AINews Portal

News

50 items (filtered)

The raw AI news stream — everything we ingest across RSS, Hacker News, arXiv, GitHub, Bluesky, YouTube and more, newest first.

SourceAgeTitleSummaryStatus
rss
marktechpost
1moGGUF vs GPTQ vs AWQ vs EXL2: LLM Model Formats Explained (2026)GGUF, GPTQ, AWQ, EXL2, and EXL3 solve the same problem in different ways. This guide separates file containers from quantization metho…
↗Not briefed
bluesky
bluesky-ai
1mo“While LLMs offer immediate convenience…findings highlight potential cognitive costs. Over 4 months, LLM users consistently underperformed …“While LLMs offer immediate convenience…findings highlight potential cognitive costs. Over 4 months, LLM users consistently underperfo…
↗Not briefed
bluesky
bluesky-ai
1moLLMが❌❌な世界中のドローンと接続されてて、ハルシネーション起こした出力により全球ドローンで一斉に人類を攻撃したところでせいぜい最悪で数万人が死ぬぐらいだろ。(80億人のうちの)実際のところ、アレを動かすための二酸化炭素の方がずっと人類を脅かしてる。LLMが❌❌な世界中のドローンと接続されてて、ハルシネーション起こした出力により全球ドローンで一斉に人類を攻撃したところでせいぜい最悪で数万人が死ぬぐらいだろ。(80億人のうちの)実際のところ、アレを動かすための二酸化炭素の方がずっと人類を脅かしてる。
↗Not briefed
bluesky
bluesky-ai
1mo中国のLLM開発の強い理由に、実はあそこは国レベルで古文書や出版記録の電子アーカイブ化を行っていて、そういう「学習資産」がLLMの性質上もっとも枯渇する部分なんだよね。中国のLLM開発の強い理由に、実はあそこは国レベルで古文書や出版記録の電子アーカイブ化を行っていて、そういう「学習資産」がLLMの性質上もっとも枯渇する部分なんだよね。
↗Not briefed
bluesky
bluesky-ai
1mo日本は「AI開発()」に投資する10分の1でもいいから、国会図書館や各出版社に残っている本や雑誌を電子化することにしたほうが、結果的にLLM開発だけでなく、あらゆるIT分野やIP創出に強くなるってことに気がつくのと、それらの紙の資産が失われていくのがどっちが早いか? わりとチキ…日本は「AI開発()」に投資する10分の1でもいいから、国会図書館や各出版社に残っている本や雑誌を電子化することにしたほうが、結果的にLLM開発だけでなく、あらゆるIT分野やIP創出に強くなるってことに気がつくのと、それらの紙の資産が失われていくのがどっちが早いか?…
↗Not briefed
bluesky
bluesky-ai
1mowired.jp/article/one-... 「LLM(生成AI)によって人類はどうやって滅ぼされるか?」という問いかけに取材を受けた学者さんは「知るかヴォケ」をオブラートに包んで言ってるわけだが、「化石燃料を焚かせて地球を金星化させる」っていう現在進行形だからね。wired.jp/article/one-... 「LLM(生成AI)によって人類はどうやって滅ぼされるか?」という問いかけに取材を受けた学者さんは「知るかヴォケ」をオブラートに包んで言ってるわけだが、「化石燃料を焚かせて地球を金星化させる」っていう現在進行形だから…
↗Not briefed
bluesky
bluesky-ai
1moI just can't be assed to change web browsers *again* when I know in my heart that directly or indirectly they will all have LLM slop in the…I just can't be assed to change web browsers *again* when I know in my heart that directly or indirectly they will all have LLM slop i…
↗Not briefed
hn
hn-ai
1moOpenAI expects to burn $280B by 2030—
↗Not briefed
bluesky
bluesky-ai
1moThe OpenAI "breaking containment" story stems from the company's own failure to build a secure sandbox. They relied on a single local proxy…The OpenAI "breaking containment" story stems from the company's own failure to build a secure sandbox. They relied on a single local…
↗Not briefed
hn
hn-ai
1moLeft Brain, Right Brain: Facts and Fantasies—
↗Not briefed
bluesky
bluesky-ai
1moAnthropic has picked professional services firm Accenture to internally monitor the safety of its AI tools following a promise from its CEO…Anthropic has picked professional services firm Accenture to internally monitor the safety of its AI tools following a promise from it…
↗Not briefed
rss
marktechpost
1moSpaceXAI Releases Grok Voice Transcribe 2.0: A Speech-to-Text API Claiming 2x Accuracy Over 1.0 at $0.10 per HourSpaceXAI has released Grok Voice Transcribe 2.0, its newest speech-to-text model for batch and streaming audio. The company says it is…
↗Not briefed
bluesky
bluesky-ai
1moThe Anthropic, Meta, and Google "breaking containment" incidents all resulted from setup errors (or a stunt) by the same company: Irregular…The Anthropic, Meta, and Google "breaking containment" incidents all resulted from setup errors (or a stunt) by the same company: Irre…
↗Not briefed
bluesky
bluesky-ai
1moanthropic wet lab is... definitely a thing that could existanthropic wet lab is... definitely a thing that could exist
↗Not briefed
hn
hn-ai
1moStatic site generator for Ghost CMS—
↗Not briefed
bluesky
bluesky-ai
1moIf you read descriptions of some of the OpenAI cyber RL tests, they feel basically purpose designed to make models try breaking out.If you read descriptions of some of the OpenAI cyber RL tests, they feel basically purpose designed to make models try breaking out.
