weekend ai reads for 2024-09-20

September 20, 2024

note: trying out a slightly different organization method below

📰 ABOVE THE FOLD: VIDEOS

Help I’m genuinely invested in the Gordon Ramsey AI cooking series / Justine Moore, Twitter (sorry) (1 minute video)

  • the swim trunks are the best

  • try it: MiniMax

  • maybe click this link in a private browser window

Lionel Messi - 80s Danceclub / RaveRaiders, TikTok (1 minute video)

  • faceswap on 80s footage so everyone is Lionel Messi

  • definitely click this link in a private browser window

  • impressive demo

Build and scale creative products with the world's most popular and intuitive video generation models in the Dream Machine API

CogVideoX-5B — image to video / thudm, Hugging Face

 

📻 QUOTES OF THE WEEK

Choose one thing you have total control over and do it exceedingly well.

Actually, screw that.

Just choose a life of stories.

Michael Thompson (source)

 

nothing like the end of the world to put things in perspective

source (neat website)

 

👥 FOR EVERYONE

AI chatbots might be better at swaying conspiracy theorists than humans — Co-author Gordon Pennycook: “The work overturns a lot of how we thought about conspiracies.” / Ars Technica (6 minute read)

Microsoft’s Hypocrisy on AI — Can artificial intelligence really enrich fossil-fuel companies and fight climate change at the same time? The tech giant says yes. / The Atlantic (14 minute read)

Asked about how much water was being allocated to the 20 data centres in the state, Del Prete said he does not “have the data because it is not in my power to request it”, but equated their water consumption to what a restaurant uses in a month.

All the while, some AI proponents have approached the art world more like bullies than collaborators, telling creators they’ll have to use AI tools or become obsolete, dismissing concerns about AI-generated art scams, and even trying to make people give companies their private work as training data. As long as the people behind AI systems seem to revel in knocking artists down a peg, why should anyone who calls themselves an artist want to use them?

  • he’s probably (annoyingly) right

 

📚 FOUNDATIONS

The moment we stopped understanding AI / Welch Labs, YouTube (17 minute video)

  • easy-to-follow video on how transformers work and where our understanding of what they do breaks down

 

🚀 FOR LEADERS

The Button Problem of AI / Napkin Math, Every (14 minute read)

As I see it, the only way that a startup can build a materially superior business is by making the previous tool’s capabilities a commodity component in their machine, then adding adjacent workflows that were previously within the purview of other applications.

Explaining AI Integration for Leaders / Enterprise Agility Magazine (7 minute read)

In this article, I will cover what I have learned through my personal AI journey and how that has shaped the AI Adoption Model, a framework to better understand AI integration in organizations.

How AI is transforming compliance — Here’s how lawyers, accountants, HR professionals, and entrepreneurs alike are cautiously embracing AI-powered tools to ensure that their organizations are following the letter of the law. / Fast Company (9 minute read)

 

🎓 FOR EDUCATORS

An AI tutor helped Harvard students learn more physics in less time — One intriguing experiment suggests the promise of using the new technology / The Hechinger Report (7 minute read)

I challenge the seasoned Times reader to try just one day of this — every 45 minutes, move through a crowded hallway to another hard chair in front of a different personality who will judge and grade you.

  • not really about AI but good

Universities Across the U.S. Are Testing A.I. as a Fundraising Tool — Universities and nonprofits are using an A.I. avatar to boost donor engagement amid staffing shortages. / Observer (4 minute read)

  • on one hand: will likely devolve into even more meaningless solicitations for AI spam filters to manage

  • on the other hand: some un- or under-paid interns’ jobs just got easier so good for them

 

📊 FOR TECHNOLOGISTS

The Data Pipeline is the New Secret Sauce — Why Data Pipelines and Inference Are AI Infrastructure’s Biggest Challenges / Heavybit (14 minute read)

  • related, Data leader’s operating guide to scaling gen AI — Deploying generative AI in the enterprise requires a data-centric road map. Leaders can use a well-defined operating model to successfully scale the technology. / McKinsey Digital (17 minute read)

Google’s DataGemma is the first large-scale Gen AI with RAG - why it matters — Google’s Data Commons serves as the foundation for supplying generative AI with information to draw upon. / Zdnet (5 minute read)

  • official statement, Grounding AI in reality with a little help from Data Commons — Google's DataGemma models bridge the gap between large language models (LLMs) and real-world data by leveraging the Data Commons knowledge graph to improve the factuality and trustworthiness of LLM responses. / Google Research (10 minute read)

It’s great time to be alive but also really scary / Ok_Maize_3709, Reddit (3 minute read)

I was procrastinating in the morning and decided to challenge myself to build an app using Claude and o1 within a day and just hit release to App Store after 6 hours of work.

It took only 6 hours, Carl! And I did not write a single line of code myself.

My approach was the following: o1 was used for planning our architecture and data structure. Claude was better at writing actual code.

  • and further in the discussion

Yes, Claude is marginally better at coding and aesthetics in my experience, so if I would choose just one - I would use Claude to be honest. Also, often when one model get stuck with an error, I use another one to review and in 80% of cases, it solves it easily (sometimes I use Gemini for that as well).

 

🎉 FOR FUN

What if AI could bring the past to life? Cartoonist Amy Kurzweil shares how she trained an AI chatbot on her late grandfather’s archives, helping her connect with a family member she never met — and see her family history in a new light. Backed by her own original drawings, she reveals the profound impact art and AI can have in keeping memories alive.

Kolors Virtual Try-On / Kwai-Kolors, Hugging Face

  • pick a person, pick an outfit, and see that person in that outfit

  • impressive that it fairly accurately renders the 3d transformations

  • also, two weeks until some high schooler gets expelled for misusing this; that guess seems low

  • related, Facecam AI — Real Time Face Swap

/

  • try it on your boss’s emails; fun!

 

🧿 AI-ADJACENT

via jim, Why wordfreq will not be updated / rspeer, GitHub (4 minute read)

The field I know as “natural language processing” is hard to find these days. It’s all being devoured by generative AI. Other techniques still exist but generative AI sucks up all the air in the room and gets all the money.

…

So I don’t want to work on anything that could be confused with generative AI, or that could benefit generative AI.

OpenAI and Google can collect their own damn data. I hope they have to pay a very high price for it, and I hope they’re constantly cursing the mess that they made themselves.

 

⋄