weekend ai reads for 2024-04-05

📰 ABOVE THE FOLD: BEYONCÉ & MUSIC & TAYLOR SWIFT

“The joy of creating music is that there are no rules,” said Beyoncé. “The more I see the world evolving the more I felt a deeper connection to purity. With artificial intelligence and digital filters and programming, I wanted to go back to real instruments.”

The Rise of AI Taylor Swift — Fans are using AI tools to synthesize the star’s voice, demonstrating how new technology is blurring the boundaries between reality and fiction. / The Atlantic

Billie Eilish and Katy Perry among 200 artists protesting AI's “assault on human creativity” — More than 200 famous musicians, including Billie Eilish, Katy Perry, and Smokey Robinson, have issued an open letter calling on AI companies to stop using AI in ways that violate and devalue the rights of human artists. / Decoder

“The best music will be the most original use of AI. It won’t be a fake Drake singing a Nirvana song. It will be a 14-year-old kid [making] something that’s totally unique,” says Stringer. But almost in the same breath, he warns tech companies off the use of AI to create music based on existing artists and their music, highlighting the delicate dance that the music industry is facing in dealing with new technology.

AI’s quiet creep into music punctuated by ‘SpongeBob’ voices and a secretive artist called Glorb — Experts who focus on artificial intelligence and music said questions surrounding copyright and ownership still linger as a new era of technology dawns in the music industry. / NBC News

 

📻 QUOTE OF THE WEEK

The task at hand for the entire quantum ecosystem is to increase the fidelity of qubits and enable fault-tolerant quantum computing so that we can use a quantum machine to unlock solutions to previously intractable problems. In short, we need to transition to reliable logical qubits — created by combining multiple physical qubits together into logical ones to protect against noise and sustain a long (i.e., resilient) computation.

Jason Zander, EVP of Microsoft's Strategic Missions and Technologies division (source)

 

🏗️ FOUNDATIONS & CULTURE

So... I wanted to know what the big themes, trends and takeaways were from #sxsw2024 over the last few days. And as I didn't go to Austin, I had to make a 3 little ai-agents to (1) listen to about 150 of the top talks, (2) analyze the talks (phew) and (3) build a (rather massive) deck.

  • as recovering conference-goers, this is a great use of AI

  • 900+ slides summarizing seemingly every talk from SXSW

  • the report [PDF, 90MB]

The U.S. House has set a strict ban on congressional staffers' use of Microsoft Copilot, the company's AI-based chatbot, Axios has learned.

  • … because the tools don’t meet federal government security and compliance requirements; just FYI to everyone who says their OpenAI wrapper is fill-in-the-blank compliant.

via george, AI Ascent 2024 (playlist) / Sequoia Capital, YouTube

  • eleven short-to-medium (~6m to ~37m) clips from Sequoia Capital's AI Ascent

  • Andrej Karpathy is always worth the time (36:58)

Most decisions in life are reversible. But some are one-way doors. Jeff Bezos talks about these: “Some decisions are consequential and irreversible or nearly irreversible—one-way doors—and these decisions must be made methodically, carefully, slowly, with great deliberation and consultation.”

Our results indicate that working under algorithmic rather than human management has substantial consequences on the way workers approach and carry out the job. Specifically, workers who receive positive feedback from human managers put significantly more care and effort into their tasks and perform more accurately, compared to workers who receive identical feedback from algorithmic management.

What Are AI Agents—And Who Profits From Them? — The newest wave of AI research is changing everything / Every

Jon Stewart On The False Promises of AI (14:54) / The Daily Show, YouTube

The 2024 MAD landscape features 2,011 logos in total.That number is up from 1,416 last year, with 578 new entrants to the map.

Here's How Generative AI Depicts Queer People — Wired investigates how artificial intelligence tools, like OpenAI’s Sora, currently portray members of the LGBTQ community. Hint: It’s a lot of purple hair. / Wired

The only voice of the customer that really matters is what actual customers — with eyeballs and index fingers and wallets — have said. Verbatim.

Execs often recoil when they hear this, claiming “we need to stay broad to maximize our market” — but that is a 101-level mistake. Focusing on one person’s understanding of their problem isn’t constraining your solution’s appeal. In fact, it is the opposite.

