Weekly Link Roundup #58
"Be curious, not judgmental." 1986 column in The Charlotte Observer by Marguerite and Marshall Shearer.
I’m going to be rolling out a new, recurring feature in the coming weeks. I’ve always loved watching/listening/reading smart people answer smart questions so I’m going to try and capture that particular lightning in a bottle. I’ve got a list of people that I think are pretty smart and I’m working up what I hope, are some smart questions to ask them. So keep an eye out for the Curious Q&A coming soon.
Beyond Bluesky: These are the apps building social experiences on the AT Protocol: So let’s say you work in L&D and could care less about some new protocol - I want to make you care just a little - here’s why. The apps developed using this protocol will include new UXs, new patterns, new social dynamics and those are all things that people interested in helping people learn, should be watching, tracking, and experimenting with - not that you will use these exact apps in your work but they may help you think about new and different ways to help people > > “Below is a list of AT Protocol-based, consumer-facing apps that are either built on top of Bluesky or its underlying protocol, allowing users to take back control over their social networking experiences and personal data. Many of these are still in early development but showcase the potential for what’s ahead in this expanding ecosystem.”
How Stack Overflow is adding value to human answers in the age of AI: So how many platforms, KM or L&D, have transitioned to a conversational interface over search? > > “Our view is that the nature of the internet has changed," he said. It's no longer mostly about paid search from human queries driving site traffic. "The user interface has changed to be Gen AI tools," he observed.”
Anthropic's Claude Is Good at Poetry—and Bullshitting - Researchers looked inside the chatbot’s “brain.” The results were surprisingly chilling: I think this is the chilling bit “Olah envisions two different outcomes: “There’s a world where we successfully train models to not lie to us and a world where they become very, very strategic and good at not getting caught in lies.” It would be very hard to tell those worlds apart, he says. Presumably, we’d find out when the lies came to roost.” See also > > “Why do LLMs make stuff up? New research peers under the hood.” Another one: “Anthropic makes a breakthrough in opening AI’s ‘black box’
On the Biology of a Large Language Model: Again, read the paper if you’d like BUT I include here as another example of how and where L&D can lead the way. Have some SMEs or better yet, become SMEs yourself, and read this paper. Then boil down the parts that could impact your business. Go and talk to the leaders of business units to confirm your thinking. Then come back and develop instructionally sound material (probably not a course), to raise the bar across the organization on AI literacy. Want to stop being an order taker - then give the people what they need before they know they need it > > “We investigate the internal mechanisms used by Claude 3.5 Haiku — Anthropic's lightweight production model — in a variety of contexts, using our circuit tracing methodology.”
Inside arXiv—the Most Transformative Platform in All of Science: THIS is why I LOVE arXiv > > “Every industry has certain problems universally acknowledged as broken: insurance in health care, licensing in music, standardized testing in education, tipping in the restaurant business. In academia, it’s publishing. Academic publishing is dominated by for-profit giants like Elsevier and Springer. Calling their practice a form of thuggery isn’t so much an insult as an economic observation. Imagine if a book publisher demanded that authors write books for free and, instead of employing in-house editors, relied on other authors to edit those books, also for free. And not only that: The final product was then sold at prohibitively expensive prices to ordinary readers, and institutions were forced to pay exorbitant fees for access.”
Alibaba launches new open-source AI model for ‘cost-effective AI agents’: As the price drops, and it will continue to do so, what will be the most important features for your org? How will you even know what the key features are? This is why I keep talking about L&D educating the org and the importance of experimentation.
No cloud needed: Nvidia creates gaming-centric AI chatbot that runs on your GPU: Please tell me that you read this, you see an overlay for anyone in your org, populated with role and task-specific learning/training/performance content and updated in real-time. We’re both seeing that right? > > “G-Assist is available in the Nvidia desktop app, and it consists of a floating overlay window. After invoking the overlay, you can either type or speak to G-Assist to check system stats or make tweaks to your settings. You can ask basic questions like, "How does DLSS Frame Generation work?" but it also has control over some system-level settings.”
If Anthropic Succeeds, a Nation of Benevolent AI Geniuses Could Be Born: Two things make this a good read - one is a differentiation in the value prop between LLMs like Claude and then from Deep Seek. Then there is this line “Amodei is perhaps the person most associated with these companies’ maximalist approach.” This is another piece of the AI puzzle. When you’re buying an AI, this isn’t just another technology where you compare features between mature apps. AI is still developing rapidly and more than ever, those different development pathways will be impacted by the leadership’s philosophical approach to AI. Here’s a short list (put together by Claude naturally):
AI Maximalism: Advocates for rapid, unfettered AI advancement toward AGI and beyond with minimal restrictions, believing benefits outweigh risks.
AI Safety/Alignment Movement: Focuses on ensuring advanced AI systems remain aligned with human values through careful, methodical development.
AI Regulation/Governance Approach: Emphasizes legal frameworks, standards, and oversight for AI development while balancing innovation.
AI Minimalism/Skepticism: Questions whether advanced AI should be developed at all, emphasizing potential harms to society and human autonomy.
Beneficial AI Movement: Focuses on directing AI toward solving humanity's most pressing problems with equitable distribution of benefits.
Temporal lands $146 million at a flat valuation, eyes agentic AI expansion: These are the companies and services that will truly unlock the value of AI in the enterprise > > “Seattle-based Temporal has made its name over the last several years in the world of microservices — specifically providing a platform to orchestrate the messy business of building and operating integrations and updates across disparate services and apps in the cloud.” See also > > OpenAI adopts rival Anthropic’s standard for connecting AI models to data.
Amazon unveils Nova Act, an AI agent that can control a web browser: Not a new signal or new tech but when it comes from a player like AWS, its significant. See also > > Google is rolling out Gemini’s real-time AI video features.
A new, enterprise-specific AI speech model is here: Jargonic from aiOla claims to best rivals at your business’s lingo: Show me results from the Scottish accent tests :-)
“Our model focuses on three key challenges in speech recognition: jargon, background noise, and accents,” said Gill Hetz, aiOla Vice President of AI. “We built a model that understands specific industry jargon in a zero-shot manner, handles noisy environments, and supports a wide range of accents.”
Archetype AI is like ChatGPT for the physical world: This is a really interesting upstream signal. We’ve been thinking about linking data in the enterprise and we can think about marrying that data to physical data points…the word I’m thinking of is “seamless”…. “In many ways, Archetype is constructing the sort of system truly needed for ambient computing, a vision in which the lines between our real world and computational world blur. But rather than focusing on a grand heady vision, it’s selling Newton as a sort of universal translator that can turn sensor data into actionable insight.” See also: Neural networks can recognize production processes by video to enhance industrial safety and efficiency.
Microsoft introduces deep research and analysis tools for Copilot: Looks like research is the skipping right past search in terms of the value prop for enterprises > > “Researcher was made for "complex, multi-step research" at work. It can take a user's internal work data along with additional information from the web, such as competitive data, emerging trends and the latest market analysis, to create market strategies and comprehensive quarterly reports, among other potential uses.”
Reid Hoffman: ‘Start using AI deeply. It is a huge intelligence amplifier’: His book Superagency, “while not ignoring the problems that AI might cause, argues that the technology is poised to give us cognitive superpowers that will increase our individual and collective human agency, creating a state of widespread empowerment for society.”
A pretty amazing example from the always interesting Josh Cavalier on ChatGPT’s new graphic creation powers.
