動詞 (@IMG_5955) header
動詞 (@IMG_5955) profile picture

エンタメAIをやってます|TTS、LLM、動画生成モデルの学習やサービングなど|ジェットコースター、カラオケ、ホラー作品、宗教、歴史、サイケロックが好き

Joined October 2023
Photo by 動詞 (@IMG_5955)Photo by 動詞 (@IMG_5955)Photo by 動詞 (@IMG_5955)Photo by 動詞 (@IMG_5955)Photo by 動詞 (@IMG_5955)Photo by 動詞 (@IMG_5955)Photo by 動詞 (@IMG_5955)Photo by 動詞 (@IMG_5955)Photo by 動詞 (@IMG_5955)Photo by 動詞 (@IMG_5955)
以前作った「PDFのレイアウトを保ったまま日本語に翻訳するツール」を発展させて、論文読みをめちゃくちゃ楽にするMacアプリを作りました! ご自身のChatGPTのサブスクを介して各種AI機能を利用するので追加費用はかかりません。ぜひ使ってみてください! 🔗: doushi.me/minisco/
5
65
5
727
413,990
動詞 retweeted
The Lean FRO and Mathlib Initiative have released a joint statement on the Lean verification of the Riemann Hypothesis. On November 20, Levent Alpöge released a proof of the Riemann Hypothesis as a Lean library of 17 million lines of code. Over the past day, the Lean and Mathlib communities worked together to verify it. The statement certifies the proof as correct and describes how it meets all necessary verification criteria. Congratulations to Levent on this achievement, and thank you to everyone in the community who contributed to the verification. Full statement: leanprover-community.github.…
544
5,302
2,080
28,444
3,377,300
人間の脳が張りうる空間の中で劣化せずに表現できるような証明は存在しないが、問いだけは立てることは可能な範囲の問題って星の数ほどあると思っている、ので、どんどん人間にとって理解不能な証明が出てきてほしいし、それを検証することがどんどん人間を介さない状態になってほしい
1
1
7
721
全ての宇宙の謎を人間が損失なく理解できるわけがあるか?と思う いくら簡潔にしても100TBぐらいある証明とかあると思う
2
2
267
引っ越し後の段ボールまみれの散らかった部屋ってエモいよなと思うので、しばらく片付けをせずにエモがってたら次第に魔法が解けてきて普通に汚い部屋という感じ方になってきた
14
648
Some of you thought that Inkling 2 Small was too big at 594B parameters. We heard you: Inkling 2 Mini is a looped transformer with just 26B parameters, small enough to run on your laptop, powerful enough for agentic work with a score of 20.1 on AA Index. thinkingmachines.ai/news/ink…
680
1,776
1,083
19,257
3,653,428
Can an image generator learn its own hierarchy of abstractions without a separately trained encoder? Meet HiT. Trained end to end from pixels, it learns to dynamically group patches across multiple levels—and semantic structure emerges without masks or region labels. Check our video below: background collapses into a few groups, the main subject gets many, and words light up the regions they refer to! It also gets 1.48 FID on ImageNet 256, uses 2.4× less training compute than JiT-H/16 at matched quality, and outperforms models with 6× more parameters on average across text-to-image benchmarks. Website, paper, and code below, followed by how it works and what emerges 🧵
13
80
18
653
78,658
楽天証券で一等当選してなくてガチで悔しい
1
13
983
GPT-6.2 Astra、相変わらずネチネチネチネチしてるからレビューには本当に向いてるな。絶対に実装はしてもらいたくない
1
5
67
3,806
動詞 retweeted
新しいsuicaこれで行こう
9
1
74
3,356
Opus 5.5が英語話しちゃう問題について、日本語でもう一回言い直させるのをhookで自動化できたので共有 zenn.dev/discus0434/articles…
19
1,298
動詞 retweeted
弊社Kotoba Technologiesが音声AI最大手のElevenLabsに買収されることになりました! 自分自身は引き続き音声AIの研究開発&エンジニアリングをやっていくことになりそうです とっても楽しみ!!!!!!!!
Today, I’m incredibly excited to announce that Kotoba (@kotoba_tech) is joining ElevenLabs (@ElevenLabs) to push the boundaries of voice AI in Japan and East Asia. From day one, Kotoba’s mission has been to build world-class AI models and products from Japan that make a meaningful difference in people’s lives. I’m incredibly proud of what we’ve built over the past three and a half years—from our work on Agentforce and voice AI models powering products at major global companies to our collaboration with YOASOBI and the popular Kotoba translation app. In that time, we grew from two people to a team of 15 and raised more than $23 million, all while staying lean and focused on bringing our research into people’s hands. As proud as I am of what we’ve achieved, I’m even more excited about what we can accomplish with ElevenLabs. With Kotoba’s research talent and deep understanding of the region alongside ElevenLabs’ global leadership in voice AI, we’re building a true dream team! Voice AI is opening up new ways for people to interact with AI and connect across languages. People and businesses in Japan and East Asia deserve the very best of it—and we’re committed to delivering it. Together, we’re taking a major step toward building a frontier AI lab in Tokyo: developing world-class voice AI models for a global audience and accelerating localization to bring the best voice AI products to Japan and East Asia. A huge thank you to Mati (@mati) and Piotr(@dabkowski_piotr) for envisioning this partnership with us. And to the Kotoba team, our investors, partners, friends, and families—thank you for believing in us and supporting us over the past three and a half years. Chapter 2 starts now! Nori
22
29
1
423
20,376
最近雨多すぎて出勤するとき二度寝して様子を見るか損切りして死ぬかで毎回迷う
4
411
自作パーソナルエージェントのバックエンドをCodex App ServerからPi Durableに変えた。Codexの$200解約してClaude Ultra x2にするかな〜
パーソナルエージェントのエンジンとして使う程度でもそろそろ(Opus 5.5やGemini 4と比べて)GPT-6.1は時代遅れな性能してる感じしてきたなー
1
9
1,141
と思ったけどPi DurableはAPI経由でしかClaude使ってくれないので、ClaudeのAgent SDKを使うしかないらしい moat築こうとすな❗️
1
2
266
マジ!?!?
. @Kotoba_tech, the leading voice AI lab in Japan, is joining ElevenLabs. Together, we’ll bring Kotoba’s Japanese voice research to ElevenAgents, so more businesses can serve customers naturally in Japanese. Welcome to ElevenLabs, @noriyuki_kojima, @jungokasai, and team.
5
541
パーソナルエージェントのエンジンとして使う程度でもそろそろ(Opus 5.5やGemini 4と比べて)GPT-6.1は時代遅れな性能してる感じしてきたなー
2
2
20
3,919
変なXYプロッタが作りたいという動機で少しずつ作っていた、リング式XYプロッタが完成しました!  ※わりと耳障りな音が出ますので注意。 紹介動画はOpus 5.5に作ってもらったものです。解説版の長い動画、解説記事のリンクはリプに貼る予定です。
🤖 Made with AI
48
1,187
180
4,816
596,519