 

🎓 EDUCATION

Crafting ethical AI landscapes in K-12 education — The journey toward integrating AI into K-12 education ethically is a collaborative endeavor requiring engagement at all levels / E-School News

using AI to discover educational resources is an emerging trend

  • related (1), The Curricula — search for something to learn

  • related (2), EduHunt — Instantly find quality educational content on YouTube

    • free trial requires sign-up

  • archive link because WSJ’s website continues to be hostile towards subscribers

MathGPT Pro — Starting with an AI Math Tutor, we are envisioning a future where quality, affordable, and personalized learning experiences are accessible to all people at any time.

Universities can also help their faculty manage the changed competitive landscape by encouraging and allowing them to be more risk-taking. The comparative advantage of academic researchers in AI is to do more high-risk exploration, and incentive structures at universities must change to account for this.

  • option #1 is “give up!”; slightly more positive from there

 

📊 DATA & TECHNOLOGY

The Data Provenance Initiative is a large-scale audit of AI datasets used to train large language models. As a first step, we've traced 1800+ popular, text-to-text finetuning datasets from origin to creation, cataloging their data sources, licenses, creators, and other metadata, for researchers to explore using this tool. The purpose of this work is to improve transparency, documentation, and informed use of datasets in AI.

Onscreen Entities: These are entities that are currently displayed on a user’s screen

Conversational Entities: These are entities relevant to the conversation. These entities might come from a previous turn for the user (for example, when the user says “Call Mom”, the contact for Mom would be the relevant entity in question) or from the virtual assistant (for example, when the agent provides a user with a list of places or alarms to choose from).

Background Entities: These are relevant entities that come from background processes that might not necessarily be a direct part of what the user sees on their screen or their interaction with the virtual agent; for example, an alarm that starts ringing or music that is playing in the background

  • ignore the clickbait headline

  • it’s hard to tell how much of Apple’s recent flurry of AI releases are meant for wall street to give them the Microsoft/Nvidia/Meta boost, and how much is signaling changes in their products; the cynic in us says the former for at least another product cycle

  • the paper, ReALM: Reference Resolution As Language Modeling / arXiv

There are two templates to generate the synthetic data. The first “base” template includes mentions, entities, and possible slot values if necessary. The second “language” template template imports the base template and adds different variations of queries that can be used for targeted cases with references defined in the base template.

Your AI Product Needs Evals — How to construct domain-specific LLM evaluation systems. / Hamel Husain

Microsoft insists that Copilot runs on the NPU instead of the GPU to minimize the impact on battery life.

Here’s Proof You Can Train an AI Model Without Slurping Copyrighted Content — OpenAI claimed it’s “impossible” to build good AI models without using copyrighted data. An “ethically created” large language model and a giant AI dataset of public domain text suggest otherwise. / Wired

 

🎉 FUN and/or PRACTICAL THINGS

  • fun seeing these filters on a large screen

The new tech bringing loved ones back to life through AI — The virtual reality tool called "live forever mode" features digital avatars who can simulate a person's voice, mannerisms and movements after just 30 minutes of the user being observed. / Sky News

Dr. Watermark — Remove Unwanted Watermarks

  • enjoy before they get sued into oblivion

Terra — companion for mindful wandering; designed with the science of AI and the wisdom of mindfulness, Terra is the incredible, pocket-sized compass that lets you wander without getting lost.

Every journey begins with a prompt. Terra’s sophisticated AI translates your intentions, available time, and precise location into a tailored trail of GPS coordinates. Users don't know where Terra will take them; the only certainty is finding their way back to the starting point.

GPT Author / mshumer, GitHub

This project utilizes a chain of GPT-4, Stable Diffusion, and Anthropic API calls to generate an original fantasy novel. Users can provide an initial prompt and enter how many chapters they'd like it to be, and the AI then generates an entire novel, outputting an EPUB file compatible with e-book readers.

A 15-chapter novel can cost as little as $4 to produce, and is written in just a few minutes.

  • not sure this is a good idea, because it’s creating more middling-quality “content”, but no one reading this is going to do anything with this, so we don’t feel too bad

  • didn’t listen to this all the way through, either

 

🧿 AI-ADJACENT

  • there have been surreal and mediocre Sora music videos starting to circulate; the Vampire Weekend video collage contrasts with what AI does

 

🛎️ ICYMI

The most-clicked link from last week was Guiding Principles for the Church’s Use of Artificial Intelligence / The Church of Jesus Christ of Latter-day Saints