Hi everyone,
As promised in the previous issue, this newsletter is now called Generative AI News! (It was previously Mobile Apps News, publishing since 2014).
In this issue we’ll look at:
The recent updates to Claude, making it much more capable.
Thoughts from the Library Copyright Alliance about why AI training should remain fair use.
How (and Why) the University of Michigan Built Its Own Closed Generative AI Tools.
Our University of Arizona Libraries generative AI resources (guides, tutorials, assignment, FAQ) - these are free to share with a CC4.0 license.
The World of AI: How libraries are integrating and navigating this powerful technology - American Libraries. I was interviewed for this article, along with several others.
And much more…
Enjoy!
Foundation Model News
Anthropic unveils Claude 3, surpassing GPT-4 and Gemini Ultra in benchmark tests - Venture Beat
If you haven’t spent much time with Claude, now is a good time to try it. The free version appears to be about as good as the paid version of ChatGPT (minus some features).
Inflection AI launches new model for Pi chatbot, nearly matches GPT-4 - Venture Beat
Pi is worth trying. It’s marketed mostly as a “friend,” but you can also use it in the ways you use ChatGPT. It searches the web for current information to base answers on, but it doesn’t provide links to sources like Microsoft Copilot or Perplexity.
OpenAI adds ‘Read Aloud’ voiceover to ChatGPT, allowing it to speak its outputs - Venture Beat
The mobile app already could speak answers to you if you started with voice input, but now you can hear answers read aloud on the web version after you’ve started by typing. Choose from five different voices in the Settings. (I like Juniper or Sky).
Captain's log: the irreducible weirdness of prompting AIs - Ethan Mollick
Advice on prompting from professor Mollick of the Wharton School.
Mistral Introduces Two New High Performance LLMs and Neither Are Open-Source - Synthedia
Mistral (located in Paris) started with open source models and is now also offering commercial models. People are calling it the “OpenAI of Europe.”
The killer app of Gemini Pro 1.5 is video - Simon Willison’s Weblog
This is about using video as input and asking questions based on the video (not creating videos). He describes interesting results from uploading a video of books on his bookshelf to Gemini.
Adobe adds AI assistant to Acrobat, Reader in effort to maintain relevance in PDF market - Venture Beat
“The AI Assistant, which is currently in beta testing, lets users ask questions about a PDF’s contents and receive summarized answers.” It seems eventually every PDF may have AI-questioning built-in.
Meta’s new AI model learns by watching videos - Fast Company
Their new model, V-JEPA, learns from watching videos. This could lead to more efficient training that requires much less computing power and costs less. “Meta says it’s releasing the V-JEPA model under a Creative Commons noncommercial license so that researchers can experiment with it and perhaps expand its capabilities.”
Cohere for AI launches open source LLM for 101 languages - Venture Beat
This model, Aya, was built with over 3,000 collaborators around the world, including teams from 119 countries. Ivan Zhang, CTO of Cohere, posted on X that “we’re releasing human demonstrations across 100+ languages to further scale intelligence and ensure that it serves more of humanity than just the english literate world.”
Generative AI for multimedia
Bill Weihl continues to fight for climate policy progress while battling ALS - Eleven Labs
Bill Weihl, founder of ClimateVoice, lost his voice to ALS. He used Eleven Labs to create a clone of his voice based on his past speeches and used it to give a speech as GreenBiz 24.
Alibaba’s new AI system ‘EMO’ creates realistic talking and singing videos from photos - Venture Beat
”Researchers at Alibaba‘s Institute for Intelligent Computing have developed a new artificial intelligence system called “EMO,” short for Emote Portrait Alive, that can animate a single portrait photo and generate videos of the person talking or singing in a remarkably lifelike fashion.”
Adobe’s new prototype generative AI tool is the ‘Photoshop’ of music-making and editing - The Verge
It’s an early experiment, not yet public, but looks interesting. “… giving creatives the same kind of deep control to shape, tweak, and edit their audio. It’s a kind of pixel-level control for music.”
How a Hollywood Director Uses AI to Make Movies - Dan Shipper
”Making movies has traditionally been a risky, expensive business reserved for rich industry insiders. Dave believes AI is democratizing the process, empowering him to visualize his sci-fi short and “test the waters with these techniques.” “[E]veryone has a chance to create something incredible,” he says.”
How Sora Works (and What It Means) - Dan Shipper
You’ve likely heard of Sora, the new video generation technology from Open AI (not yet public). (If you haven’t, watch this 12-min video about it). In this post about Sora, Dan Shipper says, “In the same way that the iPhone and the internet led to the flourishing of vlogs and makeup tutorials, AI filmmaking will create its own genre of film, with its own style and form.”
Copyright
Training Generative AI Models on Copyrighted Works Is Fair Use - Association of Research Libraries, Library Copyright Alliance
“… as champions of fair use, free speech, and freedom of information, libraries have a stake in maintaining the balance of copyright law so that it is not used to block or restrict access to information.”
Licensing research content via agreements that authorize uses of artificial intelligence - Rachael G. Samberg, Timothy Vollmer, and Samantha Teremi, professionals within the Office of Scholarly Communication Services at UC Berkeley Library
”Newly-emerging content license agreements that prohibit usage of AI entirely, or charge exorbitant fees for it as a separately-licensed right, will be devastating for scientific research and the advancement of knowledge. Our aim with this post is to empower scholars and academic librarians with legal information about why those licensing outcomes are unnecessary, and equip them with alternative licensing language to adequately address rightsholders’ concerns. “
AI & The Copyright & Plagiarism Dilemma - Lance Eaton
”… by using “plagiarism” to describe the way AI draws upon sources does not feel like the right term. And I see the use of the term “plagiarism” be set up to be exactly like the way we use the term “piracy” when discussing digital files. It’s an inaccurate metaphor that traps us into and limits our ways of thinking about something new and different.”
