I just watched the ACM TechTalk "Ethical and Responsible Large Language Models: Challenges and Best Practices" https://on.acm.org/t/ethical-and-responsible-large-language-models-challenges-and-best-practices/2901

For some reason I thought it was to be about Ethical and Responsible #LLMs, but during the talk I made a (for me) shocking discovery: When #AI researchers talk about #ExplainableAI they mean that they are able to locate the level in the deep learning network that decided the next token in the response.

#EthicalAI

1/n

Delighted to share my latest #phd research in #ExplainableAI. 'Can we Agree? On the Rashōmon Effect and Reliability of Post-Hoc Explainable AI' investigates how to find a explainable consensus between multiple #ML models. Discover our findings on arXiv https://arxiv.org/abs/2308.07247

How Important Is Explainability in Cybersecurity AI? - Artificial intelligence is transforming many industries but few as dramatically as... - https://readwrite.com/how-important-is-explainability-in-cybersecurity-ai/ #dataandsecurity #explainableai #security #ai

Explainable Artificial Intelligence (XAI)

https://www.techinsiderbuzz.com/2023/07/explainable-artificial-intelligence-xai.html

Discover the importance of Explainable AI (XAI) and how it can increase transparency and trust in AI systems. Read on to learn more.

#AI #artificialintelligence #artificial_intelligence #artificialintelligencetechnology #tech #technology #technologies #blog #blogger #blogpost #blogging #XAI #explainableai #TechInsiderBuzz

女児虐待の疑い、AIの「保護率39％」評価参考に児相が保護見送り：朝日新聞デジタル
https://digital.asahi.com/articles/ASR7B5F6NR7BONFB009.html?ptoken=01H4ZVKHVVHWVZJ8901XSMVDK3

これは責任が問われかねないやつやな。24時間全文無料。

#AI #ExplainableAI

Our pick of the week by Dennis Fucci: "Explanations for Automatic Speech Recognition" (Wu et al., 2023 #ICASSP).

https://ieeexplore.ieee.org/document/10094635

#NLProc #NLP #Speech #Recognition #ASR #SpeechRecognition #explanation #explainableAI #AI

Three more refined questions, when we are interested in explaining the decision of a #MachineLearning classifier (an area of #explainableai
What is the reason for a decision on an example?
What are the minimal aspects of the example that will guarantee the decision on that example?
What are the minimal changes to the example that will lead to a different decision?

https://arxiv.org/abs/2305.05172​​​​​​​

#logic

SHAP を用いた機械学習への解釈性付与│データ分析とケモインフォ
https://datadriven-rnd.com/shap/#:~:text=SHAP%E3%81%A8%E3%81%AF,%E3%82%92%E6%B1%82%E3%82%81%E3%82%8B%E3%81%93%E3%81%A8%E3%81%8C%E3%81%A7%E3%81%8D%E3%82%8B%E3%80%82

#datascience #ai #explainableai #shap

It surprises me that researchers are surprised that #LLMs like #ChatGPT, #Bing, and #Bard affect the way you think.

https://dl.acm.org/doi/pdf/10.1145/3544548.3581196

#AI changes who you are, so choose your thinking tools carefully.
I like you.com ( #YouChat ). At least it was trained to TRY to give you the #truth. And it is #ExplainableAI, it gives references. I can see the humans that inspired its answers.
#AgentGPT also looks promising as a way to ground LLMs in the truth and reduce bias.

PhD studentship at Brunel University London: Funded opportunity to pursue a PhD in Computer Science in Validating, Monitoring and Explaining model decisions.  

Further details including funding and application process are:
https://www.findaphd.com/phds/project/validating-monitoring-and-explaining-model-decisions/?p157744

The application deadline is Friday 26th of May 2023.
#phdfunding #academicchat #academicchatter #research #explainableai

5 emerging trends in deep learning and artificial intelligence - Explore five emerging trends in deep learning and artificial inte... - https://cointelegraph.com/news/5-emerging-trends-in-deep-learning-and-artificial-intelligence #reinforcementlearning #federatedlearning #transferlearning #explainableai #deeplearning #transparency #dataprivacy

Fascinating.

"This project aims to lay out a method to explain the behavior of music-tagging AI systems by breaking down tracks into their instrumental sources (stems) and seeing how the classification changes. This method was inspired by a well-known Explainable AI (XAI) concept called Shapley Values."

https://towardsdatascience.com/making-music-tagging-ai-explainable-through-source-separation-2d9493547a7e

#ArtificialIntelligence #AI #DeepLearning #ExplainableAI #Music #Analysis #Classification #Tag #Tagging #GameTheory #ShapleyValues #Coalition #Outcome

#ExplainableAI 👇
https://technews.social/@geekinchief/110068764212456133

@emilymbender @zoeschiffer @caseynewton

An important and maybe instructive precedent of attempting to regulate "AI" concerns algorithms that determine access to the financial system (KYC, Credit Scoring etc).

This domain is "ripe" for disruption and a grave risk under a deregulatory regime. While ML type algorithms have been tried decades ago, and are marginal on their own, this was under very different privacy, volume of data collection and triangulation cultures.

Bigtech / adtech is very eager to get out of its sandbox and colonise the circulatory system of society. I would speculate that the reason it hasn't happened yet is because there is effective resistance from other vested interests who would rather keep the surplus for themselves.

The trojan horse that will be used seems to be what has been termed "#explainableAI"

https://www.equifax.com.au/knowledge-hub/risk-solutions/explainable-ai-credit-scoring

https://en.wikipedia.org/wiki/Explainable_artificial_intelligence

RT @IJCAIconf
The 1977 IJCAI Computer and Thoughts Award went to Douglas Lenat #cyc 👇 For his latest interview check out the @lexfridman link below.

🚨 #Call for IJCAI-23 Awards is open and awaiting your input👉 https://www.ijcai.org/awards

#IJCAI2023

#reasoning #symbolic #ML #explainableAI https://twitter.com/lexfridman/status/1438182402942197774

To everyone who wants MORE than just a SIGMOID output 🙅‍♀️ 🙅‍♂️
Matthias Jakobs, his team (and me) are planning to host a Workshop on #ExplainableAI (#XAI) for #TimeSeries #Data at the #ECML #PKDD 2023 and we need your #Feedback!!
⁉️ https://forms.gle/YQJME4uS4QaYZTJb7
See you in Turin 🇮🇹😁

Accuracy is just one piece of the puzzle when it comes to #MachineLearning. With Fo-XAI, our latest open source library for Explainable AI (XAI), you can uncover the reasoning behind your models' predictions. Try it out today!

https://www.reasonfieldlab.com/case-studies/fo-xai-library-for-explainable-ai

#AI #ExplainableAI

Efficient technique improves machine-learning models’ reliability | MIT News https://triangleagency.co.uk/efficient-technique-improves-machine-learning-models-reliability-mit-news/?utm_source=dlvr.it&utm_medium=mastodon #TheTriangleAgencyNews #Efficient #ExplainableAI #GregoryWornell

Our book club had an amazing chat with @mikemunnster and @edavepitman on their XAI @OReillyMedia book. This book is a must-read for anyone who is interested in #explainableai. Highly recommended! ❤️
https://www.youtube.com/watch?v=5vn8x0aFh-4

@garymarcus next I asked whether #ChatGPT uses #ExplainableAI. I have no information about this, but got the an affirmative answer.