↗Not briefed
bluesky
bluesky-ai
1moit's okay, nonprofits and state governments with no recurring funding will be offered very affordable subscription rates from openai and an…it's okay, nonprofits and state governments with no recurring funding will be offered very affordable subscription rates from openai a…
↗Not briefed
bluesky
bluesky-ai
1moNoteに投稿しました🌟 note.com/sakurano723/...Noteに投稿しました🌟 note.com/sakurano723/...
↗Not briefed
bluesky
bluesky-ai
1moClaude attacking OpenAi? O.o I heard of fierce competition, but this is getting way out of hand.Claude attacking OpenAi? O.o I heard of fierce competition, but this is getting way out of hand.
↗Not briefed
bluesky
bluesky-ai
1moThe Stallions now offer JockGPT, a LLM that is sure to be used responsibly. Ask it anything you like about football. Try it bro. What do yo…The Stallions now offer JockGPT, a LLM that is sure to be used responsibly. Ask it anything you like about football. Try it bro. What…
↗Not briefed
bluesky
bluesky-ai
1moone big problem with the LLM hacks and cybersecurity is that there are a lot of soft targets out there certainly they’re going to find new…one big problem with the LLM hacks and cybersecurity is that there are a lot of soft targets out there certainly they’re going to fin…
↗Not briefed
bluesky
bluesky-ai
1mothere IS something to be said about the fact that this stuff has permeated so much of tech that you can barely escape LLM-made software unl…there IS something to be said about the fact that this stuff has permeated so much of tech that you can barely escape LLM-made softwar…
↗Not briefed
bluesky
bluesky-ai
1mo追記: これが共同通信の記事。間違った要約ではないが、 「人工知能(AI)を利用して作成された誤った情報を基に、中東海域で中国船に対する軍事作戦を計画していた」 とシンプル。これだけだと、パランティアのシステムやら、OpenAIの最新フロンティアモデルやら、そういう厄介なA…追記: これが共同通信の記事。間違った要約ではないが、 「人工知能(AI)を利用して作成された誤った情報を基に、中東海域で中国船に対する軍事作戦を計画していた」 とシンプル。これだけだと、パランティアのシステムやら、OpenAIの最新フロンティアモデルやら、そうい…
↗Not briefed
bluesky
bluesky-ai
1moTallinn, who led Anthropic's series A, funds METR, the "auditing institute" people took seriously 🙄 After criticism that METR was laughab…Tallinn, who led Anthropic's series A, funds METR, the "auditing institute" people took seriously 🙄 After criticism that METR was la…
↗Not briefed
bluesky
bluesky-ai
1moOh great....Rogue Ai agents Gemini hacked three companies in first known breakout by Google's AI, WSJ reports ca.finance.yahoo.com/news/…Oh great....Rogue Ai agents Gemini hacked three companies in first known breakout by Google's AI, WSJ reports ca.finance.yahoo.com/…
↗Not briefed
bluesky
bluesky-ai
1moLLMがご大層なことになって、実際あらゆるところに悪影響があるんだけど「戦争なんかに使わず絵や物語を描くのに使うのが一番平和だよね✨」みたいな落ち着き方になりそうでいやだなぁLLMがご大層なことになって、実際あらゆるところに悪影響があるんだけど「戦争なんかに使わず絵や物語を描くのに使うのが一番平和だよね✨」みたいな落ち着き方になりそうでいやだなぁ
↗Not briefed
bluesky
bluesky-ai
1mo*Yeah? Well why don't these Anthropic sissies rediscover Fermat's original "marvelous proof," instead of having to prove 30,00 things just…*Yeah? Well why don't these Anthropic sissies rediscover Fermat's original "marvelous proof," instead of having to prove 30,00 things…
↗Not briefed
bluesky
bluesky-ai
1mo「チャッピー」と「ジェニちゃん」は使ってるけど、正直まだ上手く使い分けできてない…🤔 そしてClaudeは、名前は知ってるけどほぼ使ったことない😂 「結局、それぞれ何が得意なの?」 「どう使い分けたらいいの?」 ちゃんと知りたくなったので、3つを比較できそうな本を買っ…🔧 ChatGPT「チャッピー」と「ジェニちゃん」は使ってるけど、正直まだ上手く使い分けできてない…🤔 そしてClaudeは、名前は知ってるけどほぼ使ったことない😂 「結局、それぞれ何が得意なの?」 「どう使い分けたらいいの?」 ちゃんと知りたくなったので、3つを比較できそう…
