Regulating Generative AI - LLMOps pipelines with Transparency

Debmalya Biswas
Debmalya BiswasLead Enterprise Architect - Artificial Intelligence à Philip Morris International
REGULATING GENERATIVE AI — LLMOPS
PIPELINES WITH TRANSPARENCY
DEBMALYA BISWAS, WIPRO AI
AGENDA
 Enterprise AI
 Ethical / Responsible AI
 Explainability
 Fairness & Bias
 LLMOps Architecture Patterns
 Generative AI - Responsible
Design Principles
ENTERPRISE AI
• Enterprise AI use-
cases are pervasive
RESPONSIBLE AI
“Ethical AI, also known as responsible AI, is the practice of using AI with good
intention to empower employees and businesses, and fairly impact customers
and society. Ethical AI enables companies to engender trust and scale AI with
confidence.” [1]
Failing to operationalize Ethical AI can not only expose enterprises to
reputational, regulatory, and legal risks; but also lead to wasted resources,
inefficiencies in product development, and even an inability to use data to train
AI models. [2]
[1] R. Porter. Beyond the promise: implementing Ethical AI, 2020
(link)
[2] R. Blackman. A Practical Guide to Building Ethical AI, 2020 (link)
REGULATIONS
 Good news: is that there has been a recent
trend towards ensuring that AI applications are
responsibly trained and deployed, in line with
the enterprise strategy and policies.
 Bad news: Efforts have been complicated by
different governmental organizations and
regulatory bodies releasing their own
guidelines and policies; with little to no
standardization on the definition of terms.
 For example, the EU AI Act mandates a different
set of dos & don’ts depending on the ‘risk level’
of an AI application. However, quantifying the
risk level of an AI application is easier said than
done as it basically requires you to classify how
the capabilities of a non-deterministic system
will impact users and systems who might
interact with it in the future.
ETHICAL AI PRINCIPLES
 Explainability
 Bias & Fairness
 Accountability
 Reproducibility
 Data Privacy
*D. Biswas. Ethical AI: its implications for Enterprise AI Use-
cases and Governance. Towards Data Science (link)
*D. Biswas. Privacy Preserving Chatbot Conversations.
3rd IEEE AIKE 2020: 179-182
EXPLAINABLE AI
 Explainable AI is an umbrella term for
a range of tools, algorithms and
methods; which accompany AI model
predictions with explanations.
 Explainability of AI models ranks high
among the list of ‘non-functional’ AI
features to be considered by
enterprises.
 For example, this implies having to
explain why an ML model profiled
a user to be in a specific segment
— which led him/her to receiving
an advertisement.
(Labeled)
Data
Train ML
Model
Predictions
Explanation
Model
Explainable
Predictions
EXPLAINABLE AI FRAMEWORKS - LIME
 Local Interpretable
Model-Agnostic
Explanations (LIME*)
provides easy to
understand explanations
of a prediction by training
an explainability model
based on samples around
a prediction.
 The approximate nature
of the explainability
model might limit its
usage for compliance
needs. *M. T. Ribeiro, S. Singh, C. Guestrin. “Why Should I Trust You?”
Explaining the Predictions of Any Classifier, 2016 (link)
LIME output showing the important features,
positively and negatively impacting the model’s
prediction.
EXPLAINABLE AI - FEASIBILITY
 Machine (Deep) Learning algorithms
vary in the level of accuracy and
explainability that they can provide-
the two are often inversely
proportional.
 Explainability starts becoming more
difficult as as we move to Random
Forests, which are basically an
ensemble of Decision Trees. At the
end of the spectrum are Neural
Networks (Deep Learning), which
have shown human-level accuracy.
Explainability
Accuracy
Logistic Regression
DecisionTrees
Random Forest
(Ensemble of
DecisionTrees)
Deep Learning
(Neural Networks)
EXPLAINABLE AI - ABSTRACTION
“important thing is to explain the right thing to the right person in the right way at the right
time”*
Singapore AI Governance framework: “technical explainability may not always be enlightening,
esp. to the man in the street… providing an individual with counterfactuals (such as “you would
have been approved if your average debt was 15% lower” or “these are users with similar profiles
to yours that received a different decision”) can be a powerful type of explanation”
*N. Xie, et. al. Explainable Deep Learning: A
Field Guide for the Uninitiated, 2020 (link)
AI Developer
Goal:ensure/improve
performance
Regulatory Bodies
Goal:Ensure compliance with legislation,
protect interests of constituents
End Users
Goal:Understanding of
decision,trust model output
FAIRNESS & BIAS
 Bias is a phenomenon that occurs when an algorithm
produces results that are systemically prejudiced due
to erroneous assumptions in the machine learning
process*.
 AI models should behave in all fairness towards
everyone, without any bias. However, defining
‘fairness’ is easier said than done.
 Does fairness mean, e.g., that the same proportion
of male and female applicants get high risk
assessment scores?
 Or that the same level of risk result in the same
score regardless of gender?
 (Impossible to fulfill both)
* SearchEnterprise
AI. Machine Learning bias (AI
bias) (link)
Google Photo labeling pictures of a black
Haitian-American programmer as “gorilla”
“White Barack Obama”
images (link)
A computer program used for bail and
sentencing decisions was labeled biased
against blacks. (link)
TYPES OF BIAS
 Bias creeps into AI models, primarily due
to the inherent bias already present in the
training data. So the ‘data’ part of AI
model development is key to addressing
bias.
 Historical Bias: arises due to historical
inequality of human decisions
captured in the training data
 Representation Bias: arises due to
training data that is not representative
of the actual population
 Ensure that training data is representative
and uniformly distributed over the target
population - with respect to the selected
features. Source: H. Suresh, J. V. Guttag. A Framework for
Understanding Unintended Consequences of Machine
LLMOPS: MLOPS FOR LLMS
*D. Biswas. MLOps for Compositional AI. NeurIPS Workshop on Challenges in
Deploying and Monitoring Machine Learning Systems (DMML), 2022.
*D. Biswas. Generative AI – LLMOps Architecture Patterns. Data Driven Investor,
2023 (link)
 Black-box LLM APIs: This is the
classic ChatGPT example, where
we have black-box access to a
LLM API/UI. Prompts are the
primary interaction mechanism for
such scenarios.
 While Enterprise LLM Apps have
the potential to be a multi-billion
dollar marketplace and accelerate
LLM adoption by providing an
enterprise ready solution; the
same caution needs to be
exercised as you would do before
using a 3rd party ML model —
validate LLM/training data
ownership, IP, liability clauses.
LLMOPS: MLOPS FOR LLMS (2)
*D. Biswas. Contextualizing Large Language Models (LLMs)
with Enterprise Data. Data Driven Investor, 2023 (link)
 LLMs are generic in nature.
To realize the full potential
of LLMs for Enterprises, they
need to be contextualized
with enterprise knowledge
captured in terms of
documents, wikis, business
processes, etc.
 This is achieved by fine-
tuning a LLM with enterprise
knowledge / embeddings to
develop a context-specific
LLM.
GENERATIVE AI - RESPONSIBLE DESIGN PRINCIPLES
We take inspiration from
the “enterprise friendly”
Microsoft, “developer
friendly” Google and “user
friendly” Apple — to
enable this ‘transparent’
approach to Gen AI
system design.
• Guidelines for Human-
AI Interaction by
Microsoft
• People + AI
Guidebook by Google
• Machine Learning:
Human Interface
Guidelines by Apple
Thank
You
&
Question
s
Contact: Debmalya Biswas
LinkedIn:
https://www.linkedin.com/in/debmalya-
biswas-3975261/
Medium:
https://medium.com/@debmalyabiswas
1 sur 16

