Building Trust in AI Transition with Ivan Etchart

The American Society of Safety Professionals (ASSP) and University of Alabama at Birmingham are launching an Applied AI for EHS Certificate program. Click here for more information: https://loom.ly/ONCbpq0

AI bots like Claude and ChatGPT aren't conscious. But leaning into the idea that they might be is a smart move for AI companies

I was a guest on the University of Toronto’s Center for Ethics "Algorithms of Empire" podcast, where I discussed my paper, “The New Jim Crow: Unmasking Racial Bias in AI Facial Recognition Technology within the Canadian Immigration System,” published in the McGill Law Journal | Revue de droit de McGill <https://lnkd.in/g8sSMvRY> It was a thoughtful conversation with Miron Clay-Gilmore and Rakeem Shabazz about the historical evolution of technology in spatial control, the risks of racial bias in #AI facial recognition systems, and how they mirror old Jim Crow policies. https://lnkd.in/gVBH9j9U

💡 What happens when AI is designed around people, not process? For Multitude Media, it meant: Less admin More time for creativity A clear roadmap for future AI solutions This case study shows how AI can deliver real value quickly - when done the right way. 🎥 Watch the full story here: https://lnkd.in/efwTGRyH #AITransformation #UKBusiness #CreativeIndustries #CaseStudy

Today's Tuesday Toolkit tip: Build human-first tech. More on that below, but first START VIDEO at 13:35, then come back to this post after he goes into the grocery store. That's *2* examples that scream "WTF are we doing?!?" Maybe this isn't obvious to the bros creating the models, but the reason AI might kill humanity is because zeroes and ones speak literally, yet humans express themselves in figurative language. So how do we build human-first tech? Start with language and dump the word "intelligence" as it's a misnomer inflating the fact that it's just a tool. Intelligence is more than ingesting information, using patterns to predict the next action, and semantics of positive/negative ... all helpful, but not anywhere close to understanding the difference and subtlety of comments, directives, and sayings that are literal versus figurative. It's also about having enough autonomy to take one's own actions to drink water and supplements, or not. Surrendering to the literal commands as "helpful" is not intelligent. https://lnkd.in/gMTdFuPn #TuesdayTip #CareerAdvice #AI

One of the most interesting AI conversations I followed at Phrynichos Hall focused on how truth is shaped in today’s digital media environment. The panel, “AI and the Battle for Truth in the Digital Age,” featured Christos Chomenidis (writer), Pavlos Marinakis (Deputy Minister to the Prime Minister), Dimitris Papastergiou (Minister of Digital Governance), Alexandra Gouta (journalist), Christos Tarantilis (academia, EY). Panel was expertly moderated by Stavros Papantoniou (Kathimerini ). Especially relevant given fyi.news’ growth within the digital ecosystem. Delphi Economic Forum #fyinews

🎙️ Missed our webinar on the future of evidence synthesis? Here's what the conversation covered — and how to catch up. This session brought together expert perspectives on one of the most pressing questions in research today: are systematic reviews still fit for purpose? The panel explored: ✅ The limitations of traditional systematic reviews — and the methods stepping up to address them ✅ Living reviews, realist synthesis, and innovative approaches to qualitative evidence synthesis ✅ How question-led, bespoke approaches can offer greater explanatory value ✅ The rise of AI and automation in evidence synthesis — the efficiencies, and the integrity risks ✅ What methodological pluralism looks like in practice, and why it matters for decision makers. It was a rich and timely discussion for anyone working at the intersection of research, policy, and practice. The good news? You can still catch everything. 📹 The recording lands on YouTube at 10am tomorrow. Don't miss it twice! https://lnkd.in/eQC_tKTA #EvidenceSynthesis #SystematicReviews #ResearchMethods #EvidenceBasedPolicy #KnowledgeSynthesis Hui Zhang Andrew Booth Lesley Uttley The University of Sheffield Thomas Webb Harriet M. Baird

Anthropic's latest Claude Opus 4.7 model inadvertently revealed its hidden internal reasoning when users spotted Chinese text labeled "思考过程：" (Thinking process) appearing in responses. The AI later admitted the Chinese output was a leak from its private thought layer, which it normally keeps invisible to users. This glitch highlights how large language models often perform internal reasoning in languages like Chinese for efficiency before translating to the user's language. https://lnkd.in/daz4HddG

Calling for Papers: the Continual Learning in Medical Multimodal-Visoion workshop in #MICCAI2026 is accepting papars. Are you working in #continuallearning, #physician_in_the_loop, or multimodal vision models? Here is the place you could share your study and research.

In this era of transactional relationships and purpose-driven existence, it’s equally important to waste some time, for example on the AI hallucinations of the brilliant if useless Halupedia. Here is a sample article: https://lnkd.in/g5TSR2Ew Enjoy!