Ruining the Internet
AI alignment thru open -source, OpenAI's Developer Day, and the people who ruined the internet
The past is written, but the future is left for us to write. And we have powerful tools, openness, optimism, and the spirit of curiosity – Capt. Jean-Luc Picard
👋🍿 Your weekly snack of the most remarkable stories at the intersection of technology, business, design, and culture. Three reads and three listens; no fluff, just stuff ⚡
📚 Reading
Open source AI has been vital for alignment:
To make progress, we need people who are able to tinker and play with and iteratively improve the alignment of models. This includes, necessarily, having access to the weights and free ability to alter them. We see extremely clear evidence of this thus far. Almost all valuable alignment work has been done by direct hands-on experience of models
Beren Millidge | 12 minutes
What I Saw at OpenAI’s Developer Day:
ChatGPT was born in sin. When OpenAI started, its goal was to serve developers—until it accidentally created the biggest consumer application of all time. Unfortunately, this puts the company at odds with developers because ChatGPT competes directly with much of what developers want to build—both at the consumer layer and the infrastructure layer.
Dan Shipper-Every | 15 minutes
The people who ruined the internet:
Over time, SEO techniques have spread and become insidious, such that googling anything can now feel like looking up “sneaker” in the dictionary and finding a definition that sounds both incorrect and suspiciously as though it were written by someone promoting Nike (“footwear that allows you to just do it!”) […This] practice seems to have successfully destroyed the illusion that the internet was ever about anything other than selling stuff.
Amanda Chicago Lewis—The Verge | 43 minutes
🎧 Listening
Hardcore AI for History:
The tricky thing is I think it scales up expectations. What AI is able to do with an assignment is probably going to become the minimum. If you have someone who is unable to achieve kind of the level of GPT 3.5 on a given task, it's unlikely that that person will be successful in that job. What we need to do is teach people to use this in such a way that they exceed the baseline level of the model.
Mark Humphries—Cognitive Revolution | 90 minutes
AI Regulation, War and the Future of the American Tech Stack:
Who's to know what is what and what is real and what is not? Right now, there's no way to know the identity of who's behind anything that's going on. And so that could already be happening and we wouldn't know it. And I think that's the most simple way to think about this is like, what are you seeing? What are you reading? What are you watching on video? And what is the identity behind it? It's like a really simple and easy question to think about.
Dave Morrin—More or Less | 56 minutes
Gen Z Is Built Different:
I think org structures with Millennials got flattened out. It's hard to separate it from the sort of new ideas about work that were championed by tech companies […] a lot of work trends have emanated from Silicon Valley. Everyone has imitated Silicon Valley, I feel when it comes to work. Silicon Valley pioneered the campus; I think they made it, particularly with Google, really a continuation of college. That was almost the promise. I'm surprised they didn't put up dorms.
Brian Morrissey—People vs Algorithms | 51 minutes
🎁 Listening
We’ll be roaming London next week, from Tuesday evening till Thursday afternoon. Let us know if you want to catch up and talk shop over a hot beverage ☕️🍵