AI Black Box - Cracking the Code for Startup Success with AI

Unlock the power of AI for a successful startup with "The AI Black Box: Cracking the Code for Startup Success with AI". This presentation guides you through the critical steps of leveraging AI effectively, starting with identifying the customer pain point your solution addresses. We'll then explore the diverse landscape of large language models (LLMs), helping you choose the perfect fit for your needs based on training data and capabilities. But the journey doesn't end there. We'll also discuss the importance of continuous monitoring to address model drift and ensure ongoing performance. Finally, the presentation will emphasize the foundational role of responsible AI principles in the design process. We’ll delve into strategies for mitigating bias, enhancing explainability, ensuring fairness, and protecting privacy. By embedding these principles from the outset, startups can create ethical, effective AI solutions.
Join us for an insightful journey into the intricacies of AI implementation in startups. Whether you are at the beginning of your AI journey or looking to refine your existing approach, this presentation will provide you with the tools and knowledge to crack the code for AI-driven success.

1. The AI Black Box
Cracking the Code for Startup Success with AI
1871 AI Innovation Lab -
May 28th, 2024
1
Credit: Microsoft Copilot

2. ●Academic Background
○Ph.D. in Computing Science from Aberdeen University (Scotland)
■Leveraged Semantic Technologies to facilitate inter-department collaboration
■Introduced an approach to map entities across semantic schemas (Thesis)
■Contributed to the development of W3C standards, such as SKOS

●Work Experience
○Contribute to research, and events focused on ethical, responsible, and inclusive AI development and deployment as a
Global Fellow with AI 2030
○Guide companies in developing and refining their AI-driven solutions, focusing on Responsible AI and go-to-market
strategies
○Led a cross-functional team demonstrating the potential of Generative AI on process optimization and product innovation
○Applied jobs-to-be-done framework to develop solutions that tangibly demonstrate the values of AI in transforming existing
customer experiences across different industries
○Championed a "lean innovation" mindset, encouraging team members to experiment, learn from failures, and rapidly
iterate, driving a culture of continuous improvement and innovation
○Awarded multiple Innovation Awards for contribution to the delivery of novel AI solutions
Who am I?
The AI Black Box:
Cracking the Code for Startup Success with AI

3. Agenda
●Introduction
●Finding the Needle in the Haystack
●The Right Fit for the Right Job
●From Launch to Longevity
●Building AI Solutions with Responsibility
●Key Takeaways
3
The AI Black Box:
Cracking the Code for Startup Success with AI

4. What is Generative AI?
4
The AI Black Box:
Cracking the Code for Startup Success with AI
Generative AI is a branch of AI that focuses
on creating entirely new content. This
content can come in many forms, including
text, images, audio, and even videos.
Credit: Microsoft Copilot

5. How Generative AI is Transforming Industries?
5
The AI Black Box:
Cracking the Code for Startup Success with AI
Source: The economic potential of generative AI: The next productivity frontier
75% of Generative AI value focuses on:
-Customer Operations
-Marketing & Sales
-Software Engineering
-Research & Development

6. Agenda
●Introduction
●Finding the Needle in the Haystack
●The Right Fit for the Right Job
●From Launch to Longevity
●Building AI Solutions with Responsibility
●Key Takeaways
6
The AI Black Box:
Cracking the Code for Startup Success with AI

7. Leveraging Diverse Perspectives
7
The AI Black Box:
Cracking the Code for Startup Success with AI
Source: The Power of "Jobs-to-be-Done" in Crafting Winning AI Solutions
By assembling multi-disciplinary teams, we harness the
unique perspectives and expertise of individuals from various
backgrounds. This fosters a 360° view of the problem, sparks
innovation through the cross-pollination of ideas.

8. Uncover Customer Pain Points
8
The AI Black Box:
Cracking the Code for Startup Success with AI
Credit: Jobs To Be Done Framework
The jobs-to-be-done framework is an approach to
product development that focuses on understanding
the underlying motivations and goals that drive
customers to use a product or service.
Source: The Power of "Jobs-to-be-Done" in Crafting Winning AI Solutions

9. Validate Early, Iterate Often
9
The AI Black Box:
Cracking the Code for Startup Success with AI
The Lean startup method is a
customer-centric approach to
developing products. It emphasizes
gathering early feedback through
rapid iterations to validate
product ideas and minimize waste.
Source: Lean Startup: How the lean method works

10. Find Your Ideal Customer Partners
10
The AI Black Box:
Cracking the Code for Startup Success with AI
Source: The Power of "Jobs-to-be-Done" in Crafting Winning AI Solutions
Credit: Diffusion of Innovation
The diffusion of innovation theory explains how new
ideas, technologies, or products spread and gain
acceptance within a social system or market over time.

11. Agenda
●Introduction
●Finding the Needle in the Haystack
●The Right Fit for the Right Job
●From Launch to Longevity
●Building AI Solutions with Responsibility
●Key Takeaways
11
The AI Black Box:
Cracking the Code for Startup Success with AI

12. Large Language Models (LLMs) Wild West
12
The AI Black Box:
Cracking the Code for Startup Success with AI
Source: Large Language Model Highlights (Apr/2024)

13. Evaluating LLMs for Maximum Impact
13
The AI Black Box:
Cracking the Code for Startup Success with AI
Source: Introducing DBRX: A New State-of-the-Art Open LLM
Source: Trustbit LLM Benchmarks for April 2024
Source: Artificial Analysis, Comparison of Models: Quality, Performance &
Price Analysis

14. Don't Get Blinded by the Shine
14
●Potential of overfitting: Training data
may include knowledge of benchmark
●Limited Task Diversity: Benchmarks
cover limited set of tasks
●Rapid model evolution: Benchmarks
don’t evolve as same speed as models
●Narrow focus: Don’t assess specific
knowledge required for your industry
Credit: Microsoft Copilot
The AI Black Box:
Cracking the Code for Startup Success with AI

15. TechType Rocket: A Practical Example
15
The AI Black Box:
Cracking the Code for Startup Success with AI
TechType Rocket on GitHub - Tool to help authors with
TechType Rocket© helps authors with
overcoming the “blank page”
syndrome. It also includes guardrails to
ensure that the input and output does not
contain toxic content.

16. TechType Rocket: A Practical Example
16
The AI Black Box:
Cracking the Code for Startup Success with AI
Source: It' Official Gemini Advanced Admits It Can't Count - Reddit

17. Agenda
●Introduction
●Finding the Needle in the Haystack
●The Right Fit for the Right Job
●From Launch to Longevity
●Building AI Solutions with Responsibility
●Key Takeaways
17
The AI Black Box:
Cracking the Code for Startup Success with AI

18. Adapting to a Changing World
18
The AI Black Box:
Cracking the Code for Startup Success with AI
Keeping up with rapidly evolving generative
AI models is crucial to mitigate risks from
opaque training data, leverage improved
capabilities, ensure regulatory compliance,
and meet rising user expectations.
Credit: Microsoft Copilot

19. Navigating Model Drift
19
Rigorous monitoring is crucial to detect
issues like performance drift caused by
different training sets used across versions.
The AI Black Box:
Cracking the Code for Startup Success with AI
Credit: Microsoft Copilot

20. Agenda
●Introduction
●Finding the Needle in the Haystack
●The Right Fit for the Right Job
●From Launch to Longevity
●Building AI Solutions with Responsibility
●Key Takeaways
20
The AI Black Box:
Cracking the Code for Startup Success with AI

21. AI with a Conscience
21
The AI Black Box:
Cracking the Code for Startup Success with AI
Source: Don't Get Left Behind: Why Responsible AI is the Key to Customer Success
Responsible AI refers to the practice of
developing and deploying AI systems in an
ethical, transparent, and accountable
manner, while mitigating potential risks
and negative impacts.
Credit: Microsoft Copilot

