No shocker here: You have to make decisions in work and in life. It would seem the choices you make get harder and have bigger, longer-lasting consequences as one gets older—adulting, right? But many of us had to make at least a few huge decisions right out of the gate, for instance choosing a college and signing up for many years of college loan repayments (or not). Unless you can latch on to someone else to manage your shit, you’ve got to make choices. Not to say that allowing someone else to manage a subset of decisions is a bad thing but there’s the extreme example of never growing up and hoping your parents figure out your life. Likewise, for married couples ideally it’s not just one person making all the choices—for example if your spouse is good at planning travel and you’re better at fixing the house that’s a great division of decision-making efforts. Anyway, making choices is part of the game. Literally you’ll be in a world of pain if you don’t. And you have a time limit for most decisions. If you keep waffling around in your mind things could get bad real quick. …a man, being...

Humans excel at processing vast arrays of visual information, a skill that is crucial for achieving artificial general intelligence (AGI). Over the decades, AI researchers have developed Visual Question Answering (VQA) systems to interpret scenes within single images and answer related questions. While recent advancements in foundation models have significantly closed the gap between human and machine visual processing, conventional VQA has been restricted to reason about only single images at a time rather than whole collections of visual data. This limitation poses challenges in more complex scenarios. Take, for example, the challenges of discerning patterns in collections of medical images, monitoring deforestation through satellite imagery, mapping urban changes using autonomous navigation data, analyzing thematic elements across large art collections, or understanding consumer behavior from retail surveillance footage. Each of these scenarios entails not only visual processing across hundreds or thousands of images but also necessitates cross-image processing of these findings. To address this gap, this project focuses on the “Multi-Image Question Answering” (MIQA) task, which exceeds the reach of traditional VQA systems. Visual Haystacks: the first “visual-centric” Needle-In-A-Haystack (NIAH) benchmark designed to rigorously evaluate Large Multimodal Models (LMMs) in processing long-context visual information.

Google AI Creates Fictional Folksy Sayings Google AI Creates Fictional Folksy Sayings a phrase that once may have sounded like science fiction is now rooted in reality, and it’s grabbing plenty of attention. Are we watching the rise of machine-made wisdom or just glitchy entertainment wrapped in nostalgia? As Google rolls out its new AI Overview feature, many users are discovering its ability to fabricate convincing but completely fake idioms. If you’re curious about how artificial intelligence is shaping our language and potentially misguiding anyone who trusts its responses read on. Also Read: Fully automated warehouse Understanding Google’s AI Overview Feature Google’s AI Overviews were introduced as part of its evolution toward generative search. The feature uses Google’s Gemini AI model to provide summarized answers at the top of search results, aiming to save users time by skipping irrelevant web pages. Instead of showing a list of links like traditional search, AI Overview distills information into conversational responses using data scraped from various corners of the internet. The convenience comes at a cost. Within days of the wider rollout in May 2024, users began noticing some unusual output. One viral example included a response advising people to add glue to...

More women are turning to ChatGPT for emotional support, using the AI chatbot as a stand-in therapist as mental health systems buckle under pressure. With long wait times and soaring costs, AI is filling a growing gap. Mental health care is harder to access than ever. In the UK, NHS data shows patients are eight times more likely to wait over 18 months for mental health treatment than for physical health. Private therapy isn’t always an option either, with sessions costing £60 or more. In that vacuum, ChatGPT has become a surprising outlet. Real voices, real feelings Charly, 29, from London, turned to ChatGPT while grappling with her grandmother’s terminal illness: “It’s been so helpful to ask the crass, the gruesome, the almost cruel questions about death… the things I feel twisted for wanting to understand.” Ellie, 27, from South Wales, said it helped her feel seen when no one else was around: “It didn’t have full context to my life like my therapist does, but it was accessible and non-judgmental in times of crisis.” Julia, 30, in Munich, used it when her therapist was booked up. The responses felt similar to a therapy app: “I was surprised at how...

Imagine a radiologist examining a chest X-ray from a new patient. She notices the patient has swelling in the tissue but does not have an enlarged heart. Looking to speed up diagnosis, she might use a vision-language machine-learning model to search for reports from similar patients. But if the model mistakenly identifies reports with both conditions, the most likely diagnosis could be quite different: If a patient has tissue swelling and an enlarged heart, the condition is very likely to be cardiac related, but with no enlarged heart there could be several underlying causes. In a new study, MIT researchers have found that vision-language models are extremely likely to make such a mistake in real-world situations because they don’t understand negation — words like “no” and “doesn’t” that specify what is false or absent.  “Those negation words can have a very significant impact, and if we are just using these models blindly, we may run into catastrophic consequences,” says Kumail Alhamoud, an MIT graduate student and lead author of this study. The researchers tested the ability of vision-language models to identify negation in image captions. The models often performed as well as a random guess. Building on those findings, the...