- That AI Thing
- Posts
- weekend ai reads for 2024-05-24
weekend ai reads for 2024-05-24
đ° ABOVE THE FOLD: SCARLETT JOHANSSON & OPENAI
only because so much happened, in case you missed it or purposely tuned it out, the Scarlett Johansson story (so far):
May 13:
OpenAI released âSkyâ last week, and Sam Altman even not-so-cryptically tweeted just the word âherâ, probably referencing the 2013 Spike Jonze film, âHerâ, in which the lead character develops a relationship with an AI companion, voiced by Scarlett Johansson (may be spoilers at the second link; also, the movie is over 10 years old so does it really still warrant a spoiler-warning?)
May 20:
Scarlett Johansson âAngeredâ After ChatGPT Used âEerily Similarâ Voice â OpenAI claimed it used the voice of a different actress to voice âSkyâ / Rolling Stone (4 minute read)
Scarlett Johanssonâs official statement; tl;dr is she says they asked to use her voice, she declined, and she believes they did so anyway / Bobby Allyn (NPR), Threads
OpenAI released a statement in response, quoted in the same Threads link above:
May 21:
OpenAI did suspend âSkyâ while they sort this out: OpenAI suspends ChatGPT voice âthat sounds like Scarlett Johanssonâ / Sky News (2 minute read)
May 22:
as OpenAI are wont to do, they took this to the court of public opinion, sharing recordings of a actress hired to provide the voice for âSkyâ with the Washington Post: OpenAI didnât copy Scarlett Johanssonâs voice for ChatGPT, records show / Washington Post (6 minute read)
and next:
OpenAI are still going to get sued, and theyâre going to lose, based on Midler v. Ford Motor Co. and Waits v Frito-Lay, Inc.; these cases ruled and upheld, respectively, that if a characteristic of a famous personâtheir voice in these two casesâis a distinctive characteristic of that person, then it is unlawful to imitate their voice without express written consent (disclaimer: not a lawyer)
OpenAI came off looking poorly, again: The Scarlett Johansson Incident Makes OpenAI Look Desperate / New York Magazine (6 minute read)
đť QUOTE OF THE WEEK
You want to go grab the tub of ice cream and give up. Even though it felt like it was my fault, it really wasnât. The system was working against job seekers.
Victor Schwartz (source: âYouâre Fighting AI With AIâ: Bots Are Breaking the Hiring Process / Wall Street Journal)
đď¸ FOUNDATIONS & CULTURE
Introducing Copilot+ PCs / Microsoft Blog (18 minute read)
Copilot+ PCs are the fastest, most intelligent Windows PCs ever built. With powerful new silicon capable of an incredible 40+ TOPS (trillion operations per second), allâday battery life and access to the most advanced AI models, Copilot+ PCs will enable you to do things you canât on any other PC.
related (1), Microsoftâs AI chatbot will ârecallâ everything you do on its new PCs / The Guardian (3 minute read)
related (2), Microsoft Team Copilot: A virtual team member to run meetings and projects / Venture Beat (3 minute read)
maybe Ray Dalio really was ahead of his time (e.g., Ray Dalio: Itâs âFantasticâ When We Play Conversations Weâve Recorded Back To Our Employees, 2014)
Faking William Morris, Generative Forgery, and the Erosion of Art History â Buying fake William Morris prints on Etsy and other early signs of epistemological collapse / Maggie Appleton (6 minute read)
To confound matters, the Etsy stores selling these generated images also sell genuine prints by Morris, Monet, Klimt, and Matisse. These sit alongside their modern expansion packs, blending right in. But which are which? You get coerced into playing an art historian forgery spotter.
Nvidia says 20K AI startups are building on its platform / Venture Beat (9 minute read)
âThe next company who reaches the next major plateau gets to announce a groundbreaking AI, and the second one after that gets to announce something thatâs 0.3% better,â Huang explained. âTime to train matters a great deal. The difference between time to train that is three months earlier is everything.â
related, Ten Gifts Nvidia Gave Its Investors / The Information ($) (6 minute read)
via renee, Women Leaders in Tech Are Paving the Way in GenAI / Boston Consulting Group (12 minute read)
đ EDUCATION
Does Emory University really not get how technology works? / AI Log, Substack (sorry) (8 minute read)
Emory University faces a lawsuit involving a student who built an AI study tool with a $10,000 prize and then suspended him for cheating for using a Canvas API (?)
this post is one perspective but Emory does not look great here
AI Robot Gives Graduation Speech at Buffaloâs DâYouville University / New York Times (6 minute read)
Andrew Fields, a DâYouville University student who started the petition, wrote in the petition that many students âfeel disrespectedâ by the universityâs decision to have a robot address them, especially those who could not attend their high school graduations in 2020 because of the coronavirus pandemic.
