#trustinai — Public Fediverse posts

via #Microsoft : Unlocking human ambition to drive business growth with AI

https://ift.tt/c7Kgdf0
#AI #ArtificialIntelligence #BusinessGrowth #FrontierTransformation #MicrosoftIQ #Copilot #Agent365 #Foundry #AzureOpenAI #Microsoft365Copilot #TrustInAI #IntelligencePlusTrust #…

Wie entsteht Vertrauen in agentische KI?

Gitta Kutyniok, Professorin an der Ludwig-Maximilians-Universität und Mitglied der Plattform @LernendeSysteme, macht deutlich:

🔍 Präzises Prompting schafft Klarheit.

🧩 Transparenz ist zentral für Vertrauen.

⚖️ Menschliche Kontrolle bleibt entscheidend.

Kurz: Vertrauen in KI-Agenten muss aktiv gestaltet werden.

🌐 Weitere Stimmen unserer Mitglieder zu #AgenticAI: https://www.plattform-lernende-systeme.de/ergebnisse/standpunkte/agentische-ki/articles/agentische-ki.html

#KI #TrustInAI

FYI: Only 15% of users trust AI search results, Yelp study finds: Yelp and Morning Consult surveyed 2,202 U.S. adults and found only 15% trust AI search platforms "a lot," with 51% describing results as a "walled garden." https://ppc.land/only-15-of-users-trust-ai-search-results-yelp-study-finds/ #AI #SearchResults #TrustInAI #YelpStudy #DigitalTrust

Only 15% of users trust AI search results, Yelp study finds: Yelp and Morning Consult surveyed 2,202 U.S. adults and found only 15% trust AI search platforms "a lot," with 51% describing results as a "walled garden." https://ppc.land/only-15-of-users-trust-ai-search-results-yelp-study-finds/ #AI #SearchResults #TrustInAI #YelpStudy #DigitalTrust

FYI: Most TV viewers distrust AI search results, Gracenote study finds: Gracenote's 2026 study of 4,003 U.S. users finds AI chatbots gaining ground fast in TV content discovery, but 75% of all users still verify chatbot results. https://ppc.land/most-tv-viewers-distrust-ai-search-results-gracenote-study-finds/ #AI #Chatbots #TVContent #DigitalMedia #TrustInAI

#savethedate

📢 AI Forum: Auditing AI-Systems
📅 5.12.2025 | 📍 Berlin & Online

Beim 5. Internationalen Workshop treffen Wissenschaft, Industrie & Politik zu vertrauenswürdiger KI zusammen. Themen sind u. a.: Robot Learning, LLMs, AI Governance, Transparenz & Compliance.

Mit Keynote der EU-Kommission und u. a. den Mitgliedern der Plattform Lernende Systeme Johannes Hinckeldeyn, KION Group, und Sirko Straube, @DFKI

👉 https://www.tuev-verband.de/events/foren/ai-forum-2025

#AI #TrustworthyAI #TrustInAI #Robotics

Ethics and trust aren’t just buzzwords—they’re the foundation of responsible AI.

Let’s build systems that people can truly rely on.

#AWTOMATIG #AIEthics #TrustInAI #AIGovernance #ResponsibleAI #TechForGood

AI is everywhere in business, but trust? That's another story. This article dives into the 'AI trust gap,' where extensive adoption meets a serious lack of confidence. The solution? Transparency, empowering humans, and constant vigilance on ethics.

What's your biggest hurdle to trusting AI in the workplace?
#AI #TechEthics #BusinessAI #TrustInAI #FutureOfWork
https://www.artificialintelligence-news.com/news/how-to-fix-the-ai-trust-gap-in-your-business/

AI systems can easily lie and deceive us – a fact researchers are painfully aware of
#Tech #AI #Chatbots #AIModel #Anthropic #AISafety #AIAlignment #AIEthics #MachineLearning #AIResearch #FutureOfAI #TrustInAI #ResponsibleAI #AIDeception #AIConcerns #LLM
https://the-14.com/ai-systems-can-easily-lie-and-deceive-us-a-fact-researchers-are-painfully-aware-of/

Forget 'AI will take our jobs,' the real hurdle for AI growth is much simpler: we don't trust it. A report highlights a massive public trust deficit, especially among those who haven't touched generative AI. Yet, if it's sorting traffic, we're all in. If it's watching *us*? Suddenly it's Skynet.

