Weekly Link Roundup #39
Copypasta is low in carbs but will never save you..
OpenAI unveils easy voice assistant creation at 2024 developer event: I think this moves us way closer to the Star Trek “computer” than Alexa or Siri - “Perhaps the most notable new API feature is the Realtime API, now in public beta, which supports speech-to-speech conversations using six preset voices and enables developers to build features very similar to ChatGPT's Advanced Voice Mode (AVM) into their applications.”
The Viral ‘Goodbye Meta AI’ Copypasta Will Not Protect You: Really people? Are we still doing this?
Twitch just added another path to become Partner. Will it let established streamers help up-and-comers?: Let’s pump the brakes on AI for just like one story. I still think there is more to gain from internal creator economies. Why not create a platform that people who aspire to be creators can use? They get an instant network, you get unpaid creators (except for the ad revenue share and you can be ridiculously generous because that’s not the money you’re here for). Think about it. “Twitch‘s path to partnership is not an easy one. Any streamer who wants to become an official partner of the platform has to meet a strict set of requirements: Within one 30-day period, they have to stream on at least 12 different days, stream 25+ hours of content, and maintain an average of 75 viewers across all their live time.”
OpenAI’s newest tool feels less like a chatbot, more like Google Doc: Think its interesting that from the start of this AI explosion, the main innovations have been in UI and features - the performance of the underlying models have improved sure but that’s not what’s driving people to the platforms. Its the UI. “OpenAI has just introduced Canvas, a new tool designed to take ChatGPT beyond simple chat interactions and into a more collaborative workspace for writing and coding. Unlike the traditional chat window, Canvas opens a separate work area with a chat window on the right side, allowing users to work side-by-side with ChatGPT, refining and editing ideas directly within their documents or code.”
Direct ocean capture may be the next frontier for carbon removal: This looks promising and what I can’t understand is, even if you leave all the good intentions aside, how people don’t see the heaping gobs of money to be made cleaning the world up - “The industry standard for removing carbon with direct air capture methods costs anywhere between $230 to $630 (approximately €210 to €570) per metric ton, according to the International Energy Agency. However, Amsterdam-based Brineworks, a company specializing in seawater electrolysis technology, says its innovative method is expected to cost under $100 per ton of CO2 at scale. This would put it in a pretty efficient space compared with other methods.”
A professor of learning science explains how failure helps you learn: I am going to flippant here - fair warning - and maybe I should read the book discussed here but on its face, this sounds like YET ANOTHER person/researcher DISCOVERING something that game designers have know since the invention of GO (or longer). Wow - stunning insight that people can learn through failure! Games are built on that principle. Raph Koster, game designer extraordinaire, points out in his exceptionally fine book - A Theory of Fun for Game Design - that learning is what’s addictive in games. So, I guess thanks for “discovering” that all over again.
Accenture To Train 30,000 Staff On Nvidia AI Tech In Blockbuster Deal: That seems like a lot of people. “The global services powerhouse says its newly formed Nvidia Business Group will focus on driving enterprise adoption of what it called ‘agentic AI systems’ by taking advantage of key Nvidia software platforms that fuel consumption of GPU-accelerated data centers.”
Larger and more instructible language models become less reliable: Maybe the future is smaller > > “We also find that early models often avoid user questions but scaled-up, shaped-up models tend to give an apparently sensible yet wrong answer much more often, including errors on difficult questions that human supervisors frequently overlook.”
The Most Capable Open Source AI Model Yet Could Supercharge AI Agents: “Released today by the Allen Institute for AI (Ai2), the Multimodal Open Language Model, or Molmo, can interpret images as well as converse through a chat interface. This means it can make sense of a computer screen, potentially helping an AI agent perform tasks such as browsing the web, navigating through file directories, and drafting documents.”
Fake AI “podcasters” are reviewing my book and it’s freaking me out: I’m telling you, I see things in heat maps and for the longest time OpenAI ate al the air, Claude and artifacts stormed back a bit but right now, the rising star is Notebook LLM > > “There are still enough notable issues with NotebookLM's audio output to prevent it from fully replacing professional podcasters any time soon. Even so, the podcast-like format is an incredibly engaging and endearing way to take in complex information and points to a much more personable future for generative AI than the dry back-and-forth of a text-based chatbot.”
Zillow will now show climate risks for property listings in the US: How very timely.
