📟 lens: @ isekai.lens + ens: samizdat.eth limited backup social graph: @isekai@sigmoid.social
NYC
Joined May 2008
- Tweets23.4K
- Following3.9K
- Followers1.5K
- Likes26.1K
Pinned Tweet
It's always been mine, for as long as I can remember. To lucid dream while awake and interact with the waking world with as much fluency as when in more programmable environments –– dreams, imagination, virtuality, synesthesia, image synthesis... you name it
jk 🌻(📼, 🌿) retweeted
The internet feels increasingly broken. News sites are paywalled or account walled, Reddit is nag walled, Google search spams ads and SEO to the point of uselessness, and now Twitter is account walled. Web browsing feels horrible now.
jk 🌻(📼, 🌿) retweeted
Google has released an AI for generating music.
Simply describe the music you want and a track is created.
This music AI is available for free.
Here's how to access and use it:
jk 🌻(📼, 🌿) retweeted
at today's Sen #AIhearing, Sam Altman & Gary Marcus have proposed a massive #AI regulatory regime for U.S. that involves a new bureaucracy to license just about anything that lawmakers consider “dangerous,” complete with prior restraints on algorithmic innovation on the front end + mandatory regulatory audits on back end. And then an international regulator on top of it all.
This is a truly terrible, destructive model for AI that must be rejected.
jk 🌻(📼, 🌿) retweeted
In just 60-something words, SEC Commissioner @HesterPeirce has managed to articulate all that hits awfully close to home. Thank you.
jk 🌻(📼, 🌿) retweeted
Non-blue-check people no longer appear in “For you” starting tomorrow. Posts with images also have had their reach dropped. It was nice knowing you all
jk 🌻(📼, 🌿) retweeted
On the old networks, you're either paying to exist or you're just algorithm food. Decisions, decisions...🤖
jk 🌻(📼, 🌿) retweeted
“Four things stand out in the recent history of this wonderful century. They are the automobile, flying machine, moving picture and radio.
And greatest of all, in the effect it is destined to have upon the human race, is the moving picture.”
1925
jk 🌻(📼, 🌿) retweeted
The big takeaway for me is that instead of being an app that prioritizes breaking news content or, at the very least, quickly developing trends, Twitter is now Reddit moving at the speed of Tumblr. What goes viral here feels more like what goes viral everywhere else.
jk 🌻(📼, 🌿) retweeted
This paper used Stable Diffusion to generate 175 million images over 350,000 prompts and only found 109 near copies of training data. Am I right that my main takeaway from this is how good Stable Diffusion is at *not* memorizing training examples?
Models such as Stable Diffusion are trained on copyrighted, trademarked, private, and sensitive images.
Yet, our new paper shows that diffusion models memorize images from their training data and emit them at generation time.
Paper: arxiv.org/abs/2301.13188
👇[1/9]
jk 🌻(📼, 🌿) retweeted
great paper, but what an odd summary
of 175.000.000 generated images, they managed to recreate 109 "near copies" of pics in the training set, that's 0,000062%?
alt headline: it's lottery-level difficult (but not impossible) to recreate images from the training data in SD 🤷♂️
Models such as Stable Diffusion are trained on copyrighted, trademarked, private, and sensitive images.
Yet, our new paper shows that diffusion models memorize images from their training data and emit them at generation time.
Paper: arxiv.org/abs/2301.13188
👇[1/9]
jk 🌻(📼, 🌿) retweeted
“Are GPT3-based apps defensible?” is similar to asking “Are SQL database apps defensible?”
jk 🌻(📼, 🌿) retweeted
Crypto is here to stay. I will keep advocating for policies that advance crypto innovation and adoption in the United States because crypto is more than a financial investment: it’s about restoring liberty and choice to individuals.
coindesk.com/consensus-magaz…
jk 🌻(📼, 🌿) retweeted
It will take time and evolution for the reptilian self to feel ok enough to cooperate and adjust to new technology #stableaiart #stablediffusionart #stablesurreal
jk 🌻(📼, 🌿) retweeted
Software has been basically deterministic until now
Until we fully trust LLM outputs, we will start to see a UX I call "staging"
Here are a few examples