Recommandé

Generative-AI-in-enterprise-20230615.pdf par
Generative-AI-in-enterprise-20230615.pdfGenerative-AI-in-enterprise-20230615.pdf
Generative-AI-in-enterprise-20230615.pdfLiming Zhu
857 vues7 diapositives
Large Language Models - Chat AI.pdf par
Large Language Models - Chat AI.pdfLarge Language Models - Chat AI.pdf
Large Language Models - Chat AI.pdfDavid Rostcheck
720 vues19 diapositives
Unlocking the Power of Generative AI An Executive's Guide.pdf par
Unlocking the Power of Generative AI An Executive's Guide.pdfUnlocking the Power of Generative AI An Executive's Guide.pdf
Unlocking the Power of Generative AI An Executive's Guide.pdfPremNaraindas1
2.2K vues29 diapositives
The Future of AI is Generative not Discriminative 5/26/2021 par
The Future of AI is Generative not Discriminative 5/26/2021The Future of AI is Generative not Discriminative 5/26/2021
The Future of AI is Generative not Discriminative 5/26/2021Steve Omohundro
979 vues24 diapositives
AI and ML Series - Introduction to Generative AI and LLMs - Session 1 par
AI and ML Series - Introduction to Generative AI and LLMs - Session 1AI and ML Series - Introduction to Generative AI and LLMs - Session 1
AI and ML Series - Introduction to Generative AI and LLMs - Session 1DianaGray10
1.3K vues38 diapositives
What is MLOps par
What is MLOpsWhat is MLOps
What is MLOpsHenrik Skogström
2.6K vues16 diapositives

Contenu connexe

Tendances

Basics of Generative AI: Models, Tokenization, Embeddings, Text Similarity, V... par
Basics of Generative AI: Models, Tokenization, Embeddings, Text Similarity, V...Basics of Generative AI: Models, Tokenization, Embeddings, Text Similarity, V...
Basics of Generative AI: Models, Tokenization, Embeddings, Text Similarity, V...Robert McDermott
76 vues16 diapositives
GENERATIVE AI, THE FUTURE OF PRODUCTIVITY par
GENERATIVE AI, THE FUTURE OF PRODUCTIVITYGENERATIVE AI, THE FUTURE OF PRODUCTIVITY
GENERATIVE AI, THE FUTURE OF PRODUCTIVITYAndre Muscat
6.6K vues19 diapositives
Large Language Models, No-Code, and Responsible AI - Trends in Applied NLP in... par
Large Language Models, No-Code, and Responsible AI - Trends in Applied NLP in...Large Language Models, No-Code, and Responsible AI - Trends in Applied NLP in...
Large Language Models, No-Code, and Responsible AI - Trends in Applied NLP in...David Talby
666 vues36 diapositives
Responsible Generative AI par
Responsible Generative AIResponsible Generative AI
Responsible Generative AICMassociates
249 vues12 diapositives
LLMs Bootcamp par
LLMs BootcampLLMs Bootcamp
LLMs BootcampFiza987241
134 vues12 diapositives
[DSC DACH 23] ChatGPT and Beyond: How generative AI is Changing the way peopl... par
[DSC DACH 23] ChatGPT and Beyond: How generative AI is Changing the way peopl...[DSC DACH 23] ChatGPT and Beyond: How generative AI is Changing the way peopl...
[DSC DACH 23] ChatGPT and Beyond: How generative AI is Changing the way peopl...DataScienceConferenc1
294 vues16 diapositives