Bias
Google suspends Gemini’s ability to generate people after numerous ‘woke’ inaccuracies - Venture Beat
You probably heard about this because it got a lot of press. “…the chatbot generates images of people of color and different ethnicities even in cases of historical context where they are not applicable, such as generating black Catholic popes (there has never been one), giving Vikings dark skin (most didn’t have it) and making Nazi German soldiers look Asian (most were not).”
Reducing biased and harmful outcomes in generative AI - Adobe
Describes how Adobe Firefly image generator aims to reduce bias. OpenAI has similar practices. See their DALL·E 3 system card (PDF).
Language models might be able to self-correct biases—if you ask them - MIT Technology Review
”The team found that just prompting a model to make sure its answers didn’t rely on stereotyping had a dramatically positive effect on its output…”
Deepfakes
Content Credentials Will Fight Deepfakes in the 2024 Elections - IEEE Spectrum
Learn more about the Content Authenticity Initiative.
AI Watermarking Won't Curb Disinformation - Electronic Frontier Foundation
Makes the point that watermarks are easy to remove, sadly.
Misunderstood mechanics: How AI, TikTok, and the liar’s dividend might affect the 2024 elections - Brookings
”Two camps have emerged. One warns of AI leading to a deluge of misinformation, especially on social media, and especially in the many elections coming in 2024. The other finds the fears understandable but largely overblown, pointing to social science research indicating that AI may lead to more misinformation during elections, but with little effect.”
Misinformation reloaded? Fears about the impact of generative AI on misinformation are overblown - Harvard Kennedy School
Three academics make some interesting points that may change how you think about the problem of deepfakes. Worth reading.
Generative AI in education
Generative AI in the humanities classroom - Arizona State University
“Students are currently using AI in their English classes to brainstorm topics; to get help with researching, organizing and editing their papers; and to solicit advice about second-language use,” said Krista Ratcliffe, chair of the Department of English. “In these ways, AI can serve as a powerful invention tool. But students need to develop the critical capacities of AI literacy, so that they may evaluate AI outputs as well as how and when to use them effectively.”
How (and Why) the University of Michigan Built Its Own Closed Generative AI Tools - EDUCAUSE Review
Interesting case study.
Could AI Disrupt Peer Review? - IEEE Spectrum
Publishers’ policies lag technological advances.
Should I cite the AI tool that I used? - Dr. Kristin Terrill, Iowa State University
Interesting thoughts about not how to cite, but when and whether to cite generative AI.
Generative AI in Libraries
The World of AI: How libraries are integrating and navigating this powerful technology - American Libraries
I was interviewed for this article, along with several others.
It Takes a Village: A Distributed Training Model for AI-based Chatbots - Colleen Estes, Beth Twomey, Annie Johnson
Staff from the information technology and reference departments at the University of Delaware Library, Museums and Press came together in a partnership to pilot a low-cost AI-powered chatbot.
The race to define what’s next for AI in libraries - Northwestern University Libraries
The story of building a prototype AI tool that added natural language search to part of their digital collections.
School Libraries: Ready to Lead AI Adoption - School Library Systems Association of New York
This project is working on curriculum — “we will develop a 2-3 lesson unit on AI for every grade level Pre-K through 12” and is also developing professional learning for school librarians.
Just for fun
GOODY-2: The world’s most responsible AI model
Try it. No matter what you ask it, it refuses and makes up a reason why your query would be unethical. Then read Meet the Pranksters Behind Goody-2, the World’s ‘Most Responsible’ AI Chatbot.
Laurie Anderson on making an AI chatbot of Lou Reed - The Guardian
An experimental “AI text generator that emulates the vocabulary and style of her own longtime partner and collaborator, Velvet Underground co-founder Lou Reed, who died in 2013.”
My offerings
Student guide to ChatGPT - University of Arizona Libraries
I worked with several colleagues to create this guide, along with a similar guide for instructors: AI Literacy in the age of ChatGPT. Both guides are free to copy and share, with a CC 4.0 license.
Recently, I added a chart to both guides: Which AI tool for your task? Take a look and think about the difference between tools that are “wordsmiths” vs. tools for search.
Last summer I worked with a colleague to create four Tutorials about generative AI. I recently updated them and will soon add additional tutorials about multimedia generation.
Find all of our University of Arizona Libraries generative AI resources here: https://libguides.library.arizona.edu/genAI/guides
The future
The Trust Flip - Kevin Kelly
”Now when we see a photograph we assume it is fake, unless proven otherwise. When we see video, we assume it has been altered, generated, special effected, unless claimed otherwise. The new default for all images, including photographic ones, is that they are fiction – unless they expressly claim to be real.”
Unlocking the incredible with AI - Exponential View
”AI will be able to see the hidden connections within existing knowledge. Historically, many innovations have demonstrated this potential, such as the fusion of insights from biology, observing the adhesion mechanism of burdock burrs, with materials engineering and textile technology to create the brilliant invention of Velcro (bye-bye shoelaces).”
Learn more
If you want to learn more about generative AI, contact me about doing a webinar for your group.
Later this year I’ll offer a 4-6 week asynchronous online course about generative AI. Sign up here if you’d like to be notified when it’s ready.
And as always, you can follow me on X, Mastodon, or Bluesky, where I post daily about generative AI.