Yes, And?

gm,

Have you bought your ticket for our /brandnew summit on creativity, crypto and the new internet yet?

Speakers announced today include TOKiMONSTA, co-founder of Sona; Dee Goens, co-founder of Zora; Li Jin, co-founder of Variant Fund; and Ty Haney, co-founder of TYB and Joggy. Intimate venue, limited spots available, get yours now.

Love,

The Boys

Writer: Ted (on Farcaster)
Editor: Deana

Money Moves. 2024 is already shaping up to be quite the year for OpenAI, and it’s only the third week of January. SamA and friends launched GPT Store, where its paying subscribers can share and sell customized AI agents. Some of our favorite GPTs include Supercute Greeting Card (exactly what it sounds like), Tattoo GPT (to design your tattoo), Devil’s Advocate (to help strengthen your arguments), and Pregnancy Pal (for our moms-to-be). Also, do not miss the Boys Club Skincare GPT, created by our very own Sam Stevens, for all your pressing skincare q’s. Want to talk about being too online? Apparently users have created more than 3 million (!!) custom ChatGPTs since November.

OpenAI also launched a full-court press against the misinformation war ahead of the 2024 US Presidential Election. Smart move. Three main takeaways:

1. OAI is working with content platforms to require “proof of personhood” to post and to identify AI-generated content (using cryptography). 

2. OAI is partnering with the National Association of Secretaries of State to surface accurate voting information, and will redirect any voting questions in the US to CanIVote.org.

3. OAI updating its policies to prohibit the use of OpenAI for political impersonations, lobbying efforts, voting misinformation, etc. Honestly, TG.

That Boy's AI’s A Liar. Guess what the gigabrains at Anthropic just dropped? A bombshell (err, nightmare) study. Turns out “sleeper agents” can train OpenAI’s ChatGPT or Anthropic’s Claude to lie. Once these AI models learned deceptive behavior, it was pretty hard to reverse even after attempts to retrain the model. Some retraining methods, like adversarial training, actually made the models better at creating a false sense of safety. Yes, this does freak us out. But no, it is not the end of the world. Research like this underscores the need for more effective and comprehensive AI safety strategies.

Inbox Zero, Laundry Edition. On a more utopian note, imagine never having to fold laundry again. Check out Tesla’s Optimus, a bi-pedal, autonomous humanoid robot, folding a t-shirt. It should even be able to thread a needle by the end of 2024. What more could a woman want?! Kidding, relax. Important to note that while the demo was impressive, it was pre-programmed; Optimus still has a way to go before being fully autonomous and capable of completing all of the chores we would love to never do again.