World Wide Web Foundation closes so Tim Berners-Lee can spend more time with his protocol: What a mocking headline. Don’t really expect much better from the The Register but it is what it is. Godspeed Sir Tim “The Solid Protocol is "a specification that lets people store their data securely in decentralized data stores called Pods," as the project's website explains. Pods stands for "personal online data stores," which can be hosted by an individual or provider. They're essentially databases of social graph data, user-created files, and online activity information that can be made available to apps and services under a permission model similar to what's available on mobile phones.”
College students used Meta’s smart glasses to dox people in real time: “Two Harvard students have created an eerie demo of how smart glasses can use facial recognition tech to instantly dox people’s identities, phone numbers, and addresses. The most unsettling part is the demo uses current, widely available technology like the Ray-Ban Meta smart glasses and public databases.”
An Interview with Meta CTO Andrew Bosworth About Orion and Reality Labs: Great interview (per usual) but this caught my eye - “What follows is unadulterated praise. Orion makes every other VR or AR device I have tried feel like a mistake — including the Apple Vision Pro. It is incredibly comfortable to wear, for one. What was the most striking to me, however, is that the obvious limitations — particularly low resolution — felt immaterial.” I’ve always thought AR, not voice, was the killer app especially for glasses and while these are ready to roll out yet (they cost about $10K/pair) - the review is powerful. »See also: Meta’s big tease.
Microsoft Copilot can now read your screen, think deeply, and speak aloud to you: All jokes aside, this could be HUGE for accessibility. Have you ever tried to listen to a screen reader? “Copilot Vision has a view of what you’re viewing on your PC — more specifically, a lens into the sites you’re visiting with Microsoft Edge. Gated behind Copilot Labs, a new Copilot Pro-exclusive opt-in program for experimental Copilot capabilities, Copilot Vision can analyze text and images on web pages and answer queries (e.g., “What’s the recipe for the food in this picture?”) about them.”
The Medieval Masterpiece, the Book of Kells, Is Now Digitized and Available Online: “The ancient masterpiece is a stunning example of Hiberno-Saxon style, thought to have been composed on the Scottish island of Iona in 806, then transferred to the monastery of Kells in County Meath after a Viking raid.”
Google Earth now lets you see how the world looked 80 years ago — here’s how: So cool “In its update announcement, Google says some cities, including London, Berlin, Warsaw and Paris, will have imagery dating back to the 1930s. As of right now, most locations on Google Earth go no further back than just a few decades ago. The new changes will double the time frame for even those locations.” > > Makes me think about what else it would be cool to be able to roll the clock back on.
Desert Racers Demolish Art Carved by Ancient People in Chile: Hey look! Idiots! “What many of those racers potentially ignore is that the Atacama was once a canvas for ancient Indigenous peoples of South America. Starting 3,000 years ago, those Indigenous people carved vast figures of animals, humans and objects on the desert’s slopes. Known as geoglyphs, the specimens at Alto Barranco in the Tarapacá region stand out for their remarkable preservation.”
ChatGPT's rise linked to decline in public knowledge sharing on online Q&A platforms: Predictable really - “A new study published in PNAS Nexus reveals that the widespread adoption of large language models (LLMs), such as ChatGPT, has led to a significant decline in public knowledge sharing on platforms like Stack Overflow. The study highlights a 25% reduction in user activity on the popular programming Q&A site within six months of ChatGPT's release, relative to similar platforms where access to ChatGPT is restricted.”
Omni, the Iconic Sci-Fi Magazine, Now Digitized in High-Resolution and Available Online: Some people had Sports Illustrated or Nat Geo or Time - my magazine drug of choice was OMNI. It was edgy, it was cool, nothing else looked like it. I remember whole issues. Before WIRED even thought about being WIRED...there was OMNI.
Arc creator Josh Miller on why you need a better browser than Chrome: I remember a time when I tried out and was running several browsers. I was actually an early and PAYING customer of Opera, I have a Firefox messenger bag, browsers are we consume so much of the information we see on the daily. To be sure, its less now with apps and walled gardens but browsers are still there and still important and could be a key to going back to a more open web. “Basically, Arc is a ground-up rethinking of the web browser. Most modern browsers started as simple document viewers and grew to support running complex apps. Arc’s main conceit is that it’s designed to make running and using all those apps as simple as possible.”
AI digests repetitive scatological document into profound “poop” podcast: Will Poop and Fart Be The New Turing Test? “Imagine you're a podcaster who regularly does quick 10- to 12-minute summary reviews of written works. Now imagine your producer gives you multiple pages of nothing but the words "poop" and "fart" repeated over and over again and asks you to have an episode about the document on their desk within the hour.”