large language models (Page 2)

Latest in large language models

A team of scientists subjected AIs to a number of games, forcing them to choose between pain and gain or pleasure and loss.

Artificial Intelligence

Scientists Experiment With Subjecting AI to Pain

By Victor Tangermann

Apple engineers flagged major flaws with the "reasoning" abilities of top AI models: that it's mostly a front.

Artificial Intelligence

Before Apple’s AI Went Haywire and Started Making Up Fake News, Its Engineers Warned of Deep Flaws With the Tech

By Frank Landymore

Apple has temporarily halted its disastrous "Apple Intelligence" feature which consistently bungled up summaries of breaking news alerts.

Artificial Intelligence

Apple Halts Disastrous AI System That Was Making Up Fake News Stories and Pushing Them to iPhone Users

By Victor Tangermann

Researchers found that if a mere 0.001 percent of the training data of a given LLM is "poisoned," the entire thing falls apart.

Artificial Intelligence

If Even 0.001 Percent of an AI’s Training Data Is Misinformation, the Whole Thing Becomes Compromised, Scientists Find

By Victor Tangermann

OpenAI's recently unveiled o3 AI model can cost well over $1,000 per task when used in its most powerful mode.

OpenAI’s Latest AI Can Cost More Than $1,000 Per Query

By Frank Landymore

Scientists have found that readers tend to prefer AI-generated poems over human-written ones, including the works of Shakespeare.

Artificial Intelligence

Judging Them Blind, Humans Appear to Prefer AI-Generated Poems

By Victor Tangermann

The AI-generated summaries appear as a "Key Points" box above the main body of articles on The Wall Street Journal.

Artificial Intelligence

The Wall Street Journal Is Testing AI-Generated Summaries of Its Articles

By Frank Landymore

AI researcher Gary Marcus warns that recent signs that AI improvements are stagnating could signal a crash in the industry in the future.

Artificial Intelligence

AI Expert Warns Crash Is Imminent As AI Improvements Hit Brick Wall

By Frank Landymore

OpenAI has released a new benchmark dubbed "SimpleQA" to measure the accuracy of its AI models. The results are damning.

OpenAI Research Finds That Even Its Best Models Give Wrong Answers a Wild Proportion of the Time

By Victor Tangermann

ByteDance, the Chinese company that owns TikTok, fired an intern who apparently screwed with the company's artificial intelligence projects.

Artificial Intelligence

Intern Fired by TikTok Owner for Sabotaging Its AI Model

By Noor Al-Sibai

Frontier AI models' mathematical reasoning skills and the benchmarks used to measure them may be deeply flawed, a new study reveals.

Artificial Intelligence

Top “Reasoning” AI Models Can be Brought to Their Knees With an Extremely Simple Trick

By Maggie Harrison Dupré

Online fantasy and science fiction magazine Clarkesworld is drowning in an onslaught of AI-generated slop.

Artificial Intelligence

Editors of Sci-Fi Magazine Disgusted as They Realized Submissions Were Filling With AI Slop

By Victor Tangermann

An assistant professor of philosophy was shocked to find her students use ChatGPT to "briefly introduce" themselves.

Professor Horrified When Students Use ChatGPT to Generate “Briefly Introduce Yourself” Assignment… in Ethics Course

By Victor Tangermann

OpenAI's newfangled o1-preview AI model is still struggling with the absolute basics — like counting letters.

OpenAI’s New “Strawberry” AI Is Still Making Idiotic Mistakes

By Victor Tangermann

AI CEO Proud of Chatbot for Convincing Woman to Euthanize Her Dog

Artificial Intelligence

AI CEO Proud of Chatbot for Convincing Woman to Euthanize Her Dog

By Victor Tangermann

By analyzing a decade of scientific papers, researchers found AI models are overusing "style" words that were uncommon just a few years ago.

Artificial Intelligence

Research Shows That AI-Generated Slop Overuses Specific Words

By Frank Landymore

A new paper finds that advanced AI language models are terrible at a specificproblem, calling current industry benchmarks into question.

Artificial Intelligence

This Simple Logic Question Stumps Even the Most Advanced AI

By Maggie Harrison Dupré

A data training failure has resulted in OpenAI's new GPT-4o model spitting out spam and porn-littered Chinese-language responses.

The New ChatGPT Has a Huge Problem in Chinese

By Maggie Harrison Dupré

The service Character.AI is used by millions of young people to interact with AI versions of their favorite characters.

Artificial Intelligence

Lonely Teens Are Making “Friends” With AIs

By Frank Landymore

A large language model called "gpt2-chatbot" caused a stir among the AI community for its impressive capabilities that seem to rival GPT-4.

Artificial Intelligence

Mystery Deepens as Super Powerful AI Appears, Then Disappears

By Frank Landymore

A team of resaerchers at a Berkeley-based startup called Profulent claim to have used generative AI technologies to edit human DNA.