Tendances(20)

Basics of Generative AI: Models, Tokenization, Embeddings, Text Similarity, V... par Robert McDermott
Basics of Generative AI: Models, Tokenization, Embeddings, Text Similarity, V...Basics of Generative AI: Models, Tokenization, Embeddings, Text Similarity, V...
Basics of Generative AI: Models, Tokenization, Embeddings, Text Similarity, V...
GENERATIVE AI, THE FUTURE OF PRODUCTIVITY par Andre Muscat
GENERATIVE AI, THE FUTURE OF PRODUCTIVITYGENERATIVE AI, THE FUTURE OF PRODUCTIVITY
GENERATIVE AI, THE FUTURE OF PRODUCTIVITY
Andre Muscat6.6K vues
Large Language Models, No-Code, and Responsible AI - Trends in Applied NLP in... par David Talby
Large Language Models, No-Code, and Responsible AI - Trends in Applied NLP in...Large Language Models, No-Code, and Responsible AI - Trends in Applied NLP in...
Large Language Models, No-Code, and Responsible AI - Trends in Applied NLP in...
David Talby666 vues
Responsible Generative AI par CMassociates
Responsible Generative AIResponsible Generative AI
Responsible Generative AI
CMassociates249 vues
[DSC DACH 23] ChatGPT and Beyond: How generative AI is Changing the way peopl... par DataScienceConferenc1
[DSC DACH 23] ChatGPT and Beyond: How generative AI is Changing the way peopl...[DSC DACH 23] ChatGPT and Beyond: How generative AI is Changing the way peopl...
[DSC DACH 23] ChatGPT and Beyond: How generative AI is Changing the way peopl...
Landscape of AI/ML in 2023 par HyunJoon Jung
Landscape of AI/ML in 2023Landscape of AI/ML in 2023
Landscape of AI/ML in 2023
HyunJoon Jung2.4K vues
Let's talk about GPT: A crash course in Generative AI for researchers par Steven Van Vaerenbergh
Let's talk about GPT: A crash course in Generative AI for researchersLet's talk about GPT: A crash course in Generative AI for researchers
Let's talk about GPT: A crash course in Generative AI for researchers
𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐀𝐈: 𝐂𝐡𝐚𝐧𝐠𝐢𝐧𝐠 𝐇𝐨𝐰 𝐁𝐮𝐬𝐢𝐧𝐞𝐬𝐬 𝐈𝐧𝐧𝐨𝐯𝐚𝐭𝐞𝐬 𝐚𝐧𝐝 𝐎𝐩𝐞𝐫𝐚𝐭𝐞𝐬 par VINCI Digital - Industrial IoT (IIoT) Strategic Advisory
𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐀𝐈: 𝐂𝐡𝐚𝐧𝐠𝐢𝐧𝐠 𝐇𝐨𝐰 𝐁𝐮𝐬𝐢𝐧𝐞𝐬𝐬 𝐈𝐧𝐧𝐨𝐯𝐚𝐭𝐞𝐬 𝐚𝐧𝐝 𝐎𝐩𝐞𝐫𝐚𝐭𝐞𝐬𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐀𝐈: 𝐂𝐡𝐚𝐧𝐠𝐢𝐧𝐠 𝐇𝐨𝐰 𝐁𝐮𝐬𝐢𝐧𝐞𝐬𝐬 𝐈𝐧𝐧𝐨𝐯𝐚𝐭𝐞𝐬 𝐚𝐧𝐝 𝐎𝐩𝐞𝐫𝐚𝐭𝐞𝐬
𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐀𝐈: 𝐂𝐡𝐚𝐧𝐠𝐢𝐧𝐠 𝐇𝐨𝐰 𝐁𝐮𝐬𝐢𝐧𝐞𝐬𝐬 𝐈𝐧𝐧𝐨𝐯𝐚𝐭𝐞𝐬 𝐚𝐧𝐝 𝐎𝐩𝐞𝐫𝐚𝐭𝐞𝐬
How Does Generative AI Actually Work? (a quick semi-technical introduction to... par ssuser4edc93
How Does Generative AI Actually Work? (a quick semi-technical introduction to...How Does Generative AI Actually Work? (a quick semi-technical introduction to...
How Does Generative AI Actually Work? (a quick semi-technical introduction to...
ssuser4edc93975 vues
MLOps and Data Quality: Deploying Reliable ML Models in Production par Provectus
MLOps and Data Quality: Deploying Reliable ML Models in ProductionMLOps and Data Quality: Deploying Reliable ML Models in Production
MLOps and Data Quality: Deploying Reliable ML Models in Production
Provectus215 vues
An Introduction to Generative AI par Cori Faklaris
An Introduction  to Generative AIAn Introduction  to Generative AI
An Introduction to Generative AI
Cori Faklaris11.6K vues
MLOps Bridging the gap between Data Scientists and Ops. par Knoldus Inc.
MLOps Bridging the gap between Data Scientists and Ops.MLOps Bridging the gap between Data Scientists and Ops.
MLOps Bridging the gap between Data Scientists and Ops.
Knoldus Inc.440 vues
Leveraging Generative AI & Best practices par DianaGray10
Leveraging Generative AI & Best practicesLeveraging Generative AI & Best practices
Leveraging Generative AI & Best practices
DianaGray101.8K vues
Exploring Opportunities in the Generative AI Value Chain.pdf par Dung Hoang
Exploring Opportunities in the Generative AI Value Chain.pdfExploring Opportunities in the Generative AI Value Chain.pdf
Exploring Opportunities in the Generative AI Value Chain.pdf
Dung Hoang277 vues
Microsoft + OpenAI: Recent Updates (Machine Learning 15minutes! Broadcast #74) par Naoki (Neo) SATO
Microsoft + OpenAI: Recent Updates (Machine Learning 15minutes! Broadcast #74)Microsoft + OpenAI: Recent Updates (Machine Learning 15minutes! Broadcast #74)
Microsoft + OpenAI: Recent Updates (Machine Learning 15minutes! Broadcast #74)
Naoki (Neo) SATO1.3K vues
Machine Learning using Kubeflow and Kubernetes par Arun Gupta
Machine Learning using Kubeflow and KubernetesMachine Learning using Kubeflow and Kubernetes
Machine Learning using Kubeflow and Kubernetes
Arun Gupta2.3K vues