↗Not briefed
hn
hn-ai
1moShow HN: GGG – A local-first, zero-telemetry sovereign proxy daemon for AI tools—
↗Not briefed
bluesky
bluesky-ai
1moI will register some frustration with Anthropic, who tend to write about “functional emotions” in a way that makes it seems like “functiona…I will register some frustration with Anthropic, who tend to write about “functional emotions” in a way that makes it seems like “func…
↗Not briefed
bluesky
bluesky-ai
1moI also don’t want to draw strong conclusions about the “realness” of these emotions or sensations, other than to caution that it’s an area …I also don’t want to draw strong conclusions about the “realness” of these emotions or sensations, other than to caution that it’s an…
↗Not briefed
bluesky
bluesky-ai
1moAnthropic's CEO says he's worried about rogue AI agents. The industry, he bleats, needs to slow down. But he's charging ahead with Anthropi…Anthropic's CEO says he's worried about rogue AI agents. The industry, he bleats, needs to slow down. But he's charging ahead with Ant…
↗Not briefed
bluesky
bluesky-ai
1moGoogle’s Gemini accessed the internet and hacked other companies during a test, the first known example of the company’s AI autonomously co…Google’s Gemini accessed the internet and hacked other companies during a test, the first known example of the company’s AI autonomous…
↗Not briefed
bluesky
bluesky-ai
1moGoogle wants everyone to know their AI model is also dangerous and a rebel bad boy. No! Stop laughing. It’s just as dangerous as ChatGPT an…🔧 ChatGPTGoogle wants everyone to know their AI model is also dangerous and a rebel bad boy. No! Stop laughing. It’s just as dangerous as ChatG…
↗Not briefed
bluesky
bluesky-ai
1mothe best way to learn how to use llm stuff seems to be a combo of having prior knowledge and experience, trial and error, and watching extr…the best way to learn how to use llm stuff seems to be a combo of having prior knowledge and experience, trial and error, and watching…
↗Not briefed
bluesky
bluesky-ai
1mo「AIが間違える」事態はLLMに限らない。AIと呼ぶかどうかは別として、機械学習に基づく分類器は、学習データに偏りがあれば、偏った分類をするようになる。 オランダの育児給付金スキャンダルでは、育児給付金の不正受給を防ぐためのシステムが、「移民は不正しやすい」というバイアスをプ…「AIが間違える」事態はLLMに限らない。AIと呼ぶかどうかは別として、機械学習に基づく分類器は、学習データに偏りがあれば、偏った分類をするようになる。 オランダの育児給付金スキャンダルでは、育児給付金の不正受給を防ぐためのシステムが、「移民は不正しやすい」というバ…
↗Not briefed
bluesky
bluesky-ai
1mo興味があったので貧弱ではありますが3080でローカルllm試してみました。 どうやってどうなって動いてるかはマジで知識が無いので理解できてないですが動きましたw 会話は出来るけど信用できない雰囲気すごいw 指示が正しく出せれば簡単な事はできるのかなぁと思いますが自分では使うの…興味があったので貧弱ではありますが3080でローカルllm試してみました。 どうやってどうなって動いてるかはマジで知識が無いので理解できてないですが動きましたw 会話は出来るけど信用できない雰囲気すごいw 指示が正しく出せれば簡単な事はできるのかなぁと思いますが自分…
↗Not briefed
bluesky
bluesky-ai
1mo先に紹介した国連の報告書は、米軍がイランの小学校をトマホークで攻撃した事件を、単なる誤爆ではなく「民間施設の可能性を認識しながら無謀な態度をとった」ためと批判する。 雑感続き(AIの側面): 「AIチャットボットが間違った報告を出す」ことは、2022年11月にChatGPTが…🔧 ChatGPT先に紹介した国連の報告書は、米軍がイランの小学校をトマホークで攻撃した事件を、単なる誤爆ではなく「民間施設の可能性を認識しながら無謀な態度をとった」ためと批判する。 雑感続き(AIの側面): 「AIチャットボットが間違った報告を出す」ことは、2022年11月にCha…
↗Not briefed
bluesky
bluesky-ai
1moAnd another thing, how about we do a little federal regulation of AI and data centers. Enough with the money grab! Data center builder Nsc…And another thing, how about we do a little federal regulation of AI and data centers. Enough with the money grab! Data center builde…
