weekend ai reads for 2024-03-29

Author

That AI Thing
March 29, 2024

📰 ABOVE THE FOLD: HEALTHCARE

A new prescription / The Economist

How Digital Twins’ Virtual Realities Will Transform The World — Digital twins offer humankind the ability to command virtual replicas of forests, oil fields, cities, supply chains — and even, maybe one day, our very bodies. / Noema Magazine

Nvidia powers AI nurses Hippocratic AI — The cheap AI agents offer medical advice to patients over video calls in real-time / Quartz

Nvidia announced a collaboration with Hippocratic AI on Monday, a healthcare company that offers generative AI nurses who work for just $9 an hour. Hippocratic promotes how it can undercut real human nurses, who can cost $90 an hour, with its cheap AI agents that offer medical advice to patients over video calls in real-time.

 

📻 QUOTE OF THE WEEK

If OpenAl disappeared tomorrow, we have all the IP rights and all the capability. We have the people, we have the compute, we have the data, we have everything. We are below them, above them, around them.

Satya Nadella, Microsoft CEO (source)

 

🏗️ FOUNDATIONS & CULTURE

Another way we can categorise generative AI systems is between “visible” and “invisible” systems.

The third category, and maybe the most important one, is whether the AI is “heavy” or “light”

  • part one of a four part series on ai and video games

Have We Reached Peak AI? / Edward Zitron, Where’s Your Ed At?

Altman wants to talk about the big, sexy stories of Average General Intelligences that can take human jobs because the reality of OpenAI — and generative AI by extension — is far more boring, limited and expensive than he'd like you to know.

Fairly Trained — We certify fair training data use in Generative AI.

There is a divide emerging between two types of generative AI companies: those who get the consent of training data providers, and those who don’t, claiming they have no legal obligation to do so.We believe there are many consumers and companies who would prefer to work with generative AI companies who train on data provided with the consent of its creators.Fairly Trained exists to make it clear which companies take a more consent-based approach to training, and are therefore treating creators more fairly.

 

🎓 EDUCATION

Fixing Ed-Tech Investing’s Lemons Problem — The sector currently suffers from a version of the classic “lemons problem.” / Stanford Social Innovation Review

  • building on this, investors should:

    • treat edtech differently and prioritize the development and selection of solutions based on rigorous evidence of their impact on learning outcomes, and

    • bridge the knowledge gaps that exist between different parties, ensuring that development and implementation is informed by the needs of learners and teachers, as well as by evidence of what works

Universities build their own ChatGPT-like AI tools — As concerns mount over the ethical and intellectual property implications of AI tools, universities are launching their own chatbots for faculty and students. / Inside Higher Ed

LAUSD developed Ed through a public-private partnership with AllHere, a developer of AI-powered digital applications. The Boston-based ed-tech company won a $6 million contract over five years to guide the LAUSD effort, the Los Angeles Times reported.

I suspect Ed will spend the rest of the school year disappointing optimists and providing ammunition for skeptics. Future headlines and follow-up articles will likely be warnings (not that we need more) about why new technology should be rolled out incrementally and with care.

Though potentially transformative, the benefits of AI integration into K12 education appear distant and speculative. Our current focus is on keeping our heads above water, ensuring our students’ safety and mental well-being, and simply keeping our schools staffed and our doors open.

Domain-Focused Models: Math LLMs / Alex Irina Sandu, Substack (sorry)

The End of Foreign-Language Education — Thanks to AI, people may no longer feel the need to learn a second language. / The Atlantic

  • related (?), Fluently — Speaking Copilot for Non-Native Professionals

 

📊 DATA & TECHNOLOGY

Models All The Way Down / Knowing Machines

  • we appreciate a well-told visual narrative

Investigating training sets is an essential avenue to understanding how generative AI models work; the ways they see and re-create the world.

Scrutinizing these sets is perhaps the only way to get a clear look at the models that are trained on them.

Contrary to what most large AI companies claim, the release of Common Corpus aims to show it is possible to train Large Language Model on fully open and reproducible corpus, without using copyright content.

via adam, Inside the Creation of DBRX, the World's Most Powerful Open Source AI Model — Startup Databricks just released DBRX, the most powerful open source large language model yet—eclipsing Meta’s Llama 2.

  • useful on both sides of the procurement stick

GPT LLM Trainer / mshumer, GitHub

Simply input a description of your task, and the system will generate a dataset from scratch, parse it into the right format, and fine-tune a LLaMA 2 or GPT-3.5 model for you.

  • not quite as simple as advertised, but more simple than many other approaches

  • fully open-source and free

Found means fixed: Introducing code scanning autofix, powered by GitHub Copilot and CodeQL — Now in public beta for GitHub Advanced Security customers, code scanning autofix helps developers remediate more than two-thirds of supported alerts with little or no editing. / The GitHub Blog

 

🎉 FUN and/or PRACTICAL THINGS

Realtime — Today’s Top Data Stories

Our automated engine continuously tracks public data feeds, detects key changes in real-time, and uses AI to distill them into bite-size stories to keep you informed.

GPT4All — A free-to-use, locally running, privacy-aware chatbot. No GPU or internet required.

  • faster than LM Studio, our current local chatbot, but you still get what you pay for

The Promenade — The RPG for AI Adventurers

  • requires signup

Claros — What do you want to buy?

  • uses AI and Reddit (?) to recommend products based on a prompt

AI Wedding Toast — Unique and Memorable Wedding Speeches with AI

  • not everything has to be AI-ified

  • the AI song embedded AI in the article doesn’t seem to work, but is still up here

 

🧿 AI-ADJACENT

9. AI methods that don’t leverage computation.

  • and

Things that work: Dogs, vegetables, index funds, jogging, sleep, lists, learning to cook, drinking less alcohol, surrounding yourself with people you trust and admire.

101 things I would tell my self from 10 years ago / Leila Clark, Approach with Alacrity

29. At lot of your work will involve taking some flow in the world and optimizing some property of them: latency, throughput or bandwidth. To do this, you must figure out what limits them and then remove that bottleneck. This sounds simple, but you will not really understand the depth of this discipline until you work with a master of it.

 

⋄