GPT-4 is judged more human than humans in displaced and inverted Turing tests Rathi et al.: https://
arxiv.org/abs/2407.08853 #AIAgent #ArtificialIntelligence #DeepLearning
RESEARCH
-
GPT-4 Passes Inverted Turing Test, Appears More Human
By
–
-
GLP-1 Drugs Study Obesity Patients Without Diabetes
By
–
TriX Global Network with index event between 1 January 2016 and 31 March 2024. Details are in the paper and supplemental info, includes any GLP-1 drug new user for at least 6 months in people with obesity, without diabetes
-
FLUX LoRAs Achieve Human-Level Image Generation Quality
By
–
Text has definitely passed the threshold. Even human likeness with this current wave of FLUX LoRAs. I can’t even imagine what we’ll see in the next way of CVPR and ICCV papers
-
Unconscious AI Consumption and User Awareness Gap
By
–
Yeah when you ask people to look up they will be in a discriminative mindset and more likely to spot it. But as we see with boomers on fb — they’re nomming on ai without even knowing or caring
-
AI Content Authenticity: How Much Online Content Is Genuinely Human?
By
–
IMO: 100% AI content? Easy to spot. Human + AI mix? Much harder. Makes you wonder: how much online content is truly authentic human expression? Some even argue for a "Dead Internet Theory" – that bots overtook us circa 2017. Wdyt? How often do we encounter AI content online?
-
OODA Loop Model Ultimate Decision Making Framework Explained
By
–
See text and graphic here: https://
oodaloop.com/the-ooda-loop-
explained-the-real-story-about-the-ultimate-model-for-decision-making-in-competitive-environments/
… -
Is Xi Jinping an AI Doomer? China’s AI Risk Perspective
By
–
Is Xi Jinping an #AI doomer? https://
economist.com/china/2024/08/
25/is-xi-jinping-an-ai-doomer
… -
Scammers Exploit Messaging Apps for 45% of Attack Vectors
By
–
Scammers are increasingly using messaging and social media apps to attack: Meta platforms, alongside Telegram, are among the growing number of sites used as a form of contact in 45% of scams. https://
zdnet.com/article/scamme
rs-are-increasingly-using-messaging-and-social-media-apps-to-attack/?utm_source=dlvr.it&utm_medium=twitter#ftag=RSSbaffb68
… #BigData #DataScience #IoT -
Chihuahuas or Muffins: Image Recognition Challenges in ML
By
–
Is it chihuahuas or muffins? This meme illustrates challenges in training machines on image recognition. Here are the risks associated with #MachineLearning in #AI: Misclassification Risks: AI models can struggle to differentiate between objects with similar visual features
-
Synthetic Data Enables Superhuman LLM Performance via RL Training
By
–
Synthetic data for LLMS and RL/RLHF/DPO can both train superhuman performance, model permitting.