Staff Engineer at https://teapaw.org/t.co/5gRXZa7a7f. Previously Head of R&D and CTO co-founder at https://teapaw.org/t.co/yC0WgVCmUb and Research Group Leader in Computer Vision and Natural Language Processing.
Rocky Mountain House, Canada
Joined September 2013
- Tweets302
- Following832
- Followers238
- Likes1.3K
I'm doing an experiment in whether AI, equipped with R&D skills, can bootstrap a company. The website is CraterView.ai, a web app and API to enhance, denoise, and restore photos.
Born from models I've been working on since university and a mandate ...
1
61
Bernardt Duvenhage retweeted
🚨 VEED Fabric 1.0 is now 30% FASTER 🚨
Create ultra-realistic AI talking head videos…in less time than ever.
API now live on @replicate
Comment “VEED” for access in DMs
Bernardt Duvenhage retweeted
You needs a LipSync API?
go try it
Its on @FAL
Bernardt Duvenhage retweeted
UGC AI coming to VEED soon.
Our research team has done an amazing job building a zero shot dubbing model.
Excited to roll this out in the product in the coming months.
Pipio.ai is working on something interesting with the awesome people at @AmiiThinks and we're looking for an ML intern to help us out. If you have experience in text to speech, voice style transfer, etc. please consider applying to amii.bamboohr.com/careers/77
1
94
Bernardt Duvenhage retweeted
On behalf of the entire Google Research & @GoogleAI communities, I'm excited to share an overview of some of our research in 2020.
Thanks to everyone who helped make this work possible!
ai.googleblog.com/2021/01/go…
Bernardt Duvenhage retweeted
The ✨ 2nd AfricaNLP Workshop ✨ will be taking place at #EACL2021! They have a magnificent programme 🤩
Please do submit 🌠 So we can build together 🌍
📅 Submission date: 5 Feb 2021
📝 Short (4pg) or long (8pg) papers
👇🏿All details within👇🏿
sites.google.com/view/africa…
Please help flesh out the African NLP map by completing the survey at decoloniale.typeform.com/to/… .
We are looking for an exceptional Site Reliability Engineer / DevOps Engineer ... praekelt.bamboohr.com/jobs/v…
Bernardt Duvenhage retweeted
ICYMI - The keynote interview with @shakir_za is available now! #twimlfest
Bernardt Duvenhage retweeted
In this special #TWIMLfest edition of the podcast, we’re joined by @Shakir_Za Mohamed, a Senior Research Scientist at @DeepMind, and leader of @DeepIndaba to discuss decolonizing AI, the origin of the Indaba, the phases of community, and much more. twimlai.com/decolonizing-ai-…
I really enjoyed "How to win Kaggle competitions with Anthony Goldbloom" Gradient Dissent (by @wandb) episode soundcloud.com/wandb/anthony…. They discuss how Kaggle has become a professional sport and which skills are developed by competing and following the competition forums
2
Bernardt Duvenhage retweeted
Read about changes to conference registration this year, which opens just days from now!
medium.com/@NeurIPSConf/neur…
Bernardt Duvenhage retweeted
Thank you @bigdata for having me on the Data Exchange Podcast! I've enjoyed the past episodes so it was fun to chat about Human-in-the-Loop Machine Learning in the newest one!
thedataexchange.media/human-…
I'm still working through the papers I added to my reading list at NeurIPS :-)
A cool paper on multilingual language models for translation and other tasks is papers.nips.cc/paper/8928-cr… from Facebook research. Implementation on github. Likely useful for @MasakhaneMt community.
1
This was great! For everyone that couldn’t attend I can recommend watching the #NeurIPS2019 keynote talk by @blaiseaguera
A GAN is a society of two. Locally each "actor" is doing gradient descent. Put together, the combined system is not doing gradient descent...
and life is like this at all levels (e.g. cells within a body or bacteria where the environment is each other) @blaiseaguera #NeurIPS2019
ALT Vector field with cycles for Wasserstein GAN. Locally, each "actor" is doing gradient descent of its own well-defined loss function. Put together, the combined system is not doing gradient descent, nor can we in general write down a loss function that is being optimized... ...and life is like this at all levels
1
5
Finished 'Rebooting AI' by @GaryMarcus & Davis. While it is clear that many problems remain to give machines 'common sense', I liked the discussions around trust and sound engineering principles in ML. For me it is also a good complement to 'The Book of Why' by Pearl & Mackenzie.