Two years after the release of ChatGPT, it may not be surprising that creative work is used without permission to power AI products. Yet the notion remains disturbing to many artists and professionals who feel that their craft and livelihoods are threatened by programs. Transparency is generally low: Tech companies tend not to advertise whose…
[A] popular type of generative AI model can provide turn-by-turn driving directions in New York City with near-perfect accuracy — without having formed an accurate internal map of the city. Despite the model’s uncanny ability to navigate effectively, when the researchers closed some streets and added detours, its performance plummeted. When they dug deeper, the researchers…
Everyone who is reading this article in 2024 grew up in an era where a photograph was, by default, a representation of the truth. A staged scene with movie effects, a digital photo manipulation, or more recently, a deepfake — these were potential deceptions to take into account, but they were outliers in the realm…
Chatbots are designed to give conversational answers and keep people engaged. Names and links to sources for answers range from nonexistent to hidden, though some companies are starting to make them more visible. Even when AI does include a source, it adds it after the fact, said Jevin West, a professor and co-founder of the…
I worked on LLMs, and now I got opinions. Today, let’s talk about when LLMs make mistakes. On AI Slop You’ve already heard of LLM mistakes, because you’ve seen them in the news. […] LLMs could be used to summarize sources. Something that’s fairly obvious in my journal club is that many researchers are…
Over the past couple of years, researchers at Jigsaw, a Google subsidiary that focuses on online politics and polarization, have been studying how Gen Zers digest and metabolize what they see online. The researchers were hoping that their work would provide one of the first in-depth, ethnographic studies of Gen Z’s “information literacy.” But the…
Already facing scandal, the Washington Post‘s new-ish CEO and publisher, Will Lewis, has announced that the newspaper will be pivoting to artificial intelligence to turn around its dismal financial situation. […] The paper’s chief technology officer, meanwhile, announced to staffers that going forward, WaPo is to have “AI everywhere in our newsroom,” according to Tani. It’s unclear,…
I guess now we know why for the past decade or so Google doesn’t seem to be valuing what used to be magical about Google — the ability to rank pages so effectively and connect users to web pages that met their needs. That’s just not what’s driving Google these days. With AI search results…
Here’s the underlying problem. We have an operating image of thought, an understanding of what thought is and how to make thought productive of knowledge and truth through the use of various methods. This image and its methods are, of course, products of human history. They change over time, in different cultures, and so on.…
After learning of his AIDS diagnosis, artist Keith Haring created the work, “Unfinished Painting” (1989), which is mostly a blank canvas, with streaks that evoke his own interrupted, incomplete life. Someone who thought it was “so sad” that the painting was unfinished used AI to “complete what he couldn’t finish.” Needless to say, the bot…
Some 2300 years ago in ancient Greece, Plato wrote a dialogue featuring his mentor Socrates, who argued that the ability to churn out the longest written compositions on trivial topics or the shortest compositions on important topics is a shallow skill that has nothing to do with human understanding, much like demonstrating that you can…
This story is about the admissions process, not about academics. Duke University is no longer assigning numerical ratings to applicants’ standardized test scores and essays, a university spokesperson confirmed in an email to Inside Higher Ed. The change was made earlier this year and has been in effect for the latest round of applications. […] Dean…
No names in this single-source anecdote out of Hong Kong, credited to “Senior Superintendent Baron Chan Shun-ching.” The employee joined a video call with who he thought was the business’s chief financial officer. He was initially suspicious after a message from the CFO mentioned a ‘secret transaction’, suggesting it was a phishing scam…. However, after other…
Researchers had programmed the various large language models (LLMs) to act in what they termed malicious ways, and the point of the study was to see if this behaviour could be removed through the safety techniques. The paper, charmingly titled Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training, suggests “adversarial training can teach models…
Far from being “stochastic parrots,” the biggest large language models seem to learn enough skills to understand the words they’re processing. […] A trained and tested LLM, when presented with a new text prompt, will generate the most likely next word, append it to the prompt, generate another next word, and continue in this manner,…
I’m sad thinking of all the students whose academic careers and personal intellectual growth will suffer because they depend on generative text software — whether or not they get “caught” for plagiarism. In my mind, the fundamental error that almost everyone is making is in believing that Generative AI is tantamount to AGI (general purpose…
In a writing-intensive class, students need to write extensively to the point that the teacher cannot possibly grade all of it. — Edutopia via NCTE Good advice from an article by Kara Douma, reprinted by NCTE. I need to hear that. The referees don’t score every practice. The coaches don’t give you feedback after…
The results showed that out of the 512 questions, 259 (52%) of ChatGPT’s answers were incorrect and only 248 (48%) were correct. Moreover, a whopping 77% of the answers were verbose. […] According to the study, the well-articulated responses ChatGPT outputs caused the users to overlook incorrect information in the answers. “Users overlook incorrect information…
OpenAI has already taken a toll on human workers. It’s chilling to contemplate that students and creatives are comfortable with relying on this product in any meaningful way. The business model depends on devaluing human labor, and benefitting the powerful. The 51 moderators in Nairobi working on Sama’s OpenAI account were tasked with reviewing texts,…
We ask students to cite encyclopedia articles, dictionaries, news interviews with experts, and in-person conversations. A generative text-generation tool that scrapes published work and uses a statistical model to emulate a response that mimics the way the humans in its database have responded to similar prompts is formally something like a conversation, and the MLA’s…