"The hottest new programming language is English.” – Andrej Karpathy, former AI head at Tesla
In Today's Meteor
- Create: Ukrainian artist escapes the war
- Think: AI music is getting really good
- Compress: If you sold a gorilla pic for $900K would you quit your day job?
- Disrupt: What would AI Jesus do?
After fleeing war and a surgery that left her bedridden, she found AI art
Olena is a Ukrainian photographer who fled to Turkey when the war began. An operation left her bedridden for three months, unable to shoot photos. She turned to AI art.
"All I had was my imagination and my fingertips," she said. "AI opened a new door. For me it's like clean air."
Olena's images are both beautiful and haunting. She says it's an expression of her inner mind, but not about the conflict. "I decided not to make art about the war. It's too hard for me."
"Blockchain artist" wants Koreans to be free to say what they really think
A new social art experiment is attempting to reveal how Koreans think and what they are willing to say anonymously (which could be mostly commentary about mixed martial arts by the looks of things).
Conceived by pseudonymous "blockchain artist" Peti Triot, the project features an online messenger that anyone can use to enter text and have it published on YouTube and Twitter stripped of any identifying metadata. The project promises to make the source untraceable using Qamon, a messaging platform operating on the Everscale blockchain that advertises anonymous p2p communication. A real world version will be installed in April in Seoul's Desiego gallery.
Have artists and art projects we should feature? Send them to firstname.lastname@example.org
Dancing with the prompts
A team at Google has pulled off an incredible new text-to-music AI. Just type in a phrase to create irresistible melodies and beats. The samples will blow your mind.
MusicML joins a crowded field, but sets a new standard. OpenAI’s Jukebox, Amper, Aiva and others have been available publicly for years, and have achieved good results using inputs like genre and style. Aiva has written an “Opus 1 for Piano Solo,” released an album and composed music for a video game. Due to training data and other limitations, none before have been able to reliably produce complex music from nothing more than a text prompt.
By contrast, MusicLM delivers off of something like this:
The main soundtrack of an arcade game. It is fast-paced and upbeat, with a catchy electric guitar riff. The music is repetitive and easy to remember, but with unexpected sounds, like cymbal crashes or drum rolls.
Although it can glitch sometimes, it's hard to exaggerate how good the results are.
There's more. MusicLM can be fed combinations of pictures and captions, or be directed to generate audio that sounds like a given instrument in any genre, and even specify the experience level of the “musician.” Places, historical eras or situational awareness (e.g. music for morning mediation) can all be used to alter the outputs.
It won't rival ChatGPT in popularity any time soon, but that's only because there are no plans to release it publicly at the moment, and no code was included in the paper describing it. As with other generative AI products, copyright is a concern, with about 1 percent of the music created by MusicML referencing recognizable songs, according to the team.
"A flying dog with a red cape"
Meta AI has a new generative AI, MAV3D (Make-A-Video3D), that turns text prompts into full 3D motion video (aka 4D). It is believed to be an industry first.
This was inevitable
The NFT project Kevin Rose thought he was buying into when he got hacked
Porsche lawyers have opened a can of NFTs
Purchasers of the car maker's 911 mint were asked to waive their 14-day window for requesting a refund, which is the default under the law. This guy didn't.
We put this under work, but it is really about how not to work.
If you sold a gorilla picture for $900,000 would you quit your day job?
The human army that's feeding AI
To cite a bot, or to not cite a bot?
Some do, some don't
Crypto is as safe as you want it to be. With a steady news beat of OG crypto investors getting ripped off, you'd think putting your NFTs in 18 different cold storage wallets is like parking a Ferrari in Times Square with the ignition running. There are many ways to make sure you don't fall victim to scams, you just have to learn about them and use them. Drive a car? Wear a seat belt. Ride a motorcycle? Wear a helmet. Trade in crypto?
1) Use a cold storage wallet, like Ledger, and keep your seed phrase separate and secure.
2) Only use a hot wallet when you are actively trading.
3) Use a hot wallet with pro-active security features, like Coinbase Wallet, which offers Web site scam alerts, seaport alerts, and transaction previews so you can see what will happen if you execute a smart contract before you sign it. (The Chrome extension Fire Wallet advertises similar protections, and while it seems safe to try it has not been thoroughly vetted.)
4) Use Revoke.Cash to reverse transactions and limit wallet permissions that allow dApps like NFT marketplaces to access your funds.
5) Separate your assets into multiple wallets so if one gets compromised the damage will be limited.
It's risky out there, learn how to be safe.
What Would AI Jesus Do?
Ask him, via Bible GPT.
Thanks for reading! Send your thoughts and comments to email@example.com. We are accepting story pitches. Hit us up at firstname.lastname@example.org.