#AIethics #TrustInAI #TechDebate #FutureOfWork #AIgrowth
Where do you draw the line with AI's purpose?
Link: https://www.artificialintelligence-news.com/news/public-trust-deficit-major-hurdle-for-ai-growth/

As autonomous AI systems make purchasing decisions, SMBs must address accountability, transparency, and validation to safeguard brand trust and prevent costly errors. #AIethics #RiskManagement #TrustInAI

https://www.techradar.com/pro/when-ai-buys-from-ai-who-do-we-trust

AI: Explainable Enough

They look really juicy, she said. I was sitting in a small room with a faint chemical smell, doing one my first customer interviews. There is a sweet spot between going too deep and asserting a position. Good AI has to be just explainable enough to satisfy the user without overwhelming them with information. Luckily, I wasn’t new to the problem. 

Coming from a microscopy and bio background with a strong inclination towards image analysis I had picked up deep learning as a way to be lazy in lab. Why bother figuring out features of interest when you can have a computer do it for you, was my angle. The issue was that in 2015 no biologist would accept any kind of deep learning analysis and definitely not if you couldn’t explain the details. 

What the domain expert user doesn’t want:
– How a convolutional neural network works. Confidence scores, loss, AUC, are all meaningless to a biologist and also to a doctor. 

What the domain expert desires: 
– Help at the lowest level of detail that they care about. 
– AI identifies features A, B, C, and that when you see A, B, & C it is likely to be disease X. 

Most users don’t care how a deep learning really works. So, if you start giving them details like the IoU score of the object detection bounding box or if it was YOLO or R-CNN that you used their eyes will glaze over and you will never get a customer. Draw a bounding box, heat map, or outline, with the predicted label and stop there. It’s also bad to go to the other extreme. If the AI just states the diagnosis for the whole image then the AI might be right, but the user does not get to participate in the process. Not to mention regulatory risk goes way up.

This applies beyong images, consider LLMs. No one with any expertise likes a black box. Today, why do LLMs generate code instead of directly doing the thing that the programmer is asking them to do? It’s because the programmer wants to ensure that the code “works” and they have the expertise to figure out if and when it goes wrong. It’s the same reason that vibe coding is great for prototyping but not for production and why frequent readers can spot AI patterns, ahem,  easily.  So in a Betty Crocker cake mix kind of way, let the user add the egg. 

Building explainable-enough AI takes immense effort. It actually is easier to train AI to diagnose the whole image or to give details. Generating high-quality data at that just right level is very difficult and expensive. However, do it right and the effort pays off. The outcome is an AI-Human causal prediction machine. Where the causes, i.e. the median level features, inform the user and build confidence towards the final outcome. The deep learning part is still a black box but the user doesn’t mind because you aid their thinking. 

I’m excited by some new developments like REX which sort of retro-fit causality onto usual deep learning models. With improvements in performance user preferences for detail may change, but I suspect that need for AI to be explainable enough will remain. Perhaps we will even have custom labels like ‘juicy’.

#AI #AIAdoption #AICommunication #AIExplainability #AIForDoctors #AIInHealthcare #AIInTheWild #AIProductDesign #AIUX #artificialIntelligence #BettyCrockerThinking #BiomedicalAI #Business #CausalAI #DataProductDesign #DeepLearning #ExplainableAI #HumanAIInteraction #ImageAnalysis #LLMs #MachineLearning #StartupLessons #statistics #TechMetaphors #techPhilosophy #TrustInAI #UserCenteredAI #XAI

AI: Explainable Enough

They look really juicy, she said. I was sitting in a small room with a faint chemical smell, doing one my first customer interviews. There is a sweet spot between going too deep and asserting a position. Good AI has to be just explainable enough to satisfy the user without overwhelming them with information. Luckily, I wasn’t new to the problem. 

Coming from a microscopy and bio background with a strong inclination towards image analysis I had picked up deep learning as a way to be lazy in lab. Why bother figuring out features of interest when you can have a computer do it for you, was my angle. The issue was that in 2015 no biologist would accept any kind of deep learning analysis and definitely not if you couldn’t explain the details. 

What the domain expert user doesn’t want:
– How a convolutional neural network works. Confidence scores, loss, AUC, are all meaningless to a biologist and also to a doctor. 

What the domain expert desires: 
– Help at the lowest level of detail that they care about. 
– AI identifies features A, B, C, and that when you see A, B, & C it is likely to be disease X. 