Microsoft and Khan Academy offer a free AI assistant to all US teachers â With Khanmigo for Teachers, K-12 educators can spend less time prepping lessons and more time with students. / Zdnet (3 minute read)
related, 5 great things to read or watch this summer / Bill Gates, Gates Notes (5 minute read)
referring to Sal Khanâs new book, Brave New Words: How AI Will Revolutionize Education (and Why That's a Good Thing)
Opinion: What Is Higher Edâs Role in Providing AI Job Skills? / Government Technology (7 minute read)
In addition to programming and technical skills, the next generation of AI developers may also need training in subjects traditionally aligned with liberal-arts education, such as ethics, problem-solving and communication.
related, Computer-Science Majors Graduate Into a World of Fewer Opportunities / Wall Street Journal ($) (4 minute read)
đ DATA & TECHNOLOGY
MOMENT: A Family of Open Time-series Foundation Models / arXiv (57 minute read)
we are bullish on time-series foundation models
Linearly probing MOMENT achieves near state-of-the-art performance on most datasets and horizons, and is only second to PatchTST which generally achieves the lowest MSE. On many datasets and horizons, forecasting models based on LLMsâ TimeLLM and GPT4TS perform worse than MOMENT. Notably, N-BEATS outperforms several recent methods, emphasizing the importance of comparing forecasting performance beyond transformer-based approaches.
AIâs Black Boxes Just Got a Little Less Mysterious / New York Times (6 minute read)
Researchers at the A.I. company Anthropic claim to have found clues about the inner workings of large language models, possibly helping to prevent their misuse and to curb their potential threats.
For example, they discovered that if they forced a feature linked to the concept of sycophancy to activate more strongly, Claude would respond with flowery, over-the-top praise for the user, including in situations where flattery was inappropriate.
related (1), from Anthropic: Mapping the Mind of a Large Language Model / Anthropic Blog (11 minute read)
related (2), the paper: Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet (132 minute read)
Introducing the Frontier Safety Framework / Google DeepMind Blog (6 minute read)
related (1), the report: Frontier Safety Framework [PDF]
related (2), Responsible Generative AI Toolkit / Google for Developers Blog
GPT-4oâs Chinese token-training data is polluted by spam and porn websites â The problem, which is likely due to inadequate data cleaning, could lead to hallucinations, poor performance, and misuse. / MIT Technology Review (11 minute read)
(Perhaps) Beyond Human Translation: Harnessing Multi-Agent Collaboration for Translating Ultra-Long Literary Texts / arXiv (111 minute read)
the researchers used a series of agents to replicate what a real translation company would do
and the agents produced outputs that were preferred over human-written translations
we are also bullish on agents
đ FUN and/or PRACTICAL THINGS
Google's AI Feature Suggested Using Glue to Keep Cheese on a Pizza / Business Insider (4 minute read)
A screenshot of the summary it generated, shared on X, shows it responded with âcheese can slide off pizza for a number of reasons,â and that the user could try adding âabout â cup of non-toxic glue to the sauce to give it more tackiness.â
avoid the rush: Ten Blue Links â How to Turn Off AI Overview in Google and Set âWebâ as Default
or just add âudm=14â to the search URL
finds items similar to some highfalutin one, at a lower cost, using vision AI
Arrange â Generate a plan for anything and add it to your calendar in seconds.
looks promising; waitlist
Infinite Wonderland â This is an AI experiment where the timeless classic Aliceâs Adventures in Wonderland is endlessly reimagined by artists, AI and you.
Wikipedia Citation Needed / Chrome Web Store
The Future Audiences team at the Wikimedia Foundation built an experimental new AI tool to check what Wikipedia has to say about what youâre reading, no matter where you are on the web, and weâd love your feedback.
đ§ż AI-ADJACENT
Black Games Archive â Context, Culture, and Representation
Black Games Archive is a multimedia, public-facing database of games, digital resources, accessible scholarship, and designer interviews that are relevant to the intersections between Black culture, games, and play.
â