AI Landscape in Africa (with MEST Africa)

*AI landscape in Africa*

Tecmie Research Projects
Ethics and Regulation

---

Exploring the Machine Thinking Model
SAM.ADA methodology for architecting good machines

Transcript

1. AI LANDSCAPE IN AFRICA Overview on the adoption of AI across Africa and a deep dive into the Machine Thinking Model
2. How much is Africa’s biggest startup acquisition worth?
5. .regulation.
6. how can africa benefit from AI? African Union Convention on Cyber Security and Personal Data Protection Africa, as the continent with the youngest and fastest-growing population, has a unique opportunity and responsibility to harness AI for its development and transformation.
7. .research.
8. tecmie.com
9. can diagnostic reports be better? LAB.AI
10. can i find the right investors for my startup? VC.AI
11. who do I speak to when I feel overwhelmed? CHAT.KJV
12. who do I speak to when I feel overwhelmed? CHAT.KITAB
13. Numbers don’t lie Global religious organizations market size is expected at $420.12B by 2027 at a growth rate of 3.7%.
14. can machines reason and perform actions on their own accord? RETTIWT
15. can reasoning machines manage my corporate identity? WOOTIV
21. .machine. .thinking. A methodology used to architect computer systems that generate creative, iterative, and user-centered solutions based on their understanding (training) of human needs, values, and contexts.
22. 3 1 2 Leveraging a Customer Data Platform to drive product intelligence Previously In an age of AI. Your MOAT is your data Integrate a data ingress/egress strategy into your user workflows Integrate a customer data platform
23. ANDREW MIRACLE Andrew Miracle CEO TECMIE AI CONSULTANT PRODUCT/ BLOCKCHAIN/SOFTWARE ENTREPRENEUR
24. .basic definitions. What is AI, Machine Learning, Deep Learning, and LLMs (Large Language Models)?
25. .applied A.I. What does A.I look like when applied in the business context. from sales, marketing and operations?
26. .machine learning. What are the various machine learning algorithms and their different approaches in business
27. .large language. What are the various machine learning algorithms and their different approaches models
28. .ethics and business. What ethical considerations should I make when deploying AI in my business?
29. .tools and techniques. What tools can we use to identify AI opportunities and integration potentials in our business
30. .ADA & SAM. A Machine Thinking Model
31. ADA Understanding where AI fits
32. A. breakdown your business value chain, each components in the value chain from inbound logistics, operations, outbound logistics, marketing, sales and services analyze operations. inbound. marketing
33. D. define the business challenge, understand what aspect of your strategy is at risk or open to innovation. Assess the current cost or lost opportunities due to the problem. This helps build a strong justification for investing in an AI solution. define affinity map. cohort segmentation.
34. A. Ensure the chosen problem aligns with your organization's overall strategy and objectives. Show how AI can contribute to achieving those goals. align alignment with strategy
35. SAM Framing the AI opportunity
36. S. Research various AI techniques that could address the problem. Explain how each approach works and its potential benefits. scout
37. A. assess the technical feasibility of each solution, considering data availability, infrastructure requirements, and development effort. Project the potential ROI or cost savings for each option. assess
38. M. Based on your analysis, select the AI approach that delivers the best balance of value, feasibility, and alignment with your goals. Tell a story that explains the problem, the potential of AI, and the chosen solution in a clear and concise way. Tailor the message to resonate with your audience, be it technical decision-makers or executive leadership. model
39. .ethical. .considerations. Acknowledge potential challenges like data privacy, bias, and explain your mitigation strategies. Build trust and transparency to secure buy-in.
40. important tips Focus on business outcomes: Keep the focus on how AI will solve real-world problems and deliver tangible benefits. Involve stakeholders: Get buy-in from key decision-makers and potential users early on. Their involvement strengthens the case and ensures smooth implementation. Use data and evidence: Back up your claims with data, research, and case studies to build credibility and conviction. Be flexible and adaptable: Be prepared to adjust your plan based on feedback and learnings as you progress.