Most users don’t care how a deep learning really works. So, if you start giving them details like the IoU score of the object detection bounding box or if it was YOLO or R-CNN that you used their eyes will glaze over and you will never get a customer. Draw a bounding box, heat map, or outline, with the predicted label and stop there. It’s also bad to go to the other extreme. If the AI just states the diagnosis for the whole image then the AI might be right, but the user does not get to participate in the process. Not to mention regulatory risk goes way up.

This applies beyong images, consider LLMs. No one with any expertise likes a black box. Today, why do LLMs generate code instead of directly doing the thing that the programmer is asking them to do? It’s because the programmer wants to ensure that the code “works” and they have the expertise to figure out if and when it goes wrong. It’s the same reason that vibe coding is great for prototyping but not for production and why frequent readers can spot AI patterns, ahem,  easily.  So in a Betty Crocker cake mix kind of way, let the user add the egg. 

Building explainable-enough AI takes immense effort. It actually is easier to train AI to diagnose the whole image or to give details. Generating high-quality data at that just right level is very difficult and expensive. However, do it right and the effort pays off. The outcome is an AI-Human causal prediction machine. Where the causes, i.e. the median level features, inform the user and build confidence towards the final outcome. The deep learning part is still a black box but the user doesn’t mind because you aid their thinking. 

I’m excited by some new developments like REX which sort of retro-fit causality onto usual deep learning models. With improvements in performance user preferences for detail may change, but I suspect that need for AI to be explainable enough will remain. Perhaps we will even have custom labels like ‘juicy’.

#AI #AIAdoption #AICommunication #AIExplainability #AIForDoctors #AIInHealthcare #AIInTheWild #AIProductDesign #AIUX #artificialIntelligence #BettyCrockerThinking #BiomedicalAI #Business #CausalAI #DataProductDesign #DeepLearning #ExplainableAI #HumanAIInteraction #ImageAnalysis #LLMs #MachineLearning #StartupLessons #statistics #TechMetaphors #techPhilosophy #TrustInAI #UserCenteredAI #XAI

AI: Explainable Enough

They look really juicy, she said. I was sitting in a small room with a faint chemical smell, doing one my first customer interviews. There is a sweet spot between going too deep and asserting a position. Good AI has to be just explainable enough to satisfy the user without overwhelming them with information. Luckily, I wasn’t new to the problem. 

Coming from a microscopy and bio background with a strong inclination towards image analysis I had picked up deep learning as a way to be lazy in lab. Why bother figuring out features of interest when you can have a computer do it for you, was my angle. The issue was that in 2015 no biologist would accept any kind of deep learning analysis and definitely not if you couldn’t explain the details. 

What the domain expert user doesn’t want:
– How a convolutional neural network works. Confidence scores, loss, AUC, are all meaningless to a biologist and also to a doctor. 

What the domain expert desires: 
– Help at the lowest level of detail that they care about. 
– AI identifies features A, B, C, and that when you see A, B, & C it is likely to be disease X. 

Most users don’t care how a deep learning really works. So, if you start giving them details like the IoU score of the object detection bounding box or if it was YOLO or R-CNN that you used their eyes will glaze over and you will never get a customer. Draw a bounding box, heat map, or outline, with the predicted label and stop there. It’s also bad to go to the other extreme. If the AI just states the diagnosis for the whole image then the AI might be right, but the user does not get to participate in the process. Not to mention regulatory risk goes way up.

This applies beyong images, consider LLMs. No one with any expertise likes a black box. Today, why do LLMs generate code instead of directly doing the thing that the programmer is asking them to do? It’s because the programmer wants to ensure that the code “works” and they have the expertise to figure out if and when it goes wrong. It’s the same reason that vibe coding is great for prototyping but not for production and why frequent readers can spot AI patterns, ahem,  easily.  So in a Betty Crocker cake mix kind of way, let the user add the egg. 

Building explainable-enough AI takes immense effort. It actually is easier to train AI to diagnose the whole image or to give details. Generating high-quality data at that just right level is very difficult and expensive. However, do it right and the effort pays off. The outcome is an AI-Human causal prediction machine. Where the causes, i.e. the median level features, inform the user and build confidence towards the final outcome. The deep learning part is still a black box but the user doesn’t mind because you aid their thinking. 

I’m excited by some new developments like REX which sort of retro-fit causality onto usual deep learning models. With improvements in performance user preferences for detail may change, but I suspect that need for AI to be explainable enough will remain. Perhaps we will even have custom labels like ‘juicy’.