22. A Framework for Sustainable Growth
The main pillars of Responsible AI are:
●Fairness: Treat all customers fairly, without
discrimination or unfair bias
●Transparency & Explainability: Be opened
about the algorithms, data used
●Accountability & Reliability: Maintain
detailed audit trails of interactions
●Privacy & Security: Safeguard data to
prevent data breaches and misuse

22
The AI Black Box:
Cracking the Code for Startup Success with AI
Source: Don't Get Left Behind: Why Responsible AI is the Key to Customer Success

23. Putting Responsible AI into Action (Regulatory)
In force:
●EU AI Act prohibits certain applications like social
scoring, and defines requirements for high-risk
applications
●TN ELVIS Act protects artists from having their
voices and pictures used in AI.
●CA Executive Order N-12-23 outlines a plan for the
responsible development and use of AI through
conducting risk assessments

In consideration:
●Stop Spying Bosses Act proposes requirements for
employers with respect to the collection and
disclosure of certain worker data.
●Generative AI Copyright Disclosure Act of 2024
proposes the disclosure of copyrighted material
used to train models

23
The AI Black Box:
Cracking the Code for Startup Success with AI
Credit: Microsoft Copilot
Credit: Microsoft Copilot
Credit: Microsoft Copilot

24. Putting Responsible AI into Action (Voluntary)
●NIST AI Risk Management Framework
provides guidelines to identify, assess,
and manage risks
●OECD AI Principles promote the
development and use of trustworthy AI
systems that respect human rights and
democratic values
●Experts from 27 nations, including the UK
government, have published an interim
report on safety of advanced AI to
minimize severe risks.
●AI 2030 promotes the development and
adoption of Responsible AI systems that
benefit humanity while minimizing
potential negative impacts

24
The AI Black Box:
Cracking the Code for Startup Success with AI
Credit: Microsoft Copilot

25. Agenda
●Introduction
●Finding the Needle in the Haystack
●The Right Fit for the Right Job
●From Launch to Longevity
●Building AI Solutions with Responsibility
●Key Takeaways
25
The AI Black Box:
Cracking the Code for Startup Success with AI

26. The Journey Continues
1.Understand the specific problem your AI
solution addresses.
2.Evaluate various LLMs to determine
suitability for your specific problem.
3.Implement continuous monitoring to ensure
your solution remains effective and relevant.
4.Embed responsible AI principles from the start
to build trust and sustainability.
26

27. How Founders Help Driving Responsible AI
●Ensure buy-in and commitment to responsible AI
principles across the organization
●Conduct regular assessments to identify
potential risks and define ways to mitigate them
●Foster a culture of open discussions around AI
ethics
●Stay informed and continually educate
yourselves about best practices and regulations
●Lead by example and drive positive change in
your industry
●Prioritize transparency by proactively
communicating your responsible AI approach to
customers, partners, and the public
27
The AI Black Box:
Cracking the Code for Startup Success with AI
Credit: Microsoft Copilot

28. Resources
●Courses
○IBM Generative AI Fundamentals Specialization on Coursera
○Microsoft Azure AI Fundamentals: Generative AI
○Google Generative AI for Developers Learning Path
●News & Trends
○TLDR AI
○Gradient Flow: Unlocking Data & AI
●Podcasts
○AI for Humans
○What’s the BUZZ? — AI in Business
○In AI We Trust?
●Communities
○AI Tinkerers
○AI Makerspace
○AICamp
○MLOps Community
28
The AI Black Box:
Cracking the Code for Startup Success with AI

29. 29
Credit: Mario ? Block

30. Let’s Connect!
●Email: [email protected]
●LinkedIn: https://www.linkedin.com/in/quentinreul/
●X (formerly Twitter): https://x.com/rqle_ai
●Personal GitHub: https://github.com/qhreul
●RQle.ai GitHub: https://github.com/RQledotai
30

31. Disclaimer
This presentation includes content that was generated with the assistance of
Generative AI. While AI technology has been used to aid in the creation of this
material, all content has been curated, reviewed, and finalized by the author.
The author assumes full responsibility for the accuracy, originality, and
integrity of the information presented.
31
The AI Black Box:
Cracking the Code for Startup Success with AI