OpenAI is nothing without its people
Jong Wook Kim 💟
357 posts
Jong Wook Kim 💟
@_jongwook_kim
Member of Technical Staff @OpenAI; previously at @nyuMARL, @SpotifyResearch, @pandoramusic, @kakaocorpglobal, and @NCSOFT
- i love the openai team so much
- 1.7x fewer tokens in Korean, which means GPT-4o feels 3.4x faster to Korean users!
- I can finally say that my voice in Whisper's ICML presentation video icml.cc/virtual/2023/p… was AI-generated 😄
- Whisper is now available in Hugging Face transformers! Happy to learn that it's become more accessible to the 🤗 community and hoping it enables lots of amazing applications!⏰➡👆We just released Whisper in 🤗 transformers! @openai’s latest speech recognition transformer trained on 680 000 hours of audio! For example use case, check this notebook : colab.research.google.com/drive/16HO7if9…
- Replying to @seeingwithsound and @ilyasutThe smallest CLIP model is available at github.com/openai/CLIP. It’s larger than typical models for mobile, but I think it can fit and run (slowly) on mobile devices.
- MusicNet Inspector: Browse the MusicNet dataset and explore how the labels look like while enjoying the music! (feedback welcome) musicnet-inspector.github.io
- CLIP is secretly a music generation model!Prompt: "a mel-spectrogram of an electric guitar playing" (VQGAN+CLIP) Won't sound great but will throw it into Griffin-Lim... You can kinda see little guitar shapes 🤣
- The largest GPT-2 model is now LIVE! So happy to have participated in the detection work which marks my first (co-authored) publication at @OpenAI :D Check out the new blog post and the report for our latest findings and thoughts.We're releasing the 1.5billion parameter GPT-2 model as part of our staged release publication strategy. - GPT-2 output detection model: github.com/openai/gpt-2-o… - Research from partners on potential malicious uses: d4mucfpksywv.cloudfront.net/papers/GPT_2_R… - More details: openai.com/blog/gpt-2-1-5…
- GPT-4o is way better at optical character recognition in non-English / non-Latin script languages!これは実用性高い 左:gpt-4o 右:gpt-4
- Make TFRecords Usable Again!Replying to @keunwoochoiNow available with a file-level iterator that works without tensorflow :D github.com/jongwook/tfrec…
- Visualizing the MusicNet dataset for sanity check... I need to find the best way (for a machine) to learn music from these
00:00 - So excited to attend @ismir2019 and present my latest work at @nyuMARL! Flying to the Netherlands today🚀