Startup Uses AI to Edit Human DNA

By Victor Tangermann

According to the Institute for Public Policy Research, up to 8 million jobs in the UK are vulnerable to being replaced by AI.

Artificial Intelligence

Report Warns AI Could Cause “Jobs Apocalypse”

By Frank Landymore

The study, which surveyed hundreds of university students, also found a link between heavy academic workload and increased ChatGPT usage.

ChatGPT Use Linked to Memory Loss, Procrastination in Students

By Frank Landymore

Be careful what you tell AI chatbots, because apparently it's very easy for hackers to figure out your prompts.

Hackers Can Spy on Your Chats With Almost Any AI, Experts Find

By Noor Al-Sibai

A recent study by Amazon AI researchers found that a "shocking amount of the web" comprises poor-quality AI-generated and translated content.

Artificial Intelligence

Huge Proportion of Internet Is AI-Generated Slime, Researchers Find

By Maggie Harrison Dupré

The winner of the Atakugawa Prize, Rie Kudan, said that about five percent of her novel was copied from ChatGPT and other generative AIs.

Novelist Wins Award, Then Reveals She Used ChatGPT

By Frank Landymore

A Microsoft researcher giving an internal presentation on an early demo of an AI tool admits that the responses were "cherry-picked."

Artificial Intelligence

In Leaked Audio, Microsoft Cherry-Picked Examples to Make Its AI Seem Functional

By Frank Landymore

In his latest tech industry scorecard, famed roboticist Rodney Brooks soberly warns of an impending "AI winter."

Artificial Intelligence

Robotics Expert Says AI Hype Is Due for a Brutal Reality Check

By Frank Landymore

Gary Gensler, chairman of the SEC, fears that the reliance on a small number of AI models could lead to the entire finance sector to it doom.

Artificial Intelligence

SEC Chair Warns AI “Herding” Could Drive Markets “Off an Inadvertent Cliff”

By Frank Landymore

The hack is "silly," but it gets ChatGPT to regurgitate incredible amounts of its training data, some of it potentially sensitive.

Hack Tricks ChatGPT Into Spitting Out Private Email Addresses, Phone Numbers

By Frank Landymore

In a new essay, Oxford researchers argue that scientists should abstain from using text-generating AI tools as research aids.

Artificial Intelligence

Scientists Warn That AI Threatens Science Itself

By Maggie Harrison Dupré

A study pitting human children against large language models in basic reasoning problems revealed AIs to have one big blind spot: innovation.

Artificial Intelligence

In New Experiment, Young Children Destroy AI at Basic Tasks

By Maggie Harrison Dupré

Google DeepMind researchers discovered something about AI models that may hamstring their employer's plans for more advanced AIs.

Google AI Researchers Found Something Their Bosses Might Not Be Happy About

By Noor Al-Sibai

Researchers have found a sad but unsurprising statistic: that nearly 10% of folks who use AI chatbots are doing so for horny purposes.

Artificial Intelligence

One in Ten Chatbot Users Are Big Time Horndogs, Researchers Find

By Noor Al-Sibai

Computer scientist Guillaume Cabanac spotted a telltale ChatGPT phrase in the paper, and is on a mission to uncover other frauds.

Paper Retracted When Authors Caught Using ChatGPT to Write It

By Frank Landymore

A New Zealand supermarket's AI recipe bot suggested users chef up a series of poisonous, deadly, and/or cannibalistic meals.

Artificial Intelligence

Supermarket’s Meal-Planning AI Suggests Deadly Poison for Dinner

By Maggie Harrison Dupré

Clippy's back! And this time, with a new daddy and a GPT-3.5 integration, it's more powerful than ever — unless Microsoft shuts it down.

Artificial Intelligence

Madman Brings Clippy Back as an AI

By Maggie Harrison Dupré

Starting in September, a popular intro-level coding course at Harvard University, CS50, will be taught by an AI professor.

Artificial Intelligence

Harvard Will Teach Students Using an AI Instructor Next Semester

By Maggie Harrison Dupré

AI researchers apparently have one particularly telling — and frankly, somewhat disturbing — meme for understanding AI systems: the Shoggoth.

Artificial Intelligence

Experts Have a Dark Joke About How AI Is Secretly a Horrifying Monster With a Friendly Mask

By Maggie Harrison Dupré

AI expert Rodney Brooks is arguing that we're overestimating OpenAI's large language models and that they're far stupider than we think.

AI Expert Says ChatGPT Is Way Stupider Than People Realize

By Victor Tangermann

Seance AI says that its tech will help the grieving find closure. It's not a happy story. But again, they never really are.

Artificial Intelligence

AI Company Says It’ll Perform a Seance on Your Dead Loved Ones

By Maggie Harrison Dupré

A team of South Korean researchers has created an AI model dubbed DarkBERT to index the Dark Web's websites and forums.

Artificial Intelligence

Scientists Train New AI Exclusively on the Dark Web

By Victor Tangermann

Fanfiction writers realized an OpenAI-powered writing assistant app was suddenly spitting out incredibly specific erotic fiction.

