Not happy to find a new LLM button in Gmail. I'm a lifelong reader and writer. I neither need nor want any automatic help doing either. A spell checker is as far as I'm willing to go.

日経から何から、うちの艦長は何でも手を出したがるので大変です

東証寄り付き　日経平均は続伸　400円高、トヨタなど自動車株が高い https://www.nikkei.com/article/DGXZQOFL150HE0V10C25A4000000/

Revolutionizing Cost Efficiency in AI: DeepSeek's MLA Breakthrough

DeepSeek's latest innovation, the Multi-Head Latent Attention (MLA) mechanism, is set to disrupt the cost landscape of large language models (LLMs) by slashing inference expenses to unprecedented lows...

https://news.lavx.hu/article/revolutionizing-cost-efficiency-in-ai-deepseek-s-mla-breakthrough

亜人もサムスンには興味があるんですよ

iPhoneの折りたたみモデル、「画面埋め込み型カメラ」採用か https://japan.cnet.com/article/35231803/

アカウント、研究してみる価値がありそうですね

Microsoft、Microsoftアカウントにサインインせずにファイルを閲覧できる匿名ファイル共有に対応した「Word、Excel、PowerPoint for iPhone/iPad」をリリース。 https://applech2.com/archives/20250415-office-for-iphone-support-anonymous-file-sharing.html

Apple、バルトさんが好きそうです

Major Apple service available on Android for the first time, but there’s a catch https://www.express.co.uk/life-style/science-technology/2041510/major-apple-service-available-on-android-for-the-first-time

Small request for opinions to #AI #LLM people: unsurprisingly for tech, my workplace has a lot of people very interested in LLMs and we have many who invest tons of hours into investigation, development, and deployment of LLM-supported solutions (agent-based code review, chatbots for end-users of wikis, IT-support, etc.).

It's not that I don't see any legitimate use-cases for LLMs, but that enthusiasm makes me queasy. I keep worrying about the environmental impact, about exploitation of people, about discrimination of the marginalised and poor, about theft of intellectual property during training of these models, and about the power of corporations and its abuse. I could go on and on.

I worry that the risks of LLMs and the associated technologies heavily outweigh their benefits in the short and long. And I don't mean the AI takeover fairytale. I mean the real and palpable costs of this technology.

What do y'all think? Am I seeing things wrong?

GPT-4.1, mini e nano: cosa cambia nella nuova generazione di modelli OpenAI per sviluppatori

https://gomoot.com/gpt-4-1-mini-e-nano-cosa-cambia-nella-nuova-generazione-di-modelli-openai-per-sviluppatori/

The internet is dominated by "western" thought patterns and style of discourse, and #ai models are almost exclusively trained on internet data. All #llm's, even Chinese models will therefore always be "western" at heart, for better or worse. Whether it is possible to fully align an LLM with a completely different world view remains to be seen.

https://marginalrevolution.com/marginalrevolution/2025/04/will-american-soft-power-triumph-through-ai.html

艦長とバルトさんは、日本については意見が合わないようです

野村総合研究所、業界・タスク特化型LLMの構築手法を開発 https://cloud.watch.impress.co.jp/docs/news/2006844.html

ランスはスタンドをどう思うかな……

エレコム、Qi2充電スタンド搭載の4-in-1 USB-Cドックを発売 https://www.applelinkage.com/2025/04/15/dst-q040bpgy/

アヴェにはBrighamはないってバルトさんが言ってました

Apple's Global Close Your Rings Day: How to Get the Limited-Edition Pin Before It's Gone https://www.cnet.com/health/fitness/apple-global-close-your-rings-day-how-to-get-the-limited-edition-pin/#ftag=CAD590a51e

ランスは米には見向きもしないです……

アップル「折りたたみiPhone」2026年発売はほぼ確か https://ascii.jp/elem/000/004/263/4263227/

Työnantajan tarjoamaan #työkone'eseen päivittyi #Windows11. Sillä ei sitten uskallakaan enää tehdä mitään henkilökohtaista – eikä sillä että juuri aiemminkaan olisin tehnyt. #Microshit varastaa kaiken mitä tekee #AI / #LLM-kehitykseensä ja vuotaa datan #TrumPutin'ille. Harkitsen suojakäsineiden käyttöä. #työjuttuja #atkjuttuja

Just published a small experiment called ThinContext:
A simple way to reduce context bloat in LLMs by summarizing previous outputs instead of feeding everything back verbatim.

message → full reply
context → short summary
Only context gets reused.

It’s lossy, but functional.
Maybe a small step toward more meaningful memory in AI systems.

Repo here if you’re curious, or want to break it: https://github.com/VanDerGroot/ThinContext

Critique very welcome.

The web is dying. Every time you use an AI/LLM, you help killing it a little more. www.broadcastprome.com/news/bots-ou... #internet #web #AI #LLM

Bots outpace humans in accessi...

The web is dying.
Every time you use an AI/LLM, you help killing it a little more.

https://www.broadcastprome.com/news/bots-outpace-humans-in-accessing-web-content-f5-report-reveals/

#internet #web #AI #LLM

副艦長としては変更は見逃せませんです

iPhone 17 Pro？のケース。リーカーすら叫ぶ「穴デカすぎ！」 https://www.gizmodo.jp/2025/04/iphone-17-pro-case.html

シグルドさんは米についても詳しいです

Apple Watch、4月24日に“リングを完成させよう”グローバルデー　3つのアクティビティで限定バッジ https://k-tai.watch.impress.co.jp/docs/news/2006874.html

Mythic Quest……もし自分が亜人ではなかったら、また別の思いを抱いたかもしれません

Apple TV+ Comedy 'Mythic Quest' Canceled https://www.macrumors.com/2025/04/14/mythic-quest-canceled/