Language Log

It's been clear for a while that "large language models" can be prompted to fulfill writing assignments, and that LLM detection doesn't work, and that "watermarking" won't come to the rescue. There's lots of on-going published discussion, and even more discussion in real life.

As documented by the MLA-CCCC Joint Task Force on Writing and AI, the conclusion seems to be a combination of bringing AI explicitly into the class, and designing some assignments where students need to function without it.

In one recent example, Joe Moxley has posted the syllabus for his course "Writing with Artificial Intelligence – Syllabus (ENC 3370)".

Eric Doty dueling with Grammarly on LinkedIn:

me: in real time
grammarly: in real-time

me: k, in real-time
grammarly: in real time

me: i'm going to smash you in real time
grammarly: i'm going to smash you in real-time

The screenshot I show everyone who tells me they're using AI for anything

[image or embed]

— Chris PG | PapaGlitch (@papaglitch.bsky.social) Aug 26, 2024 at 5:20 AM

Many Language Log readers are probably aware of the Monkey King, Sun Wukong, who is one of the leading characters in the famous Ming Dynasty novel, Journey to the West.  I wrote about him in "'Baton' and 'needle' in space" (6/17/21):

In the 16th-century novel, Journey to the West, the simian hero, Sun Wukong ("Monkey Enlightened to Emptiness") possesses a magical staff, the jīngū bàng 金箍棒 ("golden cudgel / rod / baton") that has transformational properties.  One of its forms is that of the dìnghǎi shénzhēn 定海神针 ("numinous needle that stabilizes the sea"), which was actually the original source of the jīngū bàng 金箍棒 ("golden cudgel / rod / baton").  Thus we can see that both of the objects that Martin asked about are attributes of the supernatural simian, Sun Wukong, of Journey to the West.  (Of course, the meaning of "baton" for relay racing is also operative.)

In the context of this post, It is pertinent to note that Sun Wukong is capable of flying 108,000 li / tricents (54,000 km, 34,000 mi) in one somersault.  For this and all manner of esoteric lore about the magical monkey and the novel in which he appears, see the remarkable website of Jim McClanahan, Journey to the West Research.

This continues the tradition of using terms from Chinese legend and myth for names of objects, equipment, places, etc. in space related research and technology.

Deepa Seetharaman & Matt Barnum, "There’s a Tool to Catch Students Cheating With ChatGPT. OpenAI Hasn’t Released It.", WSJ 8/4/2024:

OpenAI has a method to reliably detect when someone uses ChatGPT to write an essay or research paper. The company hasn’t released it despite widespread concerns about students using artificial intelligence to cheat.

The project has been mired in internal debate at OpenAI for roughly two years and has been ready to be released for about a year, according to people familiar with the matter and internal documents viewed by The Wall Street Journal. “It’s just a matter of pressing a button,” one of the people said.

Deep Learning "AI" systems are doing ever more impressive things, although there continues to be plenty of skepticism Out There about the underlying technologies.

And now there's increasing skepticism about the economic impact. An 8/2/2024 Atlantic article has the title "The Generative-AI Revolution May Be a Bubble", and the subtitle "Tech firms have been spending historic amounts of money on AI—but will it pay off?" A recent Goldman-Sachs report had the title "GEN AI: TOO MUCH SPEND, TOO LITTLE BENEFIT?", featured in a 7/24 Washington Post article "Big Tech says AI is booming. Wall Street is starting to see a bubble."

One idea about AI socio-economics has always been that wider and wider swaths of the world's population will want to pay for its participation in their daily lives, replicating the successes of social media and portable networked devices. But even if this is true, there are some hiccups along the way, such as Google's "Dear Sidney" Olympics ad:

Locally consistent hallucinations, anyhow… Zoë Hannah, "We pushed this ChatGPT game to the limits, but playing it the right way is more fun", Polygon 7/30/2024:

Apparently, we all like playing god, and we all like doing it badly. I bet none of us thought that removing the ladder from our Sims’ pools was such a universal experience until it became a pretty popular meme, and it’s no secret that lots of mods are centered on adding, uh, explicit elements to games. So, naturally, when I started playing around with DeepGame, Utile Labs’ ChatGPT-based choose-your-own-text-adventure game, I put my best sicko foot forward.

