Opens profile photo
Follow
dave
@dmvaldman
Thinking about language and AI. Formerly co-founder . Angel investing.
San Francisco, CAdmvaldman.github.ioJoined June 2009

dave’s Tweets

One step closer to text-to-cat gif
Quote Tweet
Tune-A-Video: One-Shot Tuning of Image Diffusion Models for Text-to-Video Generation is out github: github.com/showlab/Tune-A
Embedded video
0:11
11.7K views
3
Show this thread
An underrated feature of MusicLM is that is can condition on humming. Consider the editing version of this: you're a composer, and the trumpet is coming in wrong. Not "ba BUM ba BUM" you say, but "BA bum BA bum". Again, from the top!
Quote Tweet
MusicLM: Generating Music From Text abs: arxiv.org/abs/2301.11325 project page: google-research.github.io/seanet/musiclm
Show this thread
Embedded video
2:16
77.8K views
1
12
"In societies where modern conditions of production prevail, all of life presents itself as an immense accumulation of spectacles. Everything that was directly lived has moved away into a representation." - Society of the Spectacle (1967)
Quote Tweet
Midjourney is getting crazy powerful—none of these are real photos, and none of the people in them exist.
Show this thread
Image
Image
Image
Image
14
A great bit for anyone in the generative AI space. From in his recent podcast. What good may look like between open-ended and constrained textual UIs. [Sorry for this cringe video format 🫣. Wish Twitter did audio files!]
13
The value of starting early in AI is that I'm already well past the "but surely it won't be able to do that" phase. Need to remind myself to have empathy when I meet my former self out there.
3
90
Hmm, a spectrum perhaps? 🤔 RPC = no semantics REST = constrained semantics GPT = freeform semantics
Quote Tweet
"GPT is all you need for backend". This was the most inspirational project from the hackathon over the weekend, hard to stop thinking about. LLM is a kind of equivalent of the Python interpreter, except it interprets English, and has knowledge and common sense. twitter.com/DYtweetshere/s…
4
One year later, Sam tweets of having the highest talent density at OpenAI he's ever seen.
Quote Tweet
OpenAI’s chief scientist: expresses curiosity/openness about a mysterious idea, caveats with “may”. Meta’s chief AI scientist: the certainty of "nope". Probably explains a lot of the past 5 years. Dear Meta AI researchers: My email address is sama@openai.com. We are hiring!
Show this thread
5
Now when I watch engaging content I think to myself "how would I do this with generative AI?", "what prompts would I use?" and it helps me notice how I am manipulated by humans on a daily basis. The creation of tension, connection, emotion when of course everyone is just acting.
2
11
The impact of AI on code may actually be one of the most underrated. Imagine how many more ppl might "stick around" if the learning curve was not: No, no, no, no, no, no... yes! And instead more like: Your code is doing X. Was Y intended? Not quite, but try this! You got it!
7
53
Show this thread
What a great time :) pics in the thread. All we did was tweet a time and place! Such potential energy in SF right now.
Quote Tweet
Got something like 50-100 people in the first 30 min! Piñata yet unharmed, @kipperrii brought her bunny, plenty of pizza left
Show this thread
Image
Image
Image
Image
3
26
Show this thread
thanks so much for bringing us all together today!!! it goes to show the excitement and energy of ml folks that everyone hung around even as the air dropped to mind numbingly cold temperatures (54 f)
Quote Tweet
SF AI meetup today! Alamo Square Park, 3:30pm (precise location below) Co-hosted with @nearcyan and @dmvaldman. Come one, come all - many folks from Scale AI hackathon coming We have pizza and a piñata
Show this thread
Image
Image
Image
1
8
Happening in 1hr! Hope to see ya there!
Quote Tweet
SF AI meetup today! Alamo Square Park, 3:30pm (precise location below) Co-hosted with @nearcyan and @dmvaldman. Come one, come all - many folks from Scale AI hackathon coming We have pizza and a piñata
Show this thread
Image
Image
Image
2
13
Show this thread
It was really cool that a bunch of strangers came together for a day to build over a shared vision. And we were 1 of 8 finalists out of over a hundred teams!
2
18
Show this thread
Come hang out Sunday you AI and AI-curious peoples ♥️
Quote Tweet
// SF meetup, this Sunday // AI meetup in SF this Sunday, jointly with @nearcyan and @dmvaldman Time: 3:30pm Location: Alamo Square Park Come one, come all!
Show this thread
Image
4
13
If you can put images and text into the same embedding space (like CLIP does), then you should be able to take the derivative of an image with respect to a word. That's how I think image editing will work eventually. And every other modality as well. The calculus of AI.
5
24
What patterns are there to how Copilot builds its context window that can be reused for "Copilot for X"? Seems like an art when it might be a science? Maybe an LLM where the context is a mix of app-specific tokens and natural language will be the norm?
2
10
Somewhere, someone is quietly working on Engelbart's "Mother of All Demos" for AI, where every human-computer interface is rethought from first principles
28
737
Show this thread
I heard no matter or energy escaped the training of GPT4. If that's compressed down to an Azure ND A100 v4 node we may be in serious trouble.
Image
10
310
Nick ran a huge chunk of the ML teams , across research and production. I've come to expect this level of quality and attention to detail. Can't wait to see where booth.ai goes to push the limits of generative AI.
Quote Tweet
Visual content is 40x more likely to get shared on social media. So, how do you scale it? This generative AI can create hundreds of photos of your product with one click: booth.ai Try it free today.
Image
Image
Image
Image
5
It's not that there's no moat in AI, it's that features aren't products anymore.
4
40
Some blame this weather on global warming, but such displays of awesome power can only stem from the anger of a spiteful God against the arrogance of
10