Similaire à Regulating Generative AI - LLMOps pipelines with Transparency

Ethical AI - Open Compliance Summit 2020 par
Ethical AI - Open Compliance Summit 2020Ethical AI - Open Compliance Summit 2020
Ethical AI - Open Compliance Summit 2020Debmalya Biswas
170 vues25 diapositives
AI and ML Trends to Impact Business 2020 par
AI and ML Trends to Impact Business 2020AI and ML Trends to Impact Business 2020
AI and ML Trends to Impact Business 2020Takayuki Yamazaki
133 vues13 diapositives
[REPORT PREVIEW] The Customer Experience of AI par
[REPORT PREVIEW] The Customer Experience of AI[REPORT PREVIEW] The Customer Experience of AI
[REPORT PREVIEW] The Customer Experience of AIAltimeter, a Prophet Company
11K vues10 diapositives
Accelerating Business and Research Through Automation and Artificial Intellig... par
Accelerating Business and Research Through Automation and Artificial Intellig...Accelerating Business and Research Through Automation and Artificial Intellig...
Accelerating Business and Research Through Automation and Artificial Intellig...CloudHesive
17 vues32 diapositives
State of AI Report 2023 - Air Street Capital par
State of AI Report 2023 - Air Street CapitalState of AI Report 2023 - Air Street Capital
State of AI Report 2023 - Air Street CapitalAI Geek (wishesh)
1.8K vues163 diapositives
Realizing_the_real_business_impact_of_gen_AI_white_paper.pdf par
Realizing_the_real_business_impact_of_gen_AI_white_paper.pdfRealizing_the_real_business_impact_of_gen_AI_white_paper.pdf
Realizing_the_real_business_impact_of_gen_AI_white_paper.pdfPhilipBasford
28 vues19 diapositives

Similaire à Regulating Generative AI - LLMOps pipelines with Transparency(20)

Ethical AI - Open Compliance Summit 2020 par Debmalya Biswas
Ethical AI - Open Compliance Summit 2020Ethical AI - Open Compliance Summit 2020
Ethical AI - Open Compliance Summit 2020
Debmalya Biswas170 vues
Accelerating Business and Research Through Automation and Artificial Intellig... par CloudHesive
Accelerating Business and Research Through Automation and Artificial Intellig...Accelerating Business and Research Through Automation and Artificial Intellig...
Accelerating Business and Research Through Automation and Artificial Intellig...
CloudHesive17 vues
State of AI Report 2023 - Air Street Capital par AI Geek (wishesh)
State of AI Report 2023 - Air Street CapitalState of AI Report 2023 - Air Street Capital
State of AI Report 2023 - Air Street Capital
AI Geek (wishesh)1.8K vues
Realizing_the_real_business_impact_of_gen_AI_white_paper.pdf par PhilipBasford
Realizing_the_real_business_impact_of_gen_AI_white_paper.pdfRealizing_the_real_business_impact_of_gen_AI_white_paper.pdf
Realizing_the_real_business_impact_of_gen_AI_white_paper.pdf
PhilipBasford28 vues
GDG Cloud Southlake 28 Brad Taylor and Shawn Augenstein Old Problems in the N... par James Anderson
GDG Cloud Southlake 28 Brad Taylor and Shawn Augenstein Old Problems in the N...GDG Cloud Southlake 28 Brad Taylor and Shawn Augenstein Old Problems in the N...
GDG Cloud Southlake 28 Brad Taylor and Shawn Augenstein Old Problems in the N...
James Anderson160 vues
AI Readiness: Five Areas Business Must Prepare for Success in Artificial Inte... par Kaleido Insights
AI Readiness: Five Areas Business Must Prepare for Success in Artificial Inte...AI Readiness: Five Areas Business Must Prepare for Success in Artificial Inte...
AI Readiness: Five Areas Business Must Prepare for Success in Artificial Inte...
Kaleido Insights 1.8K vues
Generative-AI-Exploring-beyond-the-horizons-possibilities-of-AI-WP.pdf par shashanksalunkhe12
Generative-AI-Exploring-beyond-the-horizons-possibilities-of-AI-WP.pdfGenerative-AI-Exploring-beyond-the-horizons-possibilities-of-AI-WP.pdf
Generative-AI-Exploring-beyond-the-horizons-possibilities-of-AI-WP.pdf
UNCOVERING FAKE NEWS BY MEANS OF SOCIAL NETWORK ANALYSIS par pijans
UNCOVERING FAKE NEWS BY MEANS OF SOCIAL NETWORK ANALYSISUNCOVERING FAKE NEWS BY MEANS OF SOCIAL NETWORK ANALYSIS
UNCOVERING FAKE NEWS BY MEANS OF SOCIAL NETWORK ANALYSIS
pijans7 vues
UNCOVERING FAKE NEWS BY MEANS OF SOCIAL NETWORK ANALYSIS par pijans
UNCOVERING FAKE NEWS BY MEANS OF SOCIAL NETWORK ANALYSISUNCOVERING FAKE NEWS BY MEANS OF SOCIAL NETWORK ANALYSIS
UNCOVERING FAKE NEWS BY MEANS OF SOCIAL NETWORK ANALYSIS
pijans4 vues
In the Dark? Understanding Big Data & AI: Talent Acquisition Strategies for 2018 par Yoh Staffing Solutions
In the Dark? Understanding Big Data & AI: Talent Acquisition Strategies for 2018In the Dark? Understanding Big Data & AI: Talent Acquisition Strategies for 2018
In the Dark? Understanding Big Data & AI: Talent Acquisition Strategies for 2018
artificial intelligence par haifa rzem
artificial intelligenceartificial intelligence
artificial intelligence
haifa rzem106 vues

Plus de Debmalya Biswas

Sustainable & Composable Generative AI par
Sustainable & Composable Generative AISustainable & Composable Generative AI
Sustainable & Composable Generative AIDebmalya Biswas
33 vues21 diapositives
Data-Driven (Reinforcement Learning-Based) Control par
Data-Driven (Reinforcement Learning-Based) ControlData-Driven (Reinforcement Learning-Based) Control
Data-Driven (Reinforcement Learning-Based) ControlDebmalya Biswas
127 vues17 diapositives
MLOps for Compositional AI par
MLOps for Compositional AIMLOps for Compositional AI
MLOps for Compositional AIDebmalya Biswas
64 vues13 diapositives
A Privacy Framework for Hierarchical Federated Learning par
A Privacy Framework for Hierarchical Federated LearningA Privacy Framework for Hierarchical Federated Learning
A Privacy Framework for Hierarchical Federated LearningDebmalya Biswas
607 vues17 diapositives
Edge AI Framework for Healthcare Applications par
Edge AI Framework for Healthcare ApplicationsEdge AI Framework for Healthcare Applications
Edge AI Framework for Healthcare ApplicationsDebmalya Biswas
304 vues14 diapositives
Compositional AI: Fusion of AI/ML Services par
Compositional AI: Fusion of AI/ML ServicesCompositional AI: Fusion of AI/ML Services
Compositional AI: Fusion of AI/ML ServicesDebmalya Biswas
332 vues24 diapositives

Plus de Debmalya Biswas(12)

Data-Driven (Reinforcement Learning-Based) Control par Debmalya Biswas
Data-Driven (Reinforcement Learning-Based) ControlData-Driven (Reinforcement Learning-Based) Control
Data-Driven (Reinforcement Learning-Based) Control
Debmalya Biswas127 vues
A Privacy Framework for Hierarchical Federated Learning par Debmalya Biswas
A Privacy Framework for Hierarchical Federated LearningA Privacy Framework for Hierarchical Federated Learning
A Privacy Framework for Hierarchical Federated Learning
Debmalya Biswas607 vues
Edge AI Framework for Healthcare Applications par Debmalya Biswas
Edge AI Framework for Healthcare ApplicationsEdge AI Framework for Healthcare Applications
Edge AI Framework for Healthcare Applications
Debmalya Biswas304 vues
Compositional AI: Fusion of AI/ML Services par Debmalya Biswas
Compositional AI: Fusion of AI/ML ServicesCompositional AI: Fusion of AI/ML Services
Compositional AI: Fusion of AI/ML Services
Debmalya Biswas332 vues
Privacy Preserving Chatbot Conversations par Debmalya Biswas
Privacy Preserving Chatbot ConversationsPrivacy Preserving Chatbot Conversations
Privacy Preserving Chatbot Conversations
Debmalya Biswas137 vues
Reinforcement Learning based HVAC Optimization in Factories par Debmalya Biswas
Reinforcement Learning based HVAC Optimization in FactoriesReinforcement Learning based HVAC Optimization in Factories
Reinforcement Learning based HVAC Optimization in Factories
Debmalya Biswas667 vues
Delayed Rewards in the context of Reinforcement Learning based Recommender ... par Debmalya Biswas
Delayed Rewards in the context of Reinforcement Learning based Recommender ...Delayed Rewards in the context of Reinforcement Learning based Recommender ...
Delayed Rewards in the context of Reinforcement Learning based Recommender ...
Debmalya Biswas249 vues
Building an enterprise Natural Language Search Engine with ElasticSearch and ... par Debmalya Biswas
Building an enterprise Natural Language Search Engine with ElasticSearch and ...Building an enterprise Natural Language Search Engine with ElasticSearch and ...
Building an enterprise Natural Language Search Engine with ElasticSearch and ...
Debmalya Biswas305 vues
Privacy-Preserving Outsourced Profiling par Debmalya Biswas
Privacy-Preserving Outsourced ProfilingPrivacy-Preserving Outsourced Profiling
Privacy-Preserving Outsourced Profiling
Debmalya Biswas530 vues
Privacy Policies Change Management for Smartphones par Debmalya Biswas
Privacy Policies Change Management for SmartphonesPrivacy Policies Change Management for Smartphones
Privacy Policies Change Management for Smartphones
Debmalya Biswas518 vues

