From Efficiency to Innovation: Transforming Business Value through Gen AI

The world of Al is undergoing a metamorphosis. Traditional Al, programmed for specific tasks like playing chess, is being eclipsed by the new era of learning Al. This new breed can adapt, analyze data, and even create content. This shift is a game-changer for enterprises. Repetitive tasks can be automated, vast datasets can be analyzed for insights, and even entirely new products can be Al-powered. But the workforce needs to adapt too. Collaboration with Al tools will be key, requiring new skillsets like critical thinking and problem-solving. Generative Al, with its ability to craft images, music, and even code, holds immense promise. However, current offerings are in their infancy they can be impressive but prone to stumbles and biases.

The future of business is a partnership with Al. Businesses must carefully assess current tools and invest in human-Al collaboration and continuous learning. This will be the key to navigating the exciting, but uncertain path ahead. Eventually, we must not lose sight of the true purpose of an enterprise to provide value to the consumers, in order to improve their lives, and to do so responsibly, and in a sustainable way that provides acceptable returns to stakeholders.

1. From Efficiency to Innovation
Transforming Business Value through Gen AI
Sameer Verma, Ph.D.
Professor and Chair, Information Systems
Lam Family College of Business
San Francisco State University
https://faculty.sfsu.edu/~sverma
[email protected]
May 30, 2024

2. Introduction
Let’s ask Gemini!

3. Gen AI

4. Gen AI

5. Gen AI

6. Gen AI

7. Gen AI

8. Gen AI

9. Gen AI

10. Gen AI
Not too bad!
Summarized well.
Missed a few things.
Took some convincing, but satisfied my ego
??????

11. me

12. Teaching
●Information Systems and Technology
–Networks
–Cybersecurity
–Cloud Computing
–Content Management
–Business Analytics
–Strategy, Policy, Governance

13. Research
●Intelligent Hypermedia Systems
●Decision Support Systems
●Online Learning
●Free and Open Source
●Content Management
●Edge and Fog Computing

14. Service
●Hackathons
–Civic Tech, Civic Engagement
–http://hackathon.sfsu.edu
●One Laptop per Child
–Jamaica, India, Madagascar, Tuva…
–http://olpcsf.org
●WiRED International
–Peru, Kenya, Armenia, India...
–http://wiredinternational.org

15. Oodles of free time...
https://www.instagram.com/v3rmaji/

16. Accountability

17. Deconstruct...
●Efficiency
●Innovation
●Business Value
●Transforming
●Gen AI
Disruptive!

18. I like ice cream
Construct

19. I like ice cream
I
like
want
need
ice cream
bread pudding
rum cake
gizzada
coconut drops
Think auto-complete on
your phone. How does it
know what comes next?

20. Tensor
Introduction to Artificial Neural Networks and Deep Learning: A Practical Guide with Applications in Python by S. Raschka http://leanpub.com/ann-and-deeplearning

21. Tensor
Data Types - Scalars, Vectors, Matrices and Tensors https://www.stephanosterburg.com/math_data_types

22. Tensor math
●Matrix manipulation for
tensors
●Use CPU
●Use GPU
●Use ASICs
–Tensor Processing Unit
(Google)
–Apple Neural Engine
–Groq Language Processing
Unit
Google TPU 3

23. Neural Network
●A model
●Inspired by the structure
and function of biological
neural networks in our
brains.
●Has neurons, synapses,
signals and weights.
●Adjusting parameters
determines “learning”.
●Train and Use

24. An experiment
●nanoGPT
–Runs on GPT 2
●Train on “Tiny Shakespeare”.
–40000 lines of text from Shakespeare’s works.
●Iterations
–1000, 10000, 100000
●Can it recreate Shakespeare-like text?

25. 1000 iterations
Sor ound doull ethat peet remives, yor ston?
FREO:
And to not to my imther, my heis theath, Mang
An hoow wearlle of sit is if thave preans a to a the ind,
Thatt we lor for you in on doncee.

26. 10000 iterations
JULIET:
I cannot me stood me, beg them to the
Of his mind cannot presence to the world
Of my King learning so unlumberance,
And great no tender you migh

27. 100000 iterations
FRIAR LAURENCE:
A man, sir; yet so, must not fit Bolingbroke;
They, though the jumpeth of my rapiring.
Suffording, I promised
The messenger, to emboste upon thee; and thee,
That from you loves his brother lie,
And seem so mind her business, which thou art,
Those of will receive where made to free hall profer.

28. Infinite Monkey Theorem
https://en.wikipedia.org/wiki/Infinite_monkey_theorem

29. Stochastic Parrots
●Learn patterns
●Reward-driven
–“Polly wanna
cracker”
–Polly does not
understand what this
means.
Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. On
the Dangers of Stochastic Parrots: Can Language Models Be Too Big? . In Proceedings of the
??????
2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT '21). Association for
Computing Machinery, New York, NY, USA, 610–623. https://doi.org/10.1145/3442188.3445922

30. Large Language Models
●Learn statistical relationships from text documents.
●Computationally intensive training process.
●Post-training, generate the next word.
–Generative AI.
–Generation is not as intensive.
●Acquire patterns about syntax, semantics and
ontology over time.
●Very cool, but can it reason?

31. Heuristics vs Reasoning
●Human heuristics
–Attribute substitution
–Cognitive load reduction
–Quick and frugal
●Human reasoning
–Driven by logic
–How?
–Why?
●Cost of reason vs pattern
Verma, R. K. (1986). Offshore seismic exploration. Data acquisition, processing, and interpretation.
National Geophysical Research Institute, Hyderabad.

32. LLM Landscape
Large Language Models: A Survey https://doi.org/10.48550/arXiv.2402.06196

33. LLM families
Large Language Models: A Survey https://doi.org/10.48550/arXiv.2402.06196

34. Timeline

35. Hallucinations
●Could it all be a set of hallucinations?
–Disagreements = hallucinations
●Errors get compounded quickly
–Generated in the training data
–Insufficient training
–Model overfitting
–Vector encoding
●Trust, but verify!

36. An example

37. Explainable AI
●How do we explain the “why”?
–Weights in the neural network.
–Retrofitting some sense of reason.
●Out of band solutions
●Validation of results

38. Business Value
●Collaboration & Partnership
●Responsible and sustainable
●Returns to stakeholders
●Provide value

39. Supply and Demand

40. Value chain

41. Stakeholder perspective
RETURNS

42. Strategy
Business Organization IT
Mission
Vision
Values
Org Structure
Management Control
Work Culture
Technology Innovation
Portfolio Evolution
Total Cost of Ownership
Driver Innovation

43. Governance
Business Organization IT
Mission
Vision
Values
Org Structure
Management Control
Work Culture
Technology Innovation
Portfolio Evolution
Total Cost of Ownership
Managing Expectations
Through Decision Rights

44. Policy
●US Executive Order (E.O.) 14410, Safe, Secure, and
Trustworthy Development and Use of Artificial
Intelligence.
●EU AI Act.
●China: "A Next Generation Artificial Intelligence
Development Plan" (State Council Document No. 35).
●G7: Eleven guiding principles for the design,
production and implementation of advanced artificial
intelligence systems.

45. NIST AI Risk Management
https://www.nist.gov/itl/ai-risk-management-framework

46. Challenges and Risks
●Scalability and integration
●Business model
disruption
●Workforce transformation
●Technology dependency
●Algorithmic bias
●Hallucinations and
misinformation
●Ethical and moral dilemmas
●Intellectual property
concerns
●Privacy and cybersecurity
●Regulation and compliance
●Technological obsolescence
●Digital divide

47. Scale. Scope. Context.
Modes of Engagement

48. Simple Insights
http://data.gov.jm/

49. Copilot in Excel
●Use the Copilot ribbon.
–"Create a bar graph showing the sales growth
between Q2 and Q3."
–"Add a new column showing the percentage
difference between column A and column C."

50. Open LLMs
●Meta Llama
●Mistral AI
●Google Gemma
●Apple OpenELM
●Microsoft Phi
●Databricks MPT
Awesome-LLM https://github.com/Hannibal046/Awesome-LLM
Run locally!

51. Enterprise Systems
Bring Your Own LLMs
BYOLLM

52. Salesforce Einstein

53. Business Models
Kanbach, D.K., Heiduk, L., Blueher, G. et al. The GenAI is out of the bottle: generative artificial intelligence from a
business model innovation perspective. Rev Manag Sci 18, 1189–1220 (2024).
https://doi.org/10.1007/s11846-023-00696-z

54. Why efficiency?
What comes next...
Age of Gen AI?

55. Nascent
●Still very early
–Very disruptive, but similar to other disruptive technologies.
●Focus on business value
–Role of Mission and Vision remain unchanged.
●Re-examine governance
–Losing control to external AI entities.
●“AI Positive” approach
–Leverage generously.
●Human-centric
–The machine is here for us. We are not here for the machine.

56. Accountability

57. Closing

58. Articles
●Chowdhury, S., Budhwar, P., & Wood, G. (2024). Generative Artificial Intelligence in Business:
Towards a Strategic Human Resource Management Framework. British Journal of Management.
●Kanbach, D. K., Heiduk, L., Blueher, G., Schreiter, M., & Lahmann, A. (2024). The GenAI is out of the
bottle: generative artificial intelligence from a business model innovation perspective. Review of
Managerial Science, 18(4), 1189-1220.
●Prasad Agrawal, K. (2023). Towards adoption of generative AI in organizational settings. Journal of
Computer Information Systems, 1-16.
●Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021, March). On the dangers of
stochastic parrots: Can language models be too big? . In Proceedings of the 2021 ACM conference
??????
on fairness, accountability, and transparency (pp. 610-623).
●Li, Z. (2023). The dark side of ChatGPT: legal and ethical challenges from stochastic parrots and
hallucination. arXiv preprint arXiv:2304.14347.
●Arkoudas, K. (2023). ChatGPT is no Stochastic Parrot. But it also Claims that 1 is Greater than 1.
Philosophy & Technology, 36(3), 54.
●Dotan, R., Blili-Hamelin, B., Madhavan, R., Matthews, J., & Scarpino, J. (2024). Evolving AI Risk
Management: A Maturity Model based on the NIST AI Risk Management Framework. arXiv preprint
arXiv:2401.15229.

59. The machine
On the opening slide are a grid of Artificial Neural Networks (ANN).
This is a TensorFlow implementation that demonstrates the workings
of ANN.
You can try it yourself and run a neural network in your browser at
http://playground.tensorflow.org