Reimagining the structure of neural networks
Plus, reusing jokes, fighting bias in AI, and programming advances
We’re back with another edition of Machine Learnings, brought to you by the folks at Heyday.
Heyday is an AI-powered memory assistant that resurfaces content you forgot about while you browse the web.
What a week.
Apple didn’t mention AI by name at WWDC, but it was embedded throughout their demos. Marc Andreessen did mention it, calling AI the savior. Consider this the follow-up to “It’s time to build.”
Which we’re doing – building.
For the past 6 months, we’ve focused our efforts to embed AI in Heyday beyond the chatbot. AI tools are assistants and copilots, but they may best be served with another analogy. What do you think it is?
If you’d like to experience how we see AI helping you, connect to Heyday for free today.
-@samdebrule
What we're reading.
1/ Hyperdimensional computing. Save that to your brain file. It may just be a radically different way of approaching neural networks. Learn more at WIRED >
2/ What is it about the Singularity that attracts such fanaticism in technologists? Maybe the money? Learn more at The New York Times >
3/ We’ve known that generative AI tools are biased; humans are programming the tools, after all. Now we have data to show us how severe it actually is. Some great visualizations here, as well. Learn more at Bloomberg >
4/ [Listen] Rohin Shah, ML researcher at DeepMind, walks us through the balancing act of doom and boom in artificial intelligence. A balanced, feet-on-the-ground take from someone responsible for maintaining technical AI safety at Google DeepMind. Learn more at 80000 Hours >
5/ The diss track is ubiquitous in hip hop, but what’s the equivalent in technology? In some places, ChatGPT’s moment has turned into lines like “Don’t be an AI.” Learn more at The Atlantic >
6/ More jokes. Researchers just published a piece that showed 90% of ChatGPT’s aims to make us laugh come from the same 25 pre-trained jokes. Sounds like my uncle… Learn more at Ars Technica >
7/ More Google DeepMind – this time on their massive advances to programming. They’ve just published a new sorting method to lists in C++ and a way to accelerate cryptography. This is a fascinating read on how their models operate with innovation at the helm. Learn more at MIT Technology Review >
Bonus! Couple with this from WIRED on DeepMind learning the art of coding.
Research for this edition of Machine Learnings was enhanced by Heyday, the AI-powered memory assistant.