#AI #AIAdoption #AICommunication #AIExplainability #AIForDoctors #AIInHealthcare #AIInTheWild #AIProductDesign #AIUX #artificialIntelligence #BettyCrockerThinking #BiomedicalAI #Business #CausalAI #DataProductDesign #DeepLearning #ExplainableAI #HumanAIInteraction #ImageAnalysis #LLMs #MachineLearning #StartupLessons #statistics #TechMetaphors #techPhilosophy #TrustInAI #UserCenteredAI #XAI

AI: Explainable Enough

They look really juicy, she said. I was sitting in a small room with a faint chemical smell, doing one my first customer interviews. There is a sweet spot between going too deep and asserting a position. Good AI has to be just explainable enough to satisfy the user without overwhelming them with information. Luckily, I wasn’t new to the problem. 

Coming from a microscopy and bio background with a strong inclination towards image analysis I had picked up deep learning as a way to be lazy in lab. Why bother figuring out features of interest when you can have a computer do it for you, was my angle. The issue was that in 2015 no biologist would accept any kind of deep learning analysis and definitely not if you couldn’t explain the details. 

What the domain expert user doesn’t want:
– How a convolutional neural network works. Confidence scores, loss, AUC, are all meaningless to a biologist and also to a doctor. 

What the domain expert desires: 
– Help at the lowest level of detail that they care about. 
– AI identifies features A, B, C, and that when you see A, B, & C it is likely to be disease X. 

Most users don’t care how a deep learning really works. So, if you start giving them details like the IoU score of the object detection bounding box or if it was YOLO or R-CNN that you used their eyes will glaze over and you will never get a customer. Draw a bounding box, heat map, or outline, with the predicted label and stop there. It’s also bad to go to the other extreme. If the AI just states the diagnosis for the whole image then the AI might be right, but the user does not get to participate in the process. Not to mention regulatory risk goes way up.

This applies beyong images, consider LLMs. No one with any expertise likes a black box. Today, why do LLMs generate code instead of directly doing the thing that the programmer is asking them to do? It’s because the programmer wants to ensure that the code “works” and they have the expertise to figure out if and when it goes wrong. It’s the same reason that vibe coding is great for prototyping but not for production and why frequent readers can spot AI patterns, ahem,  easily.  So in a Betty Crocker cake mix kind of way, let the user add the egg. 

Building explainable-enough AI takes immense effort. It actually is easier to train AI to diagnose the whole image or to give details. Generating high-quality data at that just right level is very difficult and expensive. However, do it right and the effort pays off. The outcome is an AI-Human causal prediction machine. Where the causes, i.e. the median level features, inform the user and build confidence towards the final outcome. The deep learning part is still a black box but the user doesn’t mind because you aid their thinking. 

I’m excited by some new developments like REX which sort of retro-fit causality onto usual deep learning models. With improvements in performance user preferences for detail may change, but I suspect that need for AI to be explainable enough will remain. Perhaps we will even have custom labels like ‘juicy’.

#AI #AIAdoption #AICommunication #AIExplainability #AIForDoctors #AIInHealthcare #AIInTheWild #AIProductDesign #AIUX #artificialIntelligence #BettyCrockerThinking #BiomedicalAI #Business #CausalAI #DataProductDesign #DeepLearning #ExplainableAI #HumanAIInteraction #ImageAnalysis #LLMs #MachineLearning #StartupLessons #statistics #TechMetaphors #techPhilosophy #TrustInAI #UserCenteredAI #XAI

AI: Explainable Enough

They look really juicy, she said. I was sitting in a small room with a faint chemical smell, doing one my first customer interviews. There is a sweet spot between going too deep and asserting a position. Good AI has to be just explainable enough to satisfy the user without overwhelming them with information. Luckily, I wasn’t new to the problem. 

Coming from a microscopy and bio background with a strong inclination towards image analysis I had picked up deep learning as a way to be lazy in lab. Why bother figuring out features of interest when you can have a computer do it for you, was my angle. The issue was that in 2015 no biologist would accept any kind of deep learning analysis and definitely not if you couldn’t explain the details. 

What the domain expert user doesn’t want:
– How a convolutional neural network works. Confidence scores, loss, AUC, are all meaningless to a biologist and also to a doctor. 

What the domain expert desires: 
– Help at the lowest level of detail that they care about. 
– AI identifies features A, B, C, and that when you see A, B, & C it is likely to be disease X. 

