Explorar

阿里刚刚上新了全新：Qwen3-TTS（ 2025-11-27），支持10种语言+9种方言比上版，在自适应根据文本调节语速和韵律的能力上有提升，拟人化程度更逼真一些在MiniMax TTS multilingual test set上，WER优于MiniMax、ElevenLabs、GPT-4o-Audio-Preview 音色加到了49种 #TTS #Qwen3TTS

博客：https://t.co/Kt2dBodLsA Realtime API: https://t.co/vD8EN95sxu Offline API: https://t.co/DQRQjDm2jA

AIGCLINK

Fri Dec 05 23:46:28

RT @every: @hammer_mt tested it on invoice extraction. The data was a mess—it had different formats and labels that didn't match. - Starti…

Asst professor @MIT EECS & CSAIL (@nlp_mit). Author of https://t.co/VgyLxl0oa1 and https://t.co/ZZaSzaRaZ7 (@DSPyOSS). Prev: CS PhD @StanfordNLP. Research @Databricks.

Omar Khattab

Fri Dec 05 23:45:59

RT @every: The chef analogy: You wouldn't cook the full meal 100 times to perfect a recipe. You'd test tiny tasting bowls, note what went w…

Asst professor @MIT EECS & CSAIL (@nlp_mit). Author of https://t.co/VgyLxl0oa1 and https://t.co/ZZaSzaRaZ7 (@DSPyOSS). Prev: CS PhD @StanfordNLP. Research @Databricks.

Omar Khattab

Fri Dec 05 23:45:52

RT @every: What if you could breed better prompts, like a chef perfecting a recipe? @hammer_mt walked through how a system called GEPA dou…

Asst professor @MIT EECS & CSAIL (@nlp_mit). Author of https://t.co/VgyLxl0oa1 and https://t.co/ZZaSzaRaZ7 (@DSPyOSS). Prev: CS PhD @StanfordNLP. Research @Databricks.

Omar Khattab

Fri Dec 05 23:45:25

If you tried to send me, a rando, into battle with a whiz bang rifle that had gyros & computer aiming etc I would expect to die & instead beg you for an AK or an AR (of recent, quality make). It’s not a game & the tolerance for failure is zero.

Author. Coder. CTO. θηριομάχης. Building: https://t.co/otXT4Wy6WR. Writing: https://t.co/dBPBtyCIHw.

Jon Stokes

Fri Dec 05 23:45:14

Demis DISAGREES with me. TL;DR: «the gaps is maybe months, but even leaving chips aside, no Chinese labs or models have demonstrated the capability for innovating algorithms beyond the state of the art. Only fast-following» I wonder. What is this state of the art then?

We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1

Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)

Fri Dec 05 23:44:53

Newest first — browse tweet threads

Explorar

Newest first — browse tweet threads

RT @every: @hammer_mt tested it on invoice extraction. The data was a mess—it had different formats and labels that didn't match. - Starti…

RT @every: The chef analogy: You wouldn't cook the full meal 100 times to perfect a recipe. You'd test tiny tasting bowls, note what went w…

RT @every: What if you could breed better prompts, like a chef perfecting a recipe? @hammer_mt walked through how a system called GEPA dou…

If you tried to send me, a rando, into battle with a whiz bang rifle that had gyros & computer aiming etc I would expect to die & instead beg you for an AK or an AR (of recent, quality make). It’s not a game & the tolerance for failure is zero.

Demis DISAGREES with me. TL;DR: «the gaps is maybe months, but even leaving chips aside, no Chinese labs or models have demonstrated the capability for innovating algorithms beyond the state of the art. Only fast-following» I wonder. What is this state of the art then?

Explorar

Newest first — browse tweet threads

Explorar

Newest first — browse tweet threads

RT @every: @hammer_mt tested it on invoice extraction. The data was a mess—it had different formats and labels that didn't match. - Starti…

RT @every: The chef analogy: You wouldn't cook the full meal 100 times to perfect a recipe. You'd test tiny tasting bowls, note what went w…

RT @every: What if you could breed better prompts, like a chef perfecting a recipe? @hammer_mt walked through how a system called GEPA dou…

If you tried to send me, a rando, into battle with a whiz bang rifle that had gyros & computer aiming etc I would expect to die & instead beg you for an AK or an AR (of recent, quality make). It’s not a game & the tolerance for failure is zero.

Demis DISAGREES with me. TL;DR: «the gaps is maybe months, but even leaving chips aside, no Chinese labs or models have demonstrated the capability for innovating algorithms beyond the state of the art. Only fast-following» I wonder. What *is* this state of the art then?

Demis DISAGREES with me. TL;DR: «the gaps is maybe months, but even leaving chips aside, no Chinese labs or models have demonstrated the capability for innovating algorithms beyond the state of the art. Only fast-following» I wonder. What is this state of the art then?