ChatGPT Happy to Write Smut About Freakishly Obscure Sex Act

By Maggie Harrison Dupré

Google just revealed vision for AI-integrated search — and that vision, apparently, involves cutting digital publishers off at the knees.

Google Unveils Plan to Demolish the Journalism Industry Using AI

By Maggie Harrison Dupré

We caught up with fired Google engineer and AI ethicist Blake Lemoine to chat about the AI industry, Google, and more.

We Interviewed the Engineer Google Fired for Saying Its AI Had Come to Life

By Maggie Harrison Dupré

Former Google CEO Eric Schmidt said that he's worried humans will start falling in love with AIs, using AI chatbot tutors as a prime example.

Former Google CEO Warns That Humans Will Fall in Love With AIs

By Maggie Harrison Dupré

When Stanford researchers compared ChatGPT's medical advice on real scenarios compared to what medical experts said, they discovered a notable discrepancy.

ChatGPT Can Pass Medical Tests, But Its Actual Medical Advice Is a Lot More Dubious

By Frank Landymore

Don't expect scam emails to be as obvious as a foreign prince strapped for cash writing with terrible grammar. AI fixes that problem easily.

Scammers Are Using ChatGPT to Write Emails That Aren’t Riddled With Typos

By Frank Landymore

A team of AI scientists is claiming that just like humans can solve problems through self-reflection, Large Language Model (LLM) AIs might be able to, too.

Artificial Intelligence

AI Seems to Do Better on Tasks When Asked to Reflect on Its Mistakes

By Maggie Harrison Dupré

Concerned about OpenAI and the implications of its powerful AI tech? Fair, but CEO Sam Altman says you should probably focus your worry on its competition.

OpenAI CEO Warns That Competitors Will Make AI That’s More Evil

By Maggie Harrison Dupré

Hot on the heels of the GPT-4 drop, OpenAI has released a bunch of stats about its even-more-powerful new language model — and we're slightly freaked.

OpenAI’s GPT-4 Just Smoked Basically Every Test and Exam Anyone’s Ever Taken

By Noor Al-Sibai

Facebook's new artificial intelligence model for researchers and governments has been leaked — and naturally, it ended up on 4chan.

Facebook’s Powerful ChatGPT Competitor Leaks on 4Chan

By Noor Al-Sibai

Wondering if artificial intelligence (AI) will be taking your job? A fair question — and one that a group of professors may have just shed some light on.

Artificial Intelligence

These Are the Jobs Most Vulnerable to AI, Researchers Say

By Maggie Harrison Dupré

Watch out, Big Tech: using a generative AI like ChatGPT to power a search engine uses way more computing power than before.

Turns Out Using ChatGPT In Search Engines Would Have a Grisly Environmental Footprint

By Frank Landymore

Artificial intelligence like ChatGPT is often wrong, and yet it's somehow always certain, even tending to double down on incorrect answers. Sound familiar?

ChatGPT Is Just an Automated Mansplaining Machine

By Maggie Harrison Dupré

According to OpenAI CTO Mira Murati, folks at OpenAI had some "trepidation" about taking their AI chatbot, ChatGPT, to market. Makes sense.

Head of ChatGPT Admits She Had Concerns About Releasing It Publicly

By Maggie Harrison Dupré

ChatGPT, OpenAI's groundbreaking language modeling AI, just passed a Wharton Business School exam. Queue yet another existential crisis!

Sorry, Suits: ChatGPT Just Passed a Wharton Business School Test

By Maggie Harrison Dupré

A new AI chatbot, Historical Figures, simulates conversations with famous leaders. Just one thing: every famous person in history is perfect. Noted.

Artificial Intelligence

AI of Andrew Jackson, Who Literally Owned Slaves, Insists That He Wasn’t Racist

By Maggie Harrison Dupré

Experts found that language-modeling AIs like ChatGPT will likely make it easier and more efficient than ever to generate massive amounts of misinformation.

Artificial Intelligence

Experts Warn of Nightmare Internet Filling With Infinite AI-Generated Propaganda

By Maggie Harrison Dupré

There already appears to be an generative AI-powered Substack, and it already appears to have word-for-word plagiarized human work. Cheers.

Artificial Intelligence

Viral AI-Written Article Busted as Plagiarized

By Maggie Harrison Dupré

Despite the VC gold rush, this expert is arguing that hype — as opposed to firm data and proven results — is in the generative AI industry driving seat.

Artificial Intelligence

Deep Learning Expert Says GPT Startups May Be in for a Very Rude Awakening

By Maggie Harrison Dupré

After just a few days online (and tons of Twitter criticism), Meta-formerly-Facebook has taken down its Galactica academic paper-writing AI.

Artificial Intelligence

Facebook Takes Down AI That Churns Out Fake Academic Papers After Widespread Criticism

By Noor Al-Sibai