👋 On time for your weekend: a round-up of this week’s remarkable stories at the intersection of technology, business, design, and culture. Three reads and three listens; no fluff, just stuff⚡
️Thoughtforms will be touching grass for the next three weeks, and editions may not turn up as expected during that period. Back to normal service on Friday, September 22nd.
📚 Reading
Open challenges in LLM research:
Never before in my life had I seen so many smart people working on the same goal: making LLMs better. After talking to many people working in both industry and academia, I noticed the 10 major research directions that emerged. The first two directions, hallucinations and context learning, are probably the most talked about today. I’m the most excited about numbers 3 (multimodality), 5 (new architecture), and 6 (GPU alternatives).
Chip Huyen | 17 minutes
"Open" "AI" isn't:
In the Big Tech internet, it's freedom for them, openness for us. "Openness" – transparency, reusability and extensibility – is valuable, but it shouldn't be mistaken for technological self-determination. As the tech sector becomes ever-more concentrated, the limits of openness become more apparent. But even by those standards, the openness of "open AI" is thin gruel indeed…
Cory Doctorow—Pluralistic | 12 minutes
Now is the time for grimoires:
The corporate focus on giving AIs more data before building an infrastructure around using AI misses this point, which is not surprising because the use case of AI is radical: it puts individual workers, not the company, in charge of innovation. Instead, companies should be considering how to build libraries of prompts, grimoires of expert spells that allow practices to be scaled inside the organization
Ethan Mollick—One Useful Thing | 11 minutes
🎧 Listening
Design in Tech 2023: Design and Artificial Intelligence (Abridged)
When you think of design thinking, it's most likely set to be revolutionized [by Generative AI] because it's much easier to do design thinking with AI. You can use AI as a partner in doing your design thinking, and that's going to be disruptive and actually quite useful to a lot of organizations trying to get out of their convergence and diverge just enough to help the customer a whole lot more.
John Meeda | 26 minutes
Straight Talk on A.I. Large Language Models:
[A]re these systems actually using analog reasoning the way that we humans use it? Or are they using some other way of solving the problems? […] When ChatGPT passes a bar exam, can it go out and replace human lawyers? A human who passed the bar exam can do that. But I don't know if you can make the same assumption for a language model, because its reasoning might be quite different and not imply the same kinds of more general abilities.
Melanie Mitchell—Ground Truths | 39 minutes
You should kill your startup, What replaces globalization:
[R]elatively never do we talk about the digital trade imbalance. Our bits, as they flow out of our internet companies, literally hit a brick wall when they get to the Chinese border. And yet, Chinese internet companies can develop enormously huge market caps and monopolies and technologies inside of Chinese borders […] and nobody really talks about it from an economic perspective, that there's an enormous trade imbalance at the bits level of the internet.
Dave Morin—More or Less | 52 minutes
🛌 On the Nightstand
On the Origin of Time by Thomas Hertog. One of those things that intuitively feel right, somehow, but defy rationalizing: the laws of physics evolved with and over time.