42. AI Vision Our vision is to transform our retail operations into a data-driven, customer-centric model by Q4 2024. By implementing an AI-powered inventory system and a customer recommendation engine, we aim to increase sales by 25% and reduce inventory costs by 15%. The plan involves training our staff in data analytics, integrating AI tools by Q2 2024, and launching a pilot program. We will regularly review progress against our KPIs, adapting our strategy based on customer feedback and system performance. Stakeholders, including department heads and IT staff, will be involved throughout, ensuring a smooth transition to a more efficient, AI-enhanced operation."
44. .components of. .a good machine.
45. a good machine It should understand the meaning of text, facial expressions, user interactions and any available data within the system to build a nuaced understanding of the user’s need based on objectives and product strategy understands nlp. vision. tts. vectors
46. a good machine The process of desigining the system is not a one-shot process. A continuous process of training, reinforcement learning or other feedback mechanisms can be used to continuously improve their decision-making for these systems can improve RLHF. LoRA. Transformers. MoE
47. a good machine A good machine doesn’t just mimic existing patterns. They are able to ggenerate novel solutions that can address the user’s need in un- expected ways. is creative LLMs. Lateral Thinking. Stable Diffusion. GPT
48. a good machine A good machine is aware of the broader context in which they operate, including the potential biases and ethical implications of their decisions. This involves explicitly designing for inclusivity, fairness, and transparency, along with building safeguards to prevent discriminatory or harmful outcomes. has context RAG. Cosine. Moderation. e/acc
49. What emerging technologies might enable us to build truly user-centered AI systems? Discussion
50. .prompt. .engineering. A quick primer on instructing machines better
51. elements of a prompt ✍🏼 A prompt contains a specific task or instruction you want the model to perform instruction Large language models in the workplace
52. elements of a prompt ✍🏼✍🏽 A prompt contains external information or additional context that can steer the model to better responses context Large language models in the workplace
53. elements of a prompt ✍🏼✍🏽✍🏾 A prompt contains the input or question that we are interested to find a response for input data Large language models in the workplace
54. elements of a prompt ✍🏼✍🏽✍🏽✍🏾 A prompt contains the the type or format of the output we expect from our AI model output indicator Large language models in the workplace
55. A technique for modeling AI systems that can remember what they need to perform an action. retrieval augmentation Large language models in the workplace memory.
56. let’s prompt an AI to help us form teams for an idea case study Large language models in the workplace
57. LLMs in the workplace Prompting Techniques
58. LLMs in the workplace Evaluation & Impact of Prompts Template stickiness refers to how frequently the model answers in the desired format.
60. Bonus
61. Lawler Framework DECEMBER 2023 Smart Systems
62. Lawler Model Framework
63. Needs Assessment: Identifying the real-world problem the AI model seeks to solve and assessing the need for an AI-based solution. Stakeholder Mapping: Identifying all individuals or organizations who would be impacted by the product and including them in the consultation and feedback loops. Understanding the Problem Space
64. Sourcing the data and deciding where it will come from. preprocessing: cleaning and organising the data for training or retrieval purposes. Data consideration
65. Algorithm Selection: Deciding on the algorithmic architecture (e.g., deep learning, decision trees, etc.) most appropriate for solving the problem. Model Training: Actually building and training the AI model. Algorithm and Technical preference
66. Transparency: Creating a system that is understandable and can be interpreted by both technical and non- technical stakeholders. Bias Mitigation: Putting into place mechanisms to detect and mitigate any biases in data or algorithms. Ethics
67. Prototyping: Developing a minimum viable product (MVP) for initial testing. User Testing: Gathering user feedback and making necessary adjustments. Launch Strategy: Deciding how the product will be rolled out, whether as a phased launch or full-scale deployment. Ongoing Evaluation: Constantly monitoring for any issues that could arise post-launch, both from a technical and an ethical standpoint. Prototype Deploy & Monitor
69. Q&A
70. AI apps for Videos
71. AI apps for Voice
72. AI apps for Presentation
73. AI apps for Operations
74. AI apps for Marketing