Tea@programming.devEnglish · 2 hours agoOLMo 2 32B sets a new standard for true open-source LLMs with public code, weights, and data, outperform GPT 3.5 and GPT 4o mini.plus-squareallenai.orgexternal-linkmessage-square1fedilinkarrow-up13arrow-down10
arrow-up13arrow-down1external-linkOLMo 2 32B sets a new standard for true open-source LLMs with public code, weights, and data, outperform GPT 3.5 and GPT 4o mini.plus-squareallenai.orgTea@programming.devEnglish · 2 hours agomessage-square1fedilink
Tea@programming.devEnglish · 2 hours agoHPC-AI Tech Releases Open-Sora 2.0: An Open-Source SOTA-Level Video Generation Model Trained for Just $200K.plus-squarecomfyui-wiki.comexternal-linkmessage-square0fedilinkarrow-up13arrow-down10
arrow-up13arrow-down1external-linkHPC-AI Tech Releases Open-Sora 2.0: An Open-Source SOTA-Level Video Generation Model Trained for Just $200K.plus-squarecomfyui-wiki.comTea@programming.devEnglish · 2 hours agomessage-square0fedilink
Tea@programming.devEnglish · 3 hours agoMost AI struggles to read clocks and calendars.plus-squarewww.ed.ac.ukexternal-linkmessage-square6fedilinkarrow-up11arrow-down11
arrow-up10arrow-down1external-linkMost AI struggles to read clocks and calendars.plus-squarewww.ed.ac.ukTea@programming.devEnglish · 3 hours agomessage-square6fedilink
Tea@programming.devEnglish · 2 hours agoSesame releases CSM-1B AI voice generator as open source.plus-squarehuggingface.coexternal-linkmessage-square0fedilinkarrow-up11arrow-down10
arrow-up11arrow-down1external-linkSesame releases CSM-1B AI voice generator as open source.plus-squarehuggingface.coTea@programming.devEnglish · 2 hours agomessage-square0fedilink
Nemeski@lemm.eeEnglish · 2 days agoOpenAI declares AI race “over” if training on copyrighted works isn’t fair useplus-squarearstechnica.comexternal-linkmessage-square12fedilinkarrow-up157arrow-down10
arrow-up157arrow-down1external-linkOpenAI declares AI race “over” if training on copyrighted works isn’t fair useplus-squarearstechnica.comNemeski@lemm.eeEnglish · 2 days agomessage-square12fedilink
Nemeski@lemm.eeEnglish · 2 days agoAI search engines give incorrect answers at an alarming 60% rate, study saysplus-squarearstechnica.comexternal-linkmessage-square3fedilinkarrow-up120arrow-down10
arrow-up120arrow-down1external-linkAI search engines give incorrect answers at an alarming 60% rate, study saysplus-squarearstechnica.comNemeski@lemm.eeEnglish · 2 days agomessage-square3fedilink
Nemeski@lemm.eeEnglish · 2 days agoOpenAI calls DeepSeek 'state-controlled,' calls for bans on 'PRC-produced' modelsplus-squaretechcrunch.comexternal-linkmessage-square4fedilinkarrow-up16arrow-down10
arrow-up16arrow-down1external-linkOpenAI calls DeepSeek 'state-controlled,' calls for bans on 'PRC-produced' modelsplus-squaretechcrunch.comNemeski@lemm.eeEnglish · 2 days agomessage-square4fedilink
Nemeski@lemm.eeEnglish · 2 days agoAI-Generated Voice Evidence Poses Dangers in Courtplus-squarewww.lawfaremedia.orgexternal-linkmessage-square0fedilinkarrow-up12arrow-down10
arrow-up12arrow-down1external-linkAI-Generated Voice Evidence Poses Dangers in Courtplus-squarewww.lawfaremedia.orgNemeski@lemm.eeEnglish · 2 days agomessage-square0fedilink
Nemeski@lemm.eeEnglish · 2 days agoAnthropic CEO says spies are after $100M AI secrets in a ‘few lines of code’plus-squaretechcrunch.comexternal-linkmessage-square0fedilinkarrow-up13arrow-down12
arrow-up11arrow-down1external-linkAnthropic CEO says spies are after $100M AI secrets in a ‘few lines of code’plus-squaretechcrunch.comNemeski@lemm.eeEnglish · 2 days agomessage-square0fedilink
Tea@programming.devEnglish · edit-23 days agoGoogle DeepMind’s new AI models help robots perform physical tasks, even without training.plus-squaredeepmind.googleexternal-linkmessage-square0fedilinkarrow-up16arrow-down10
arrow-up16arrow-down1external-linkGoogle DeepMind’s new AI models help robots perform physical tasks, even without training.plus-squaredeepmind.googleTea@programming.devEnglish · edit-23 days agomessage-square0fedilink
Tea@programming.devEnglish · edit-23 days agoWelcome Gemma 3: Google's all new multimodal, multilingual, long context open LLM.plus-squarehuggingface.coexternal-linkmessage-square1fedilinkarrow-up111arrow-down14
arrow-up17arrow-down1external-linkWelcome Gemma 3: Google's all new multimodal, multilingual, long context open LLM.plus-squarehuggingface.coTea@programming.devEnglish · edit-23 days agomessage-square1fedilink