Dernier

Enabling DPU Hardware Accelerators in XCP-ng Cloud Platform Environment - And... par
Enabling DPU Hardware Accelerators in XCP-ng Cloud Platform Environment - And...Enabling DPU Hardware Accelerators in XCP-ng Cloud Platform Environment - And...
Enabling DPU Hardware Accelerators in XCP-ng Cloud Platform Environment - And...ShapeBlue
106 vues12 diapositives
Backroll, News and Demo - Pierre Charton, Matthias Dhellin, Ousmane Diarra - ... par
Backroll, News and Demo - Pierre Charton, Matthias Dhellin, Ousmane Diarra - ...Backroll, News and Demo - Pierre Charton, Matthias Dhellin, Ousmane Diarra - ...
Backroll, News and Demo - Pierre Charton, Matthias Dhellin, Ousmane Diarra - ...ShapeBlue
186 vues15 diapositives
The Power of Heat Decarbonisation Plans in the Built Environment par
The Power of Heat Decarbonisation Plans in the Built EnvironmentThe Power of Heat Decarbonisation Plans in the Built Environment
The Power of Heat Decarbonisation Plans in the Built EnvironmentIES VE
79 vues20 diapositives
CloudStack Object Storage - An Introduction - Vladimir Petrov - ShapeBlue par
CloudStack Object Storage - An Introduction - Vladimir Petrov - ShapeBlueCloudStack Object Storage - An Introduction - Vladimir Petrov - ShapeBlue
CloudStack Object Storage - An Introduction - Vladimir Petrov - ShapeBlueShapeBlue
138 vues15 diapositives
Transitioning from VMware vCloud to Apache CloudStack: A Path to Profitabilit... par
Transitioning from VMware vCloud to Apache CloudStack: A Path to Profitabilit...Transitioning from VMware vCloud to Apache CloudStack: A Path to Profitabilit...
Transitioning from VMware vCloud to Apache CloudStack: A Path to Profitabilit...ShapeBlue
159 vues25 diapositives
TrustArc Webinar - Managing Online Tracking Technology Vendors_ A Checklist f... par
TrustArc Webinar - Managing Online Tracking Technology Vendors_ A Checklist f...TrustArc Webinar - Managing Online Tracking Technology Vendors_ A Checklist f...
TrustArc Webinar - Managing Online Tracking Technology Vendors_ A Checklist f...TrustArc
170 vues29 diapositives

Dernier(20)