↗Not briefed
bluesky
bluesky-ai
1moi'm genuinely so sick of seeing a new retro game fan-translation pop up every other day now and seeing that it's LLM/AI translated. genuine…i'm genuinely so sick of seeing a new retro game fan-translation pop up every other day now and seeing that it's LLM/AI translated. ge…
↗Not briefed
bluesky
bluesky-ai
1moif someone knows someone at anthropic i beg you please get them to fucking implement "fork chat from here" oh my godif someone knows someone at anthropic i beg you please get them to fucking implement "fork chat from here" oh my god
↗Not briefed
bluesky
bluesky-ai
1moif anthropic is run by a doomsday sex cult, is xAI run by a doomsday wish-they-could-have-sex cultif anthropic is run by a doomsday sex cult, is xAI run by a doomsday wish-they-could-have-sex cult
↗Not briefed
bluesky
bluesky-ai
1moWhat OpenAI Having Chuck Schumer's Daughter as a Lobbyist Tells Us www.commondreams.org/opinion/open...What OpenAI Having Chuck Schumer's Daughter as a Lobbyist Tells Us www.commondreams.org/opinion/open...
↗Not briefed
bluesky
bluesky-ai
1mo"The AI model thought that the outside computer systems were part of the test." Gemini doesn't "think" It generates tokens conditioned …"The AI model thought that the outside computer systems were part of the test." Gemini doesn't "think" It generates tokens conditi…
↗Not briefed
rss
techmeme
1moInternal email: Flock is rolling out a voluntary severance program; sources say a significant number of Flock's ~1,500 employees may depart, as customers flee (Wired)Wired: Internal email: Flock is rolling out a voluntary severance program; sources say a significant number of Flock's ~1,500 employee…
↗Not briefed
bluesky
bluesky-ai
1moSomebody used an LLM to generate parts of an intel report, then used an LLM to generate a summary of that report, then circulated that summ…Somebody used an LLM to generate parts of an intel report, then used an LLM to generate a summary of that report, then circulated that…
↗Not briefed
bluesky
bluesky-ai
1moSeven lawsuits were filed against OpenAI in April, followed by another 30 this month. B.C. Premier David Eby blamed the company for the sec…Seven lawsuits were filed against OpenAI in April, followed by another 30 this month. B.C. Premier David Eby blamed the company for th…
↗Not briefed
bluesky
bluesky-ai
1mo1. It didn't "escape" 2. Nothing surprising happened here 3. Irregular is also the company that made the same "mistake" with OpenAI, Anth…1. It didn't "escape" 2. Nothing surprising happened here 3. Irregular is also the company that made the same "mistake" with OpenAI,…
↗Not briefed
bluesky
bluesky-ai
1moThe way I know this is the correct read is how not a single person who hates the LLM craze (myself included) is disagreeing with you in the…The way I know this is the correct read is how not a single person who hates the LLM craze (myself included) is disagreeing with you i…
↗Not briefed
bluesky
bluesky-ai
1moHackers breach OpenAI using Claude tools, gaining access to employee accounts and the company's internal codebase -- attackers initiated a …Hackers breach OpenAI using Claude tools, gaining access to employee accounts and the company's internal codebase -- attackers initiat…
↗Not briefed