Session take-aways: ChatGPT and generative AI

Take-aways from a workshop run by Nick Scott for SOS Children’s Villages, offering an overview of ChatGPT and Generative AI and how they can be a partner in fundraising growth and data storytelling.

  • AI background
    • Definitions
      • AI
      • Generative AI
      • Predictive AI
      • LLMs
    • Language tasks account for 62% of total working hours across industries. 40% of all working hours across industries could be impacted by large language models.
    • Largest likely challenges/bottlenecks for nonprofits in using AI:
      • Using AI to scale bad fundraising practices rather than fixing underlying issues
      • Limited nonprofit capacity for large datasets and AI talent
      • Lack of transparency around how commercial AI platforms use data
      • Tendency to promote large, well-known nonprofits over smaller organizations
      • Ethical concerns around privacy, accountability, and other issues
      • Nonprofit skepticism and lack of knowledge about appropriate uses for AI
      • Lack of comprehensive outcomes data needed to train AI algorithms
  • ChatGPT Prompt tips
    • Analyse - identify strong and weak points in answers
    • Roles - tell ChatGPT what role to play
    • Context - provide detailed background
    • Iterate - ask ChatGPT to go step by step.
    • Feedback - ask for changes, improvements or clarifications
    • Specialise - use ChatGPT to support your areas of expertise
    • Polite - be engaging and friendly with ChatGPT
    • Examples and constraints - provide specific information and examples of outputs you want
    • Separate - Use different chats for different topics; name each chat
  • ChatGPT Limitations
    • Data privacy - the storing of inputted data
    • 2021 data cut-off
    • No internet access on core model
    • Eager to please and makes stuff up
    • Different GPT models - 3.5: fast and free; GPT4: more intelligent but subscription needed
    • You won’t get the same answer twice
    • Free version is often inaccessible
    • Limited memory (context window)
  • Generative AI Concerns
    • Misuse for misinformation / disinformation
    • Credibility and reputational risk from fake / incorrect content
    • Bias in outputted information
    • Copyright - source information and outputs
    • Environmental impact of training and using LLMs
‣
Example ChatGPT Roles and Use cases
  1. An assistant: Let it save time on things you can do but take time
    1. A copy reviewer:Analyse and identify the biases and assumptions inherent in the proposals you make or text you create
    2. A minute taker: Turn meeting transcripts into documents with minutes, including discussion summaries and action points.
    3. An editing translator: Edit and translate texts with complicated language from one language / style into another in a highly context-sensitive way
    4. An Excel formula wizard: Ask in plain English (or any supported language) to get an Excel formula
    5. A storyteller: Develop a narrative or alternative forms of explaining complicated concepts to teams and others.
  2. A team member: adding capacity in and among your team
    1. A Copywriting Assistant: Develop edits or variations of a piece of fundraising copy
    2. A content strategist: Develop outline content marketing plans and hone them
    3. A data analyst: Review data and identify hidden trends or ask questions about it
    4. A visual communicator: Find ways to communicate ideas in more visual ways, through images, diagrams, infographics etc.
    5. An events coordinator: Plan and support the delivery of all types of online and offline events
    6. An innovations lead: Identify new ideas and changes and detail plans to test these
    7. A researcher: Investigate information and knowledge
  3. A partner: Improve the quality of your work, even in your areas of expertise
    1. A language teacher: Support you in improving your mastery of a language
    2. A co-designer: Assisting throughout the design process to generate ideas and structure
    3. An opportunity engine: Generate options for an idea, concept or initiative
    4. A negotiation coach: Helping you improve your self-management or results
    5. A risk assessor: Identify inherent risks in a piece of text, strategy or proposal
    6. A persona / impersonator: Take the role of a specific person or role and argue from their viewpoint
    7. A personal tutor: Reflect on material or learn new skills.
‣
Example Generative AI Tools

Each tool has its own intricacies, strengths and weaknesses!

  • AI chatbot: ChatGPT, Bing AI, Bard, Claude
  • Copywriting and editing: Jasper.AI, Dataro AI Assist, Fundraising Copywriters Assistant, Jasper.ai, Google Docs Duet and Microsoft Word CoPilot
  • AI presentations: Gamma, Beautiful.ai, StoryD
  • Document chat: PDF.ai, Chatdoc, Humata
  • Meeting Transcription: Otter.AI, Fireflies.AI, Airgram
  • AI graphics: Microsoft Designer, Canva, Adobe Firefly
  • AI image creation or manipulation: DALL-E, Midjourney, Ideogram
  • Text to speech (TTS): Lovo.ai, Speechify, Murf, HeyGen
  • Automation: Zapier, Make.com, Bardeen
‣
Potential Next Steps
  1. Test some of the examples provided in this session and continue the conversation with ChatGPT to explore its capabilities and limitations further.
  2. Follow AI influencers in your field to stay up-to-date on the latest developments and applications of Generative AI.
  3. Create a personal or team safe space for experimentation with Generative AI tools, where you can explore their potential and limitations in a controlled environment.
  4. Participate in the drop-in session - details in your inbox after this session!

For more on AI and digital in non-profit organisations:

  1. My name is Nick Scott
  2. Please connect with me LinkedIn: linkedin.com/in/nicholasmscott/
  3. Or email me on: nick@nickscott.digital

LinkedInX