Does your organization allow ChatGPT at work? The answer might depend on where you work. Many organizations do not allow ChatGPT at work. The truth is that for the organizations, ChatGPT is a fear of missing out and a fear of messing up. But, just like any other past new technologies such as Cloud computing and social media, the organizations eventually integrate ChatGPT or other Large Language Model (LLM). This paper is for those especially Biometrics who want to initiate ChatGPT integration at work.
This paper presents how Biometric department can lead the integration of LLM, focusing on the exemplary model ChatGPT, across an entire enterprise, even in situations where the organization restricts or prohibits ChatGPT usage at work.
The roadmap outlines key stages, starting with an introduction to LLM and ChatGPT, followed by potential risks and concerns and the benefits and diverse use cases. The roadmap will emphasize how Biometrics function leads the building of a cross-functional team to initiate ChatGPT integration and build the policy and guidelines. Then, the roadmap discusses the crucial aspect of training, emphasizing user education and engagement based on company polices. The roadmap finishes with a Proof of Concept (PoC) to validate and evaluate the ChatGPT’s applicability to organizational needs and its compliance to company policies.
This paper can serve as a valuable resource navigating the implementation journey of ChatGPT, providing insights and strategies for successful integration, even within the confines of organizational limitations on ChatGPT usage.
Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...
A fear of missing out and a fear of messing up : A Strategic Roadmap for ChatGPT Integration at Work
1. Strategic Gen AI (ChatGPT)
Integration at Enterprise-level
Kevin Lee
2. Disclaimer
The views and opinions presented here represent those of
the speaker and should not be considered to represent
any companies or organizations.
3. Agenda
➢ Gen AI (ChatGPT) Implementation
Roadmap for the whole organization
➢ Gen AI & ChatGPT Introduction
➢ Risks and Concerns
➢ Benefits and Use Cases
➢ Cross Functional Team
➢ Policy and Guidelines
➢ Training and Education
➢ PoC
➢ Evaluation
➢ Discussion
3
7. What is Gen AI?
Gen AI – a trained Machine Learning
model that generate new contents
with a simple prompt.
• Text (e.g., Large Language Models) :
Content Writing, Chatbots, Assistants,
Search
• Code : Code Generation, Data Set
Generation
• Image : Image Generation, Image Edit
• Audio : Voice Generation/Edit, Sound
creation, Audio Translation
• Video : Video Creation/Edit, Voice
Translation, Deepfake
7
12. What is ChatGPT?
• ChatGPT is an advanced LLM developed by
OpenAI.
• ChatGPT is trained on large corpus of text (~300B
tokens).
• Its main strength
• the ability to generate human-like responses in
various contexts.
• The ability to understand and generate text in a
wide range of domains.
• Its application
• Inquiry/prompt ( rather than search )
• Content Generation
• Customer support
• Interactive Storytelling
• Coding
• Image Analysis
• Art Development 12
13. ChatGPT Development
13
ChatGPT4
• March’2023
• 8K tokens
• 32k tokens (ChatGPT4-
32k)
• ~1,760B parameters
ChatGPT3.5
• April’2023
• 4K tokens (3,072 words)
• 16K tokens
(ChatGPT3.5-turbo-16k)
• 175B parameters
ChatGPT3
• June’2020
• 1,536 words
• 175B parameters
ChatGPT2
• Feb’2019
• 768 word
• 1.5B parameters
Reference Points
• Tweet : 55 words
• NY Times Article : 622 words
• Standard Novel : 90K words
• Kings James Bible : 783,137 words
Each prompt include the current tokens and previous tokens.
15. Reasons that ChatGPT is not being used
• A lack of understanding of
ChatGPT Use Cases / Benefits
• A lack of ChatGPT Usages(
Prompt Engineering)
• Concerns/Risks using ChatGPT
15
17. 17
Gen AI (ChatGPT) Implementation Roadmap
Risk and
Concerns
Benefits and
Use Cases
Cross
Functional
Team
Policy and
Guideline
Training /
Education
PoC
ROI
18. Concerns using AI (ChatGPT)
- Data Privacy & Security
- Bias
- Regulatory Compliance
- Ethical Consideration
- Black Box Feature
- Resource Requirements
- Lack of Expertise
- Cost & ROI
19. 19
Data Privacy
• Data Privacy – the right of a
person to have control over
how their personal
information is collected and
used. (e.g., clinical trial data)
• Since ML is built using data,
ML algorithms can contain
sensitive information even
though it is one of a Big Data.
• The growth of ML has
increased the possibility of
using sensitive data in ways
that may violate data
privacy. (e.g., ChatGPT)
20. 20
Bias of Machine Learning
• Phenomenon that occurs
when ML algorithms
produces results that are
prejudiced due to wrong
assumptions
• Examples :
• Amazon ML recruiting
engine show a bias on
women.
• ML was trained to
vet applicants using
10 years of resume
• Most came from
men, a reflection of
male dominance
across tech industry.
21. 21
Regulatory Compliance
• Regulatory Compliance – the
process of adhering to laws,
regulations, standards, and other
rules set by governments and
other regulatory bodies.
• For pre-market approval, AI-
based software needs to follow
regulatory guidelines (e.g., Good
Machine Learning for Medical
Device Development Guiding
Principles)
22. 22
Ethical Consideration
• Who will make decision? AI or
human?
• Can AI be responsible for their
decision?
• Is AI transparent?
23. Potential Benefits using
ChatGPT
- Higher Productivity / Efficiency
- Better Patient Engagement
- Patient’s Recruitment
- Enhancement in Image
- Market Research & Insights
24. ChatGPT Users vs Non-users
• Results of consultants
using ChatGPT
• finished 12.2% more
tasks on average
• completed tasks
25.1% more quickly
• produced 40%
higher quality results
24
25. Bar Exam Score Performance :
ChatGPT3.5 vs ChatGPT4
25
10% vs 90%
31. 31
Employee
Training
• Employees who will
be using and
managing
ChatGPT
• Security best
practices
• Data Privacy
• Data Handling
Procedures with
ChatGPT
• Ethics and
Compliance