"Hack the AI" - Adam Craven (@HackTheAI) header
"Hack the AI" - Adam Craven (@HackTheAI) profile picture
Brisbane, Queensland
Joined December 2018
Photo by "Hack the AI" - Adam Craven (@HackTheAI)Photo by "Hack the AI" - Adam Craven (@HackTheAI)
"Hack the AI" - Adam Craven retweeted
New research summary: with better masks & ventilation, & rapid tests, we could halt the outbreaks in NSW and Vic. 1/🧵 fast.ai/2021/09/03/covid-r/
12
37
11
112
"Hack the AI" - Adam Craven retweeted
Presenting DreamerV2, the first world model-based #ReinforcementLearning agent to achieve top-level performance on the Atari benchmark, learning general representations from images to discover successful behaviors in latent space. Read more at goo.gle/3ub8iZe
17
254
34
935
"Hack the AI" - Adam Craven retweeted
Introducing NFNets, a family of image classification models that are: *SOTA on ImageNet (86.5% top-1 w/o extra data) *Up to 8.7x faster to train than EfficientNets to a given accuracy *Normalizer-free (no BatchNorm!) Paper: dpmd.ai/06171 Code: dpmd.ai/nfnets
16
321
44
1,406
"Hack the AI" - Adam Craven retweeted
We’ve developed two neural networks which have learned by associating text and images. CLIP maps images into categories described in text, and DALL-E creates new images, like this, from text. A step toward systems with deeper understanding of the world. openai.com/multimodal/
90
899
328
2,872
"Hack the AI" - Adam Craven retweeted
Neural Filters is a major breakthrough in AI-powered creativity and the beginning of a complete reimagination of filters and image manipulation inside @Photoshop 🤩: adobe.ly/3jhxcj2 #AdobeMAX
168
1,928
2,118
8,059
"Hack the AI" - Adam Craven retweeted
Interested in working though all of the previously shared #AtHomeWithAI resources? We’ve compiled a master list of the researcher recommendations - see here: https://teapawt.net/t.co/GqJKWAQ6MM
11
276
14
728
"Hack the AI" - Adam Craven retweeted
Attention & memory have recently emerged as two vital components of DL. In this lecture, research scientist Alex Graves covers a range of contemporary attention mechanisms - looking at variable computation time, transformers & more. https://teapawt.net/t.co/q8WmJ7WR79 #AtHomeWithAI
4
126
6
464
"Hack the AI" - Adam Craven retweeted
dm_control is our Python toolkit for physics-based reinforcement learning tasks. We've released new models 🤖🕷🐕 & a brand new tutorial - see below! Github: git.io/dm_control Tutorial: https://teapawt.net/t.co/4ggH7Hp60n Blog: https://teapawt.net/t.co/vDQpeKs8Je Videos: https://teapawt.net/t.co/8ACgtQOskQ
7
220
12
886
"Hack the AI" - Adam Craven retweeted
We found that just as a large transformer model trained on language can generate coherent text, the same exact model trained on pixel sequences can generate coherent image completions and samples. openai.com/blog/image-gpt/
57
712
158
2,853
"Hack the AI" - Adam Craven retweeted
Gated Linear Networks (GLNs) are backpropagation-free neural nets designed for efficient online learning via local convex optimisation. We show their performance, interpretability & robustness to catastrophic forgetting on classification tasks - read here: https://teapawt.net/t.co/8NTNDkXZDT
4
227
14
852
"Hack the AI" - Adam Craven retweeted
We've made a draft of our deep learning book available as @ProjectJupyter notebooks. Important: This is for personal use, and not licensed for any redistribution or change of format or medium. Eg. don't make a PDF and share it, please! github.com/fastai/fastbook
55
835
55
2,840
"Hack the AI" - Adam Craven retweeted
I stopped doing CV research because I saw the impact my work was having. I loved the work but the military applications and privacy concerns eventually became impossible to ignore. teapawt.net/RogerGrosse/status/123…
Replying to @skoularidou
What's an example of a situation where you think someone should decide not to submit their paper due to Broader Impacts reasons?
74
771
189
2,894
"Hack the AI" - Adam Craven retweeted
Free Python course materials! I put in a FOIA request to the NSA for their Python training materials and got back a 400-page printout of their COMP 3321 training course. So, I scanned and OCR'd it. Here is a PDF (warning: 118 MB) nsa.sfo2.digitaloceanspaces.…
127
2,262
294
6,427
"Hack the AI" - Adam Craven retweeted
We're standardizing OpenAI's deep learning framework on PyTorch to increase our research productivity at scale on GPUs (and have just released a PyTorch version of Spinning Up in Deep RL): openai.com/blog/openai-pytor…
32
524
69
1,858
"Hack the AI" - Adam Craven retweeted
Given the smoothness of videos, can we learn models more efficiently than with #backprop? We present Sideways - a step towards a high-throughput, approximate backprop that considers the one-way direction of time and pipelines forward and backward passes. arxiv.org/pdf/2001.06232.pdf
7
233
21
914
"Hack the AI" - Adam Craven retweeted
The v3 of the course (part 1) is now fully ported to fastai v2 (and by ported, it's not just the code runs but the results match or exceed the ones from the version on fastai v1). All notebooks are here: github.com/fastai/fastai2/tr…
8
61
4
385
"Hack the AI" - Adam Craven retweeted
Training data is often collected through a biased process. Models trained on such data are inherently biased. We demonstrate how adversarial training through disentangled representations can reduce the effect of spurious correlations present in datasets: arxiv.org/abs/1912.03192
1
133
10
377
"Hack the AI" - Adam Craven retweeted
#NeurIPS2019 - the 33rd Conference on Neural Information Processing Systems - is now under way! We’re looking forward to this week’s tutorials, workshops, and sessions. Come say hello to the team @ our booth. We'll post schedules and updates here throughout the week!
4
40
2
238
"Hack the AI" - Adam Craven retweeted
We introduce LOGAN, a game-theory motivated algorithm, which improves the state-of-the-art in GAN image generation by over 30% measured in FID: arxiv.org/abs/1912.00953 Here are samples showing higher diversity:
12
312
12
933