The game, which runs on ChatGPT and is available to anyone with an account, generates stories in a variety of genres. You start off with a command like “Play a romantasy story” or “Surprise me” and let the GPT do its thing — and despite my desire to break the game, I found it much more enjoyable when I took it just a little more seriously.

Boone Ashworth, "Wear This AI Friend Around Your Neck", Wired 7/30/2024:

The latest attempt at an AI-powered wearable is an always-listening pendant. But it doesn’t help you be more productive, it just keeps you company.

AVI SCHIFFMANN SHOWS up to the WIRED office with a Friend hanging around his neck. It dangles there like a pendant on a necklace. It’s about the size and shape of an AirTag—a soft, round little puck that rests right next to Schiffmann’s heart, just atop the Dark Side of the Moon logo on the shirt behind it.

The Friend, to be clear, is an AI wearable. It’s a pal, a buddy, but mostly an AI chatbot that lives inside the pendant. It always has an opinion to share about what’s going on around it, which it communicates using text messages and push notifications on the phone it’s paired to.

Schiffmann and his Friend (this one’s name is Emily) have come to WIRED’s San Francisco office to meet with me and my colleague Reece Rogers to talk publicly about this new AI wearable for the first time. Before we get started, I tell Schiffmann I’d like to record our chat and ask if he’s cool with that. This is considered a good journalistic practice, sure, but also it’s a legal requirement in California, which requires two-party consent before taping a private interaction. So I ask permission to turn on a tape recorder and Schiffmann just laughs.

“I am the last person who would mind that,” he says.

That makes sense. After all, the pendant around his neck has already been listening to us this entire time.

"Augmenting parametric data synthesis with 3D simulation for OCR on Old Turkic runiform inscriptions: A case study of the Kül Tegin inscription", Mehmet Oğuz Derin and Erdem Uçar, Journal of Old Turkic Studies (7/21/24)

Abstract

Optical character recognition for historical scripts like Old Turkic runiform script poses significant challenges due to the need for abundant annotated data and varying writing styles, materials, and degradations. The paper proposes a novel data synthesis pipeline that augments parametric generation with 3D rendering to build realistic and diverse training data for Old Turkic runiform script grapheme classification. Our approach synthesizes distance field variations of graphemes, applies parametric randomization, and renders them in simulated 3D scenes with varying textures, lighting, and environments. We train a Vision Transformer model on the synthesized data and evaluate its performance on the Kül Tegin inscription photographs. Experimental results demonstrate the effectiveness of our approach, with the model achieving high accuracy without seeing any real-world data during training. We finally discuss avenues for future research. Our work provides a promising direction to overcome data scarcity in Old Turkic runiform script.

"China deploys censors to create socialist AI:  Large language models are being tested by officials to ensure their systems ‘embody core socialist values’", by Ryan McMorrow and Tina Hu in Beijing, Financial Times (July 17 2024)

Chinese government officials are testing artificial intelligence companies’ large language models to ensure their systems “embody core socialist values”, in the latest expansion of the country’s censorship regime.

The Cyberspace Administration of China (CAC), a powerful internet overseer, has forced large tech companies and AI start-ups including ByteDance, Alibaba, Moonshot and 01.AI to take part in a mandatory government review of their AI models, according to multiple people involved in the process.

The effort involves batch-testing an LLM’s responses to a litany of questions, according to those with knowledge of the process, with many of them related to China’s political sensitivities and its President Xi Jinping.

"China Puts Power of State Behind AI—and Risks Strangling It:  Government support helps China’s generative AI companies gain ground on U.S. competitors, but political controls threaten to weigh them down", by Lia Lin, WSJ (7/16/24)

Screenshot of emails circulating on social media:

Wasn’t on my 2024 Bingo Card:

US Lawmakers: We need to ban $NVDA GPUs sales into China, or else they will lead in AI and boost their military.

Chinese social meme accounts burns through valuable Huawei Ascend compute, to make Biden and Trump sing Chinese folk songs about… pic.twitter.com/T03DwIZKp4

— Marcel Münch (@_mm85) July 9, 2024