Enabling DPU Hardware Accelerators in XCP-ng Cloud Platform Environment - And... par ShapeBlue
Enabling DPU Hardware Accelerators in XCP-ng Cloud Platform Environment - And...Enabling DPU Hardware Accelerators in XCP-ng Cloud Platform Environment - And...
Enabling DPU Hardware Accelerators in XCP-ng Cloud Platform Environment - And...
ShapeBlue106 vues
Backroll, News and Demo - Pierre Charton, Matthias Dhellin, Ousmane Diarra - ... par ShapeBlue
Backroll, News and Demo - Pierre Charton, Matthias Dhellin, Ousmane Diarra - ...Backroll, News and Demo - Pierre Charton, Matthias Dhellin, Ousmane Diarra - ...
Backroll, News and Demo - Pierre Charton, Matthias Dhellin, Ousmane Diarra - ...
ShapeBlue186 vues
The Power of Heat Decarbonisation Plans in the Built Environment par IES VE
The Power of Heat Decarbonisation Plans in the Built EnvironmentThe Power of Heat Decarbonisation Plans in the Built Environment
The Power of Heat Decarbonisation Plans in the Built Environment
IES VE79 vues
CloudStack Object Storage - An Introduction - Vladimir Petrov - ShapeBlue par ShapeBlue
CloudStack Object Storage - An Introduction - Vladimir Petrov - ShapeBlueCloudStack Object Storage - An Introduction - Vladimir Petrov - ShapeBlue
CloudStack Object Storage - An Introduction - Vladimir Petrov - ShapeBlue
ShapeBlue138 vues
Transitioning from VMware vCloud to Apache CloudStack: A Path to Profitabilit... par ShapeBlue
Transitioning from VMware vCloud to Apache CloudStack: A Path to Profitabilit...Transitioning from VMware vCloud to Apache CloudStack: A Path to Profitabilit...
Transitioning from VMware vCloud to Apache CloudStack: A Path to Profitabilit...
ShapeBlue159 vues
TrustArc Webinar - Managing Online Tracking Technology Vendors_ A Checklist f... par TrustArc
TrustArc Webinar - Managing Online Tracking Technology Vendors_ A Checklist f...TrustArc Webinar - Managing Online Tracking Technology Vendors_ A Checklist f...
TrustArc Webinar - Managing Online Tracking Technology Vendors_ A Checklist f...
TrustArc170 vues
KVM Security Groups Under the Hood - Wido den Hollander - Your.Online par ShapeBlue
KVM Security Groups Under the Hood - Wido den Hollander - Your.OnlineKVM Security Groups Under the Hood - Wido den Hollander - Your.Online
KVM Security Groups Under the Hood - Wido den Hollander - Your.Online
ShapeBlue221 vues
Import Export Virtual Machine for KVM Hypervisor - Ayush Pandey - University ... par ShapeBlue
Import Export Virtual Machine for KVM Hypervisor - Ayush Pandey - University ...Import Export Virtual Machine for KVM Hypervisor - Ayush Pandey - University ...
Import Export Virtual Machine for KVM Hypervisor - Ayush Pandey - University ...
ShapeBlue119 vues
Hypervisor Agnostic DRS in CloudStack - Brief overview & demo - Vishesh Jinda... par ShapeBlue
Hypervisor Agnostic DRS in CloudStack - Brief overview & demo - Vishesh Jinda...Hypervisor Agnostic DRS in CloudStack - Brief overview & demo - Vishesh Jinda...
Hypervisor Agnostic DRS in CloudStack - Brief overview & demo - Vishesh Jinda...
ShapeBlue161 vues
Webinar : Desperately Seeking Transformation - Part 2: Insights from leading... par The Digital Insurer
Webinar : Desperately Seeking Transformation - Part 2:  Insights from leading...Webinar : Desperately Seeking Transformation - Part 2:  Insights from leading...
Webinar : Desperately Seeking Transformation - Part 2: Insights from leading...
Business Analyst Series 2023 - Week 4 Session 8 par DianaGray10
Business Analyst Series 2023 -  Week 4 Session 8Business Analyst Series 2023 -  Week 4 Session 8
Business Analyst Series 2023 - Week 4 Session 8
DianaGray10123 vues
VNF Integration and Support in CloudStack - Wei Zhou - ShapeBlue par ShapeBlue
VNF Integration and Support in CloudStack - Wei Zhou - ShapeBlueVNF Integration and Support in CloudStack - Wei Zhou - ShapeBlue
VNF Integration and Support in CloudStack - Wei Zhou - ShapeBlue
ShapeBlue203 vues
Extending KVM Host HA for Non-NFS Storage - Alex Ivanov - StorPool par ShapeBlue
Extending KVM Host HA for Non-NFS Storage -  Alex Ivanov - StorPoolExtending KVM Host HA for Non-NFS Storage -  Alex Ivanov - StorPool
Extending KVM Host HA for Non-NFS Storage - Alex Ivanov - StorPool
ShapeBlue123 vues
DRBD Deep Dive - Philipp Reisner - LINBIT par ShapeBlue
DRBD Deep Dive - Philipp Reisner - LINBITDRBD Deep Dive - Philipp Reisner - LINBIT
DRBD Deep Dive - Philipp Reisner - LINBIT
ShapeBlue180 vues
Keynote Talk: Open Source is Not Dead - Charles Schulz - Vates par ShapeBlue
Keynote Talk: Open Source is Not Dead - Charles Schulz - VatesKeynote Talk: Open Source is Not Dead - Charles Schulz - Vates
Keynote Talk: Open Source is Not Dead - Charles Schulz - Vates
ShapeBlue252 vues
Setting Up Your First CloudStack Environment with Beginners Challenges - MD R... par ShapeBlue
Setting Up Your First CloudStack Environment with Beginners Challenges - MD R...Setting Up Your First CloudStack Environment with Beginners Challenges - MD R...
Setting Up Your First CloudStack Environment with Beginners Challenges - MD R...
ShapeBlue173 vues
Migrating VMware Infra to KVM Using CloudStack - Nicolas Vazquez - ShapeBlue par ShapeBlue
Migrating VMware Infra to KVM Using CloudStack - Nicolas Vazquez - ShapeBlueMigrating VMware Infra to KVM Using CloudStack - Nicolas Vazquez - ShapeBlue
Migrating VMware Infra to KVM Using CloudStack - Nicolas Vazquez - ShapeBlue
ShapeBlue218 vues