Tea@programming.devEnglish · 5 days agoNew method significantly reduces AI energy consumption.plus-squarewww.tum.deexternal-linkmessage-square1fedilinkarrow-up119arrow-down10
arrow-up119arrow-down1external-linkNew method significantly reduces AI energy consumption.plus-squarewww.tum.deTea@programming.devEnglish · 5 days agomessage-square1fedilink
Tea@programming.devEnglish · edit-24 days agoOpenAI debuts a Responses API to help devs build agents that search the web, scan for files, and perform tasks on computers, and an Agents SDK for orchestration.plus-squareopenai.comexternal-linkmessage-square0fedilinkarrow-up14arrow-down12
arrow-up12arrow-down1external-linkOpenAI debuts a Responses API to help devs build agents that search the web, scan for files, and perform tasks on computers, and an Agents SDK for orchestration.plus-squareopenai.comTea@programming.devEnglish · edit-24 days agomessage-square0fedilink
Tea@programming.devEnglish · edit-25 days agoFoxconn says it built FoxBrain, an in-house reasoning LLM, trained in four weeks with support from Nvidia via its Taiwan-based supercomputer and consulting.plus-squarewww.morningstar.comexternal-linkmessage-square1fedilinkarrow-up17arrow-down10
arrow-up17arrow-down1external-linkFoxconn says it built FoxBrain, an in-house reasoning LLM, trained in four weeks with support from Nvidia via its Taiwan-based supercomputer and consulting.plus-squarewww.morningstar.comTea@programming.devEnglish · edit-25 days agomessage-square1fedilink
Tea@programming.devEnglish · 7 days agoZoom researchers detail a “chain of draft” method to let LLMs accurately solve reasoning problems with as little as 7.6% of the tokens used by current methods.plus-squarearxiv.orgexternal-linkmessage-square2fedilinkarrow-up110arrow-down10
arrow-up110arrow-down1external-linkZoom researchers detail a “chain of draft” method to let LLMs accurately solve reasoning problems with as little as 7.6% of the tokens used by current methods.plus-squarearxiv.orgTea@programming.devEnglish · 7 days agomessage-square2fedilink
Tea@programming.devEnglish · 4 days agoHere’s how I use LLMs to help me write code.plus-squaresimonwillison.netexternal-linkmessage-square0fedilinkarrow-up12arrow-down13
arrow-up1-1arrow-down1external-linkHere’s how I use LLMs to help me write code.plus-squaresimonwillison.netTea@programming.devEnglish · 4 days agomessage-square0fedilink
Tea@programming.devEnglish · edit-29 days agoMistral launches Mistral OCR, a multimodal API that can turn complex PDF documents into AI-ready Markdown files.plus-squaremistral.aiexternal-linkmessage-square0fedilinkarrow-up112arrow-down11
arrow-up111arrow-down1external-linkMistral launches Mistral OCR, a multimodal API that can turn complex PDF documents into AI-ready Markdown files.plus-squaremistral.aiTea@programming.devEnglish · edit-29 days agomessage-square0fedilink
Tea@programming.devEnglish · edit-29 days agoAlibaba releases QwQ-32B, an open-source reasoning model, on Hugging Face and ModelScope, claiming performance similar to DeepSeek-R1 with lower compute needs.plus-squareqwenlm.github.ioexternal-linkmessage-square0fedilinkarrow-up19arrow-down10
arrow-up19arrow-down1external-linkAlibaba releases QwQ-32B, an open-source reasoning model, on Hugging Face and ModelScope, claiming performance similar to DeepSeek-R1 with lower compute needs.plus-squareqwenlm.github.ioTea@programming.devEnglish · edit-29 days agomessage-square0fedilink
Tea@programming.devEnglish · 10 days agoMicrosoft unveils Sales Agent and Sales Chat AI agents, available in public preview in May, designed to work with Dynamics 365 business apps and with Salesforce.plus-squarewww.microsoft.comexternal-linkmessage-square1fedilinkarrow-up14arrow-down15
arrow-up1-1arrow-down1external-linkMicrosoft unveils Sales Agent and Sales Chat AI agents, available in public preview in May, designed to work with Dynamics 365 business apps and with Salesforce.plus-squarewww.microsoft.comTea@programming.devEnglish · 10 days agomessage-square1fedilink
Tea@programming.devEnglish · edit-210 days agoAMD Announces "Instella" Open-Source 3B Language Models.plus-squarerocm.blogs.amd.comexternal-linkmessage-square0fedilinkarrow-up15arrow-down10
arrow-up15arrow-down1external-linkAMD Announces "Instella" Open-Source 3B Language Models.plus-squarerocm.blogs.amd.comTea@programming.devEnglish · edit-210 days agomessage-square0fedilink