I have a friendly announcement, at this time of the year, as every year before.
It’s gonna be May 🕺
1
100
Paper naming award goes to this one 💕
Segment Everything Everywhere All at Once
Presents SEEM, a promptable, interactive model for Segmenting Everything Everywhere all at once in an image.
arxiv.org/abs/2304.06718
118
"As extensions of the human brain, they (computers) have been constructed with the same property of error, spontaneous, uncontrolled, and rich in possibilities."
Samuel Butler’s quote I found in today's @brainpicker newsletter. 💜
34
This is the byproduct of the acquaintance of living in a world where we have passively accepted that eventually, technology will have an answer for every question.
A widely spread lack of curiosity.
Biggest takeaway from the @sama-@lexfridman episode today was a creeping sense of eudread*
My new working definition of AGI now stems from the observation that we test progress by asking our AIs questions, but we are running out of questions to which we know the answer.
When we are only left asking it questions to remaining mysteries of the universe, and it gives us retrospectively obviously correct answers, it will clearly and irreversibly be over.
*a positive form of AGI dread, after stress-eustress
50
Of course what you’re saying it’s important honey 😰🫣
Live footage of *not* being on call
xitter.gq/marjo_beb/status/16395…
58
Marketing after hearing everyone complaining
We’re no longer sunsetting the Free Team plan docker.com/blog/no-longer-su…
2
104
Dreambooth/Lora Level Models With 5 Training Steps, Train in Seconds Rather Than Minutes
The paper -> arxiv.org/abs/2302.12228
#StableDiffusion2 / #StableDiffusion #AIArt #lora
254
Llama running inference (7B params) on a AWS c7g.2xlarge ($0.289/hr)
predict time = 50629.78 ms / 99.08 ms per token 🔥
52
Found this stream, AI Simpsons.
Characters are recreating the episode "Steamed Hams" where Skinner tries to convince Chalmers there's no fire in the kitchen. Except there's a completely new plot every N minutes generated by AI
twitch.tv/unlimitedsteam
2
121
🎉🎉🎉
Google AI just announced the PaLM API!
It will be released with a new tool called MakerSuite, which lets you prototype ideas, do prompt engineering, synthetic data generation and custom-model tuning. Waitlist available soon.
developers.googleblog.com/20…
22
One of the reviewers on the Llama project approved the changes to merge the PR "Save bandwidth by using a torrent to distribute more efficiently #73" 😄
github.com/facebookresearch/…
68
Detecting Deepfakes by Studying ‘Disconnected’ Facial Expressions
metaphysic.ai/detecting-deep…
1
20




















