I'm Sherman Chann (main), a former infosec/SWE guy now working as a Machine Learning Engineer at ElevenLabs.
Some time in early 2022 -- around the time stuff like Copilot or Gato or Clippy were published, I realised what the important people had figured long ago: AGI is coming, nothing else matters, drop everything and work on this.
So, I did. Not in the strongest sense, but by the time ChatGPT rolled around, I had made:
- Some TalkNet TTS models.
- Minor contributions and friends from the Stable Diffusion community
- Some basic fastapi wrappers around LLMs for Code generation and writing from huggingface
After that, I
- wasted about a month attempting to build a GPT-3-enabled AI tutor.
- Made some discord bot wrappers for multimodal LLMs like OpenFlamingo and MiniGPT-4
- Involved myself with ML discord/twitter a lot. Eleuther/Nous.
- Improved 🐢 TorToiSe-TTS with fast inference && fine-tuning
Many TTS companies reached out to me for my work on Tortoise, and I eventually accepted a few offers. Although I'm deeply interested in the frontier space of LLM development, working for private startups has mostly hindered me from making any public contributions to that space since ~April.
If you're interested in the work I was doing (webdev, CTF, general software engineering, competitive programming, game dev, etc) prior to 2022, you can read a more comprehensive account at my about page.