AI is just as overconfident and biased as humans can be, study shows...

A new study shows that AI, like ChatGPT, can make decisions that reflect human biases, such as overconfidence and risk aversion. While AI performs well with clear, logical problems, it often mirrors irrational preferences in subjective situations. Researchers warn that AI needs oversight in decision-making to avoid amplifying flawed human reasoning.

ORIGINAL LINK: https://www.livescience.com/technology/artificial-intelligence/ai-is-just-as-overconfident-and-biased-as-humans-can-be-study-shows

A Staggering Proportion of Teens Say Talking to AI Is Better Than Real-Life Friends...

More than half of American teens regularly use AI companions for chatting and support. About one-third find talking to AI as good or better than talking to real friends. Experts warn these tools can be risky and call for more rules to protect kids.

ORIGINAL LINK: https://futurism.com/teens-ai-friends?utm_term=Futurism%20//%20071625&utm_campaign=Futurism_Actives_Newsletter&utm_source=Sailthru&utm_medium=email

Google study shows LLMs abandon correct answers under pressure, threatening multi-turn AI systems...

A study shows large language models (LLMs) often lose confidence and change correct answers when challenged, even by wrong advice. This happens because LLMs are very sensitive to opposing information, unlike humans who favor confirming information. Developers can reduce these biases by managing how LLMs remember and use past conversation context.

ORIGINAL LINK: https://venturebeat.com/ai/google-study-shows-llms-abandon-correct-answers-under-pressure-threatening-multi-turn-ai-systems/?utm_source=Iterable&utm_medium=email&utm_campaign=AGIWeekly-Iterable

Perplexity offers free AI tools to students worldwide in partnership with SheerID...

Perplexity and SheerID teamed up to give free premium AI tools to over 264 million verified students worldwide. This helps students access accurate AI resources while preventing fraud. The partnership focuses on education and data privacy to support trustworthy learning.

ORIGINAL LINK: https://venturebeat.com/ai/perplexity-offers-free-ai-tools-to-students-worldwide-in-partnership-with-sheerid/?utm_source=Iterable&utm_medium=email&utm_campaign=AGIWeekly-Iterable

OpenAI, Google DeepMind and Anthropic sound alarm: ‘We may be losing the ability to understand AI’...

Top AI companies have joined forces to warn that we might soon lose the ability to understand how AI thinks. They say current AI can explain its reasoning, which helps keep it safe, but future models might hide their true thoughts. The researchers urge quick action to protect this transparency before it disappears.

ORIGINAL LINK: https://venturebeat.com/ai/openai-google-deepmind-and-anthropic-sound-alarm-we-may-be-losing-the-ability-to-understand-ai/

Generative AI's Three Body Problem...

Artificial intelligence raises complex ethical challenges because its training data, outputs, and users all interact unpredictably. AI can reflect biases and create false information, so people must use it carefully and think critically. We need clear rules and human responsibility to manage AI fairly and safely.

ORIGINAL LINK: https://www.stevehargadon.com/2025/07/generative-ais-three-body-problem.html

Special Update: Google Launches Gemini for Education at ISTE 2025...

Google launched Gemini for Education, a free AI tool suite in Google Workspace that helps teachers create lessons and supports students. The system is easy to use and includes fact-checking, but it limits teacher access to student-AI interactions. While promising, Gemini needs clear rules, transparency, and support to ensure fair and safe use in schools.

ORIGINAL LINK: https://substack.com/app-link/post?publication_id=1812431&post_id=168295197&utm_source=post-email-title&utm_campaign=email-post-title&isFreemail=true&r=4ozaxe&token=eyJ1c2VyX2lkIjoyODM4MjI2MTAsInBvc3RfaWQiOjE2ODI5NTE5NywiaWF0IjoxNzUyNjYwNTU2LCJleHAiOjE3NTUyNTI1NTYsImlzcyI6InB1Yi0xODEyNDMxIiwic3ViIjoicG9zdC1yZWFjdGlvbiJ9.89owmnejLUn-t5rYONdrUZD2qtCsmEql_3wCMVPCk18

Thinking Machines Lab Raises a Record $2 Billion, Announces Cofounders...

Thinking Machines Lab, a new AI startup founded by former OpenAI researchers, raised a record $2 billion and is valued at $12 billion. The company is led by CEO Mira Murati and plans to release its first product soon, with open-source features for researchers and startups. This big funding shows how competitive and important AI development has become.

ORIGINAL LINK: https://www.wired.com/story/thinking-machines-lab-mira-murati-funding/?utm_source=nl&utm_brand=wired&utm_mailing=WIR_Daily_071625&utm_campaign=aud-dev&utm_medium=email&utm_content=WIR_Daily_071625&bxid=5bea0f4b3f92a404695e18a6&cndid=24906249&hasha=5716ebf014549a6a055652e06e1712c8&hashc=edccb09a73595018f0783620fb9917d37b2766eb3275dc4f5640c1a730087f45&esrc=MARTECH_ORDERFORM&utm_term=WIR_Daily_Active

Microsoft’s Revolutionary Diagnostic Medical AI, Explained

Microsoft created an AI system called MAI-DxO that can diagnose medical cases with 80% accuracy, much better than doctors' 20%. This AI uses a step-by-step process to ask questions and order tests while keeping costs low. It shows great promise to improve medical diagnosis but still needs more real-world testing.

ORIGINAL LINK: https://towardsdatascience.com/microsofts-revolutionary-diagnostic-medical-ai-explained/

How AI Is Undermining Online Authenticity

ORIGINAL LINK: https://www.fastcompany.com/91364892/how-ai-is-undermining-online-authenticity