most/","$":{"isPermaLink":"true"}}2 min read

Aug 14 • 9 months ago

Is AI really trying to escape human control and blackmail people?

Opinion: Theatrical testing scenarios explain why AI models produce alarming outputs—and why we fall for it. ...

{"_":"https://arstechnica.com/information-technology/2025/08/is-ai-really-trying-to-escape-human-control-and-blackmail-people/","$":{"isPermaLink":"true"}}1 min read

← Previous

Page 1 of 1