Regulating Generative AI - LLMOps pipelines with Transparency

  • 1. REGULATING GENERATIVE AI — LLMOPS PIPELINES WITH TRANSPARENCY DEBMALYA BISWAS, WIPRO AI
  • 2. AGENDA  Enterprise AI  Ethical / Responsible AI  Explainability  Fairness & Bias  LLMOps Architecture Patterns  Generative AI - Responsible Design Principles
  • 3. ENTERPRISE AI • Enterprise AI use- cases are pervasive
  • 4. RESPONSIBLE AI “Ethical AI, also known as responsible AI, is the practice of using AI with good intention to empower employees and businesses, and fairly impact customers and society. Ethical AI enables companies to engender trust and scale AI with confidence.” [1] Failing to operationalize Ethical AI can not only expose enterprises to reputational, regulatory, and legal risks; but also lead to wasted resources, inefficiencies in product development, and even an inability to use data to train AI models. [2] [1] R. Porter. Beyond the promise: implementing Ethical AI, 2020 (link) [2] R. Blackman. A Practical Guide to Building Ethical AI, 2020 (link)
  • 5. REGULATIONS  Good news: is that there has been a recent trend towards ensuring that AI applications are responsibly trained and deployed, in line with the enterprise strategy and policies.  Bad news: Efforts have been complicated by different governmental organizations and regulatory bodies releasing their own guidelines and policies; with little to no standardization on the definition of terms.  For example, the EU AI Act mandates a different set of dos & don’ts depending on the ‘risk level’ of an AI application. However, quantifying the risk level of an AI application is easier said than done as it basically requires you to classify how the capabilities of a non-deterministic system will impact users and systems who might interact with it in the future.
  • 6. ETHICAL AI PRINCIPLES  Explainability  Bias & Fairness  Accountability  Reproducibility  Data Privacy *D. Biswas. Ethical AI: its implications for Enterprise AI Use- cases and Governance. Towards Data Science (link) *D. Biswas. Privacy Preserving Chatbot Conversations. 3rd IEEE AIKE 2020: 179-182
  • 7. EXPLAINABLE AI  Explainable AI is an umbrella term for a range of tools, algorithms and methods; which accompany AI model predictions with explanations.  Explainability of AI models ranks high among the list of ‘non-functional’ AI features to be considered by enterprises.  For example, this implies having to explain why an ML model profiled a user to be in a specific segment — which led him/her to receiving an advertisement. (Labeled) Data Train ML Model Predictions Explanation Model Explainable Predictions
  • 8. EXPLAINABLE AI FRAMEWORKS - LIME  Local Interpretable Model-Agnostic Explanations (LIME*) provides easy to understand explanations of a prediction by training an explainability model based on samples around a prediction.  The approximate nature of the explainability model might limit its usage for compliance needs. *M. T. Ribeiro, S. Singh, C. Guestrin. “Why Should I Trust You?” Explaining the Predictions of Any Classifier, 2016 (link) LIME output showing the important features, positively and negatively impacting the model’s prediction.
  • 9. EXPLAINABLE AI - FEASIBILITY  Machine (Deep) Learning algorithms vary in the level of accuracy and explainability that they can provide- the two are often inversely proportional.  Explainability starts becoming more difficult as as we move to Random Forests, which are basically an ensemble of Decision Trees. At the end of the spectrum are Neural Networks (Deep Learning), which have shown human-level accuracy. Explainability Accuracy Logistic Regression DecisionTrees Random Forest (Ensemble of DecisionTrees) Deep Learning (Neural Networks)
  • 10. EXPLAINABLE AI - ABSTRACTION “important thing is to explain the right thing to the right person in the right way at the right time”* Singapore AI Governance framework: “technical explainability may not always be enlightening, esp. to the man in the street… providing an individual with counterfactuals (such as “you would have been approved if your average debt was 15% lower” or “these are users with similar profiles to yours that received a different decision”) can be a powerful type of explanation” *N. Xie, et. al. Explainable Deep Learning: A Field Guide for the Uninitiated, 2020 (link) AI Developer Goal:ensure/improve performance Regulatory Bodies Goal:Ensure compliance with legislation, protect interests of constituents End Users Goal:Understanding of decision,trust model output
  • 11. FAIRNESS & BIAS  Bias is a phenomenon that occurs when an algorithm produces results that are systemically prejudiced due to erroneous assumptions in the machine learning process*.  AI models should behave in all fairness towards everyone, without any bias. However, defining ‘fairness’ is easier said than done.  Does fairness mean, e.g., that the same proportion of male and female applicants get high risk assessment scores?  Or that the same level of risk result in the same score regardless of gender?  (Impossible to fulfill both) * SearchEnterprise AI. Machine Learning bias (AI bias) (link) Google Photo labeling pictures of a black Haitian-American programmer as “gorilla” “White Barack Obama” images (link) A computer program used for bail and sentencing decisions was labeled biased against blacks. (link)
  • 12. TYPES OF BIAS  Bias creeps into AI models, primarily due to the inherent bias already present in the training data. So the ‘data’ part of AI model development is key to addressing bias.  Historical Bias: arises due to historical inequality of human decisions captured in the training data  Representation Bias: arises due to training data that is not representative of the actual population  Ensure that training data is representative and uniformly distributed over the target population - with respect to the selected features. Source: H. Suresh, J. V. Guttag. A Framework for Understanding Unintended Consequences of Machine
  • 13. LLMOPS: MLOPS FOR LLMS *D. Biswas. MLOps for Compositional AI. NeurIPS Workshop on Challenges in Deploying and Monitoring Machine Learning Systems (DMML), 2022. *D. Biswas. Generative AI – LLMOps Architecture Patterns. Data Driven Investor, 2023 (link)  Black-box LLM APIs: This is the classic ChatGPT example, where we have black-box access to a LLM API/UI. Prompts are the primary interaction mechanism for such scenarios.  While Enterprise LLM Apps have the potential to be a multi-billion dollar marketplace and accelerate LLM adoption by providing an enterprise ready solution; the same caution needs to be exercised as you would do before using a 3rd party ML model — validate LLM/training data ownership, IP, liability clauses.
  • 14. LLMOPS: MLOPS FOR LLMS (2) *D. Biswas. Contextualizing Large Language Models (LLMs) with Enterprise Data. Data Driven Investor, 2023 (link)  LLMs are generic in nature. To realize the full potential of LLMs for Enterprises, they need to be contextualized with enterprise knowledge captured in terms of documents, wikis, business processes, etc.  This is achieved by fine- tuning a LLM with enterprise knowledge / embeddings to develop a context-specific LLM.
  • 15. GENERATIVE AI - RESPONSIBLE DESIGN PRINCIPLES We take inspiration from the “enterprise friendly” Microsoft, “developer friendly” Google and “user friendly” Apple — to enable this ‘transparent’ approach to Gen AI system design. • Guidelines for Human- AI Interaction by Microsoft • People + AI Guidebook by Google • Machine Learning: Human Interface Guidelines by Apple