-
Reading AI summaries makes people more likely to buy something — despite alarming 60% hallucination rateA project that found AI summaries are likely to majorly influence buying decisions raises interesting and potentially disturbing questions about how much we trust AI-generated content.
-
Scientists made AI agents ruder — and they performed better at complex reasoning tasksA new project allowed AI chatbots to interrupt, stay silent or speak up the way humans do in conversation, and it made them smarter and more accurate.
4 Comments -
Your own voice could be your biggest privacy threat. How can we stop AI technologies exploiting it?Voices contain countless cues about their owners, and new research suggests that computers might use them to facilitate a range of bad behaviors.
-
AI can develop 'personality' spontaneously with minimal prompting, research shows. What does that mean for how we use it?When large language models (LLMs) are allowed to interact without any preset goals, scientists found distinct personalities emerged by themselves.
4 Comments -
Will AI ever be more creative than humans?A new study argues that AI can never be more creative than humans, but many experts argue that AI's output will only ever be as good as its input — with the goalposts shifting as AI improves in the years to come.
6 Comments -
The more that people use AI, the more likely they are to overestimate their own abilitiesResearchers found that AI flattens the bell curve of a common principle in human psychology, known as the Dunning-Kruger effect, giving us all the illusion of competence.
4 Comments -
Scientists asked ChatGPT to solve a math problem from more than 2,000 years ago — how it answered it surprised themWe've wondered for centuries whether knowledge is latent and innate or learned and grasped through experience, and a new research project is asking the same question about AI.
16 Comments -
There are 32 different ways AI can go rogue, scientists say — from hallucinating answers to a complete misalignment with humanityNew research has created the first comprehensive effort to categorize all the ways AI can go wrong, with many of those behaviors resembling human psychiatric disorders.
12 Comments -
New study claims AI 'understands' emotion better than usCommon AI models outperformed humans on emotional intelligence in a recent study, but experts caution us to look beyond the headline.
2 Comments -
Replika AI chatbot is sexually harassing users, including minors, new study claimsUser reviews of Replika, a popular AI companion, report they had been victims of sexual harassment. And some of those users claim to be minors, according to a new study.
9 Comments -
AI models can't tell time or read a calendar, study revealsChallenges in visual and spatial processing and a deficit in training data have revealed a surprising lack of timekeeping ability in AI systems
8 Comments -
AI is just as overconfident and biased as humans can be, study showsIrrational tendencies — including the hot hand, base-rate neglect and sunk cost fallacy — commonly show up in AI systems, calling into question how useful they actually are.
3 Comments -
Scientists discover major differences in how humans and AI 'think' — and the implications could be significantStudy finds that AI fundamentally lacks the human capability to make creative mental connections, raising warning signs for how we deploy AI tools.
12 Comments -
Traumatizing AI models by talking about war or violence makes them more anxiousA recent study exposing AI models to carefully designed prompts around trauma revealed they can get anxious, potentially affecting the conversation and having negative impacts on people who use such models to discuss their mental health.
2 Comments -
People find AI more compassionate than mental health experts, study finds. What could this mean for future counseling?People find AI more compassionate and understanding than human mental health experts, a new study shows. Even when participants knew that they were talking to a human or AI, the third-party assessors rated AI responses higher.
-
Older AI models show signs of cognitive decline, study shows — but not everyone is entirely convincedOlder chatbots show signs of cognitive impairment, failing on several important metrics in a test normally used on humans.
32 Comments -
Supercomputer runs largest and most complicated simulation of the universe everFrontier, the second fastest supercomputer in the world, used dark matter and the movement of gas and plasma rather than just gravity to model the observable universe.
6 Comments -
How long does it take to travel to the moon?The answer depends on many factors, including the amount of fuel needed, the moon's orbit and the mission's objectives.
-
We are instinctively turned off by stories labeled 'AI-generated' — even if they were secretly written by other people, study findsEven when a creative work is made by humans, we think it’s missing a key element of art if we think it's been written by an AI — that capacity to fully immerse you in the world being created.
-
'Future You' AI lets you speak to a 60-year-old version of yourself — and it has surprising wellbeing benefitsAn MIT-led project asked young users to talk to an AI-powered simulation of their 60-year-old selves through a chatbot interface. The experience led to decreased anxiety and a boost in optimism.
-
Common AI models believe racist stereotypes about African Americans that predate the Civil Rights movement — and they 'try to hide it when confronted'When exposed to terms common in different racial dialects, large language models make racist assumptions about people from particular racial groups, even without explicitly knowing their race.
-
Would you prefer AI make major life decisions for you? Study suggests yes — but you'd be much happier if a human did itMost people much prefer algorithms to make decisions about the redistribution of resources — but they see the outcomes of human-made decisions as more favorable.
-
New in-vehicle AI algorithm can spot drunk drivers by constantly scanning their faces for signs of intoxicationAn in-vehicle camera can continuously watch you for signs of intoxication, with a new AI algorithm pinpointing your level of drunkenness with a 75% accuracy.
-
Most ChatGPT users think AI models have 'conscious experiences'The more people use tools like ChatGPT, the more likely they are to think they are conscious, which will carry ramifications for legal and ethical approaches to AI.
