Weekly Link Roundup #44
This one really has everything....
Since there is still time for gift-buying, I thought I’d include this guide to Indigenous-owned businesses you can shop from and ask, if anyone knows more, then please feel free to add them via comments. Cheers! > > The Indigenous-Owned Gift/Business Issue.
Generative Agent Simulations of 1,000 People (PDF link): What would it take for you to rely on agent simulations? > > “We present a novel agent architecture that simulates the attitudes and behaviors of 1,052 real individuals—applying large language models to qualitative interviews about their lives, then measuring how well these agents replicate the attitudes and behaviors of the individuals that they represent. The generative agents replicate participants' responses on the General Social Survey 85% as accurately as participants replicate their own answers two weeks later, and perform comparably in predicting personality traits and outcomes in experimental replications. Our architecture reduces accuracy biases across racial and ideological groups compared to agents given demographic descriptions.”
Superhuman Knowledge Workers? AI Exoskeletons and Scaffoldings: “Scaffolding is primarily designed as a temporary support, focusing on cognitive assistance and learning processes. It is designed to encourage a more reflective and iterative approach to problem-solving.” and “In contrast, AI exoskeletons are more persistent enhancement systems. They aim for continuous augmentation of both cognitive and analytical capabilities across various domains without necessarily intending to be removed over time.”
World Labs’ AI can generate interactive 3D scenes from a single photo: The kicker > “Lots of AI systems can turn a photo into 3D models and environments. But World Labs’ scenes are unique in that they’re interactive — and modifiable.”
Data brokers may be banned from selling your social security number: Can we file this under JUDAS H PRIEST ON A PONY THAT TOOK FREAKING LONG ENOUGH? I don’t know, maybe if you systemic vulnerabilities that say leak about 200 million SSNs, then maybe you need a little more regulation. See also: FTC Bans Location Data Company That Powers the Surveillance Ecosystem.
Anthropic bets on personalization in the AI arms race with new ‘styles’ feature: Not that this is different from how humans learn to communicate, but as with all things AI, this runs the risk of enabling mistakes at scale and speed. Who designed the presets? Why don’t we want unanticipated styles and response patterns? Is this patter recognition or pattern lock-in? “The new “styles” feature, launching today on Claude.ai, enables users to preset how Claude responds to queries, offering formal, concise, or explanatory modes. Users can also create custom response patterns by uploading sample content that matches their preferred communication style.”
The Mega-Universities Disrupting Higher Education: I think ASU is the leader here in terms of maintaining their resident campus reputation and growing its online presence and reputation. “This shift is giving rise to an increasing number of “mega-universities" — massively scaled institutions that are rewriting some of the rules of higher education.”
ElevenLabs’ new feature is a NotebookLM competitor for creating GenAI podcasts: I did this with the NotebookLM (audio below)
“Voice AI startup ElevenLabs on Wednesday introduced a feature that lets you upload different types of content to create a multispeaker podcast for you, similar to Google’s NotebookLM.”
AWS SageMaker is transforming into a combined data and AI hub: Get ready for waves of this - backend consolidation of data assets with front end AI. And no shocker here but Amazon finally wading into the retail side of GenAI - Amazon launches Nova AI model family for generating text, images and videos. See also this > > Emergence’s AI orchestrator launches to do what big tech offerings can’t: play well with others.
Amazon teams up with Orbital to remove CO2 from the air at one of its data centers: Don’t bury the lede: “And if data centers do end up capturing more carbon dioxide than they generate, Amazon and other companies can sell the credits themselves, turning the system into a profit center.”
Ex-Microsoft employees get $4M from Accel to build an AI tool for product presentations: If you don’t see another blow to content production for L&D, you’re not paying attention. “Product teams often have tons of screen recordings and screenshots that end up unused. Video production using those screen recordings has been time-consuming and expensive. That’s why two former Microsoft employees started to build Lica, which is an AI tool that easily creates tutorials and product videos out of screenshots and screen recordings.”
Liquid AI’s new STAR model architecture outshines Transformer efficiency: OK, this may feel a little abstract but how many of you reading this newsletter read Attention is All You Need? This could be one of those papers/developments “Now, MIT-birthed startup Liquid AI has introduced STAR (Synthesis of Tailored Architectures), an innovative framework designed to automate the generation and optimization of AI model architectures. The STAR framework leverages evolutionary algorithms and a numerical encoding system to address the complex challenge of balancing quality and efficiency in deep learning models.”
Anthropic proposes a new way to connect data to AI chatbots: At the start of this Cambrian Integration and Consolidation, developments like this can be key (if they stick): “Anthropic is proposing a new standard for connecting AI assistants to the systems where data resides. Called the Model Context Protocol, or MCP for short, Anthropic says the standard, which it open sourced today, could help AI models produce better, more relevant responses to queries.”
Enterprises are struggling with what to do with Gen AI, say venture capitalists: They should be struggling. And experimenting. And trying stuff. And finding out what doesn’t work and thinking of new places where AI could help. I wouldn’t trust anyone who says they have it all figured out. This % should be higher - some people were lying “More than a third of our survey respondents do not have a clear vision for how generative AI will be implemented across their organizations," write the authors of the report, Menlo Ventures partners Tim Tully and Joff Redfern, and investor Derek Xiao..”
How Alok Is Transforming Ancestral Sounds Into a Global Call for Sustainability: “The Brazilian superstar, the governor of Pará, and Indigenous leaders explore groundbreaking ways to integrate traditional wisdom with modern environmental advocacy ahead of COP30.”
A lollipop interface for simulating taste in virtual environments: How about no? “A team of biomedical engineers and virtual reality researchers has developed a lollipop interface to simulate taste in a virtual environment. They have published their research in Proceedings of the National Academy of Sciences.”
The New York Times Announces its Best Books of 2024: Will check out this list but wanted to give kudos to Bookriot for saying “its” versus the NYT’s declarative statement of just “the 10 best” - its not hard to add that this list is your own opinion and that other lists will look different and that that’s OK. Sheesh. /end rant >More book lists: Lit Hub’s 38 Favorite Books of 2024. More > These are the Most Read Books of the 2024 Goodreads Reading Challenge.