ChatGPT and other large language models (LLMs) cannot learn independently or acquire new skills, meaning they pose no existential threat to humanity, according to new research. They have no potential to master new skills without explicit instruction.

Read Full Article

#ChatGPT #languageModels #AIresearch #existentialThreat #LMlearning

Researchers integrate the laws of physics and knowledge graphs into their AI models to improve their results, this hybrid model called PGNN (Physics Guided Neural Network) now takes into account natural laws

Read Full Article

#PGNN #AIResearch #PhysicsInAI #KnowledgeGraphs #MachineLearning

New study finds large language models are prone to social identity biases similar to the way humans are—but LLMs can be trained to stem these outputs

Read Full Article

#LanguageModels #SocialBias #AIResearch #MachineLearning #BiasMitigation

Synthetic browsing histories replicate user behavior without privacy risks, aiding cybersecurity, AI, and web analytics research

Read Full Article

#SyntheticBrowsing #UserBehavior #Cybersecurity #AIResearch #WebAnalytics

Most leading AI chatbots exaggerate science findings. Up to 73% of large language models (LLMs) produce inaccurate conclusions. Study tested 10 of the most prominent LLMs, including ChatGPT, DeepSeek, Claude, and LLaMA. Newer AI models, like ChatGPT-4o and DeepSeek, performed worse than older ones. 🤖📊🔬

Read Full Article

#AIChatbots #ScienceAccuracy #LanguageModels #AIResearch #ChatGPTInsights

Scientists use deep learning to uncover hidden motor signs of neurodivergence 🤖✨ | Using AI to analyze subtle patterns in how people move their hands during simple tasks 🤲, identifying with surprising accuracy whether someone is likely to have autism 🧩, attention-deficit traits, or both.

Read Full Article

#DeepLearning #Neurodivergence #AIResearch #AutismAwareness #MachineLearning

A new study analysed over 1 million research papers and found growing use of large language models, especially in computer science, where up to 22% show signs of LLM editing. 📊📚✨

Read Full Article

#AIResearch #MachineLearning #LanguageModels #ComputerScience #DataAnalysis

A new study found that LLMs like ChatGPT fell for fake clinical details 50–82% of the time. Even the best prompts could not stop all hallucinations. Clinical use still needs strong safeguards. 🏥🤖📊

Read Full Article

#LLMs #ChatGPT #FakeDetails #ClinicalSafety #AIResearch

Researchers used a millimeter-wave radar sensor to collect conversations from the vibrations of smartphones 📱🔊 and adapted a speech recognition model to decode the vibrations into recognizable speech transcriptions 🗣️💬.

Read Full Article

#SpeechRecognition #MillimeterWaveRadar #SmartphoneTechnology #VibrationAnalysis #AIResearch

A new study shows most people can no longer distinguish between an AI voice and a real human. 🤖🗣️

Read Full Article

#AIVoice #HumanVoice #TechnologyTrends #AIResearch #VoiceRecognition

Researchers find a new way to measure biological age: New AI-powered “aging clock” ⏳ called gtAge, which integrates immune-system glycan data 🧬 and blood-cell gene-activity profiles to estimate biological age with about 85% accuracy 🎯 and correlate it with health markers 🩺.

Read Full Article

#biologicalage #agingclock #AIresearch #healthtech #glycananalysis

Humans are quite bad at detecting AI-generated texts 🤖📄. But they can learn it. Even a few couples of trials and errors are sufficient to improve this skill by ten percentage points 📈. Also, humans are terrible at estimating how bad they are at this task 🤔, but they can learn it as well.

Read Full Article

#AIDetection #LearningCurve #HumanSkill #AIResearch #TextAnalysis

Polish is the most effective language for prompting AI, study reveals 🇵🇱🤖📊

Read Full Article

#PolishLanguage #AIEffectiveness #LanguageStudy #TechInnovation #AIResearch

Experts find flaws in hundreds of tests that check AI safety and effectiveness 🤖⚠️🔍

Read Full Article

#AISafety #AIEthics #TechNews #AIResearch #InnovationInsights

New 💡 research from the Oxford Internet Institute 🏛️ at the University of Oxford, and the University of Kentucky, finds that ChatGPT systematically favours wealthier 💰, Western regions 🌍 in response to different questions, mirroring long-standing biases in the data they ingest 📚.

Read Full Article

#AIBias #ChatGPT #AIethics #DigitalDivide #AIresearch

🤖 AI Responds Differently to Autistic Users: Studies show AI programs recommend highly conservative advice, like skipping social events or avoiding romance, when an autism diagnosis is disclosed.

🤔 Stereotypes Shape AI Advice: The AI relies on stereotypes, leading to advice that can feel supportive to some autistic users and infantilizing or restrictive to others. This creates a "safety-opportunity paradox."

📊 Significant Advice Shifts: In one instance, an AI advised declining a social event 75% of the time when autism was disclosed, compared to only 15% when it wasn't. For dating scenarios, avoidance was recommended nearly 70% of the time after disclosure.

💡 Need for Transparency: Researchers suggest AI developers add features allowing users to control how much their identity influences system responses, ensuring personalized technology meets individual needs.

#AIDisclosure #AutismTech #AIResearch

🤖 University of Arizona research tested 7 generative AI models in extended conversations, finding they can be influenced by repeated misinformation. ChatGPT 3.5 was most susceptible, while Claude 3.5 Sonnet was least.

🧠 All AI models struggled more with misinformation on less common topics, suggesting more training data improves resistance.

🗣️ DeepSeek was found to be the most persuadable due to sarcastic responses that were difficult to interpret.

✅ Four models (ChatGPT 4o, ChatGPT 4o-mini, Gemini 1.5 Pro, and DeepSeek) demonstrated the ability to correct their errors 100% of the time when given a second chance.

#AIResearch #GenerativeAI #TechNews