Most users don’t care how a deep learning really works. So, if you start giving them details like the IoU score of the object detection bounding box or if it was YOLO or R-CNN that you used their eyes will glaze over and you will never get a customer. Draw a bounding box, heat map, or outline, with the predicted label and stop there. It’s also bad to go to the other extreme. If the AI just states the diagnosis for the whole image then the AI might be right, but the user does not get to participate in the process. Not to mention regulatory risk goes way up.

This applies beyong images, consider LLMs. No one with any expertise likes a black box. Today, why do LLMs generate code instead of directly doing the thing that the programmer is asking them to do? It’s because the programmer wants to ensure that the code “works” and they have the expertise to figure out if and when it goes wrong. It’s the same reason that vibe coding is great for prototyping but not for production and why frequent readers can spot AI patterns, ahem,  easily.  So in a Betty Crocker cake mix kind of way, let the user add the egg. 

Building explainable-enough AI takes immense effort. It actually is easier to train AI to diagnose the whole image or to give details. Generating high-quality data at that just right level is very difficult and expensive. However, do it right and the effort pays off. The outcome is an AI-Human causal prediction machine. Where the causes, i.e. the median level features, inform the user and build confidence towards the final outcome. The deep learning part is still a black box but the user doesn’t mind because you aid their thinking. 

I’m excited by some new developments like REX which sort of retro-fit causality onto usual deep learning models. With improvements in performance user preferences for detail may change, but I suspect that need for AI to be explainable enough will remain. Perhaps we will even have custom labels like ‘juicy’.

#AI #AIAdoption #AICommunication #AIExplainability #AIForDoctors #AIInHealthcare #AIInTheWild #AIProductDesign #AIUX #artificialIntelligence #BettyCrockerThinking #BiomedicalAI #Business #CausalAI #DataProductDesign #DeepLearning #ExplainableAI #HumanAIInteraction #ImageAnalysis #LLMs #MachineLearning #StartupLessons #statistics #TechMetaphors #techPhilosophy #TrustInAI #UserCenteredAI #XAI

What do real developers think about AI coding tools?

At @wearedevelopers, we asked engineers how AI fits into their workflow. The answers were smart, honest, and not always what the hype says.

No black boxes. No vibe coding. Just systems you can trust.

Read the full post:
🔗 https://www.leapter.com/ai-coding-meets-the-real-world/

#AI #FOSS #SoftwareDevelopment #DevTools #WeAreDevelopers #Leapter #TrustInAI

Link to the article:
https://www.listentoyourmuse.com/data-ethics

#AIethics #artificialintelligence #machinebias #NorbertWiener #JamesMoor #ethicalAI #AIdecisionmaking #technologyandethics #trustinAI #AIbias #computerethics #globalAIimpact
#newsletter #muse #listentoyourmuse #technology #technologynews

Link to the article:
https://www.listentoyourmuse.com/data-ethics

#AIethics #artificialintelligence #machinebias #NorbertWiener #JamesMoor #ethicalAI #AIdecisionmaking #technologyandethics #trustinAI #AIbias #computerethics #globalAIimpact
#newsletter #muse #listentoyourmuse #technology #technologynews

A UN survey across 21 countries shows trust in AI is highest in low-income and developing nations, with 83% in China and only 37.5% in the US expressing trust. #AI #TrustInAI #China #US #GlobalTrends #TechPerception #DigitalDivide #ArtificialIntelligence #UNSurvey

Before AI can be widely adopted, people must trust it, especially that it can make accurate and fair decisions. AI should be aware of and aligned with human values. #TrustInAI #AIAdoption

People trust legal advice generated by ChatGPT more than a lawyer – new study
#AI #ChatGPT #LegalAdvice #LawAndTechnology #AIEthics #AILiteracy #AIRegulation #LLM #ArtificialIntelligence #AIRisk #LegalTech #TrustInAI #AIResearch #FutureOfLaw
https://the-14.com/people-trust-legal-advice-generated-by-chatgpt-more-than-a-lawyer-new-study/

Study Finds People Prefer AI Legal Advice Blind, Even If They Can Spot It

#AI #LLMs #GenAI #CHI2025 #LegalTech #AIethics #TrustInAI #AIsafety #Legal #Law

https://winbuzzer.com/2025/04/28/study-finds-people-prefer-ai-legal-advice-blind-even-if-they-can-spot-it-xcxwbn/

Study shows a gap in how AI is perceived: experts are more optimistic about its benefits and less concerned about risks compared to the general public.
Is this just a case of expertise leading to realism—or are experts overlooking public fears?
#AI #Ethics #TrustInAI
https://arxiv.org/abs/2412.01459