The document discusses the rise of generative artificial intelligence and its chatbot ChatGPT, outlining its potential uses for knowledge work including summarization, translation, and drafting as well as implications for the future of knowledge management and society. It suggests that development organizations prepare for building their own internal AI systems and discuss strategically how to handle societal impacts of information overload and job disruptions.
Simulation-based Testing of Unmanned Aerial Vehicles with Aerialist
How generative AI will transform knowledge work
1. Generative Artificial Intelligence
3/14/2023
Johannes Schunter
Head of Knowledge Management
Diakonie Deutschland | Brot für die Welt | Diakonie Katastrophenhilfe
Implications for knowledge work in
development using the example of ChatGPT
KM4Dev Knowledge Café
3. • 2017 - Google releases the Transformer model
• AI model for text processing via machine learning (Large Language Models)
• Initially for translation services & speech recognition, now also for text generation
• OpenAI's GPT model (2018 GPT-2; 2020 GPT-3)
• In November 2022, ChatGPT is released for free, a chatbot based on GPT-3.5
(trained on 100 billion words)
• Within 2 months, it gains 100 million users
• TikTok took 9 months
• Instagram took 28 months
• Microsoft follows with Bing Chat (based on GPT), and Google, Meta, and Amazon
are about to release their own language models.
Background
4. Example: Research on any topic that was ever written about
1. Entry point for research
Prompt: What is ‘shrinking space’ and why is the concept important for development?
2. Search for examples
Prompt: Give me five examples of shrinking space from concrete countries. Display
them in a table with the columns country, situation, effects and impact for development
Research
5. Example: Quickly understanding a UN resolution (https://documents-dds-
ny.un.org/doc/UNDOC/GEN/N21/416/87/PDF/N2141687.pdf?OpenElement)
1. Summaries of large texts
Prompt: Summarize the following in bullet points and simple language:
„[COPIED TEXT]“
2. „Translation“ of difficult texts (legal/medical/scientific/bureaucratic)
into easy language
3. Follow up on specific parts or aspects of the text
Summaries of texts
6. Example: Message to partners on International Women‘s Day
Prompt: Write a message on International Women‘s Day 2023 to our partners in a colleagial and
warm tone. Congratulate them on their efforts in advancing women‘s rights, thank them for being a
valuable partner in our projects. Point out the special burden for women during the past COVID-19
pandemic (mention some reasons) and do a motivating call to action to collaborate further in
empowering women. Close with a hopeful quote from an international female leader from the global
South (quote must be real).
Assistance in drafting texts (Part 1)
7. General format: „Write [DOCUMENT TYPE] to/for
[TARGET AUDIENCE] in [TONE/STYLE] language on [TOPIC] to achieve [GOAL].
Example: Complaint letter to service provider
Prompt: Write an Email to Mr. Brown, the Head of Sales at the printing company Pronto in a friendly
but firm tone, regarding the botched delivery of flyer prints for the anniversary of our organisation (the
margins have been cut the in the wrong way), in order to get a full refund.
Suitable for standard communications: Protocols, press releases,
job advertisments, announcements, event invites, questionnaires,
bureaucratic communication, etc.
Assistance in drafting texts (Part 2)
8. 1. Shorten texts
Prompt: Make the text a bit shorter without loosing the key points.
2. Tailor to different audiences
Prompt: Rewrite the text so it is speaking to [TARGET AUDIENCE]
3. Adjust the tone
Prompt: Make the text more diplomatic / joyful / serious / motivating / practical / demanding / …
Prompt: Rewrite the text in the style of [AUTHOR]
Rewrite text from the perspective of [CHARACTER]
4. Generate titles/slogans
Prompt: Suggest 10 attention grabbing headlines for this message
Assistance in drafting texts (Part 3)
9. Example: Grant application for Global Partnership for Education
https://www.globalpartnership.org/content/program-development-grant-application
Prompt: I need to write a grant application for my project in the following structure:
[TEMPLATE STRUCTURE]. I will paste the template prompts for each section and you write up the
section based on my points. Here comes the first section:
„Describe the context for the proposed activities, including the status of the ESP or TEP, and
whether it will be a stand-alone program or contribute to an existing one, in 400 words or less.”
And here are my notes you can use:
- My notes about links to Education Sector Plan (ESP) or Transitional Education Plan (TEP)
- My notes about the context
Assistance in drafting texts (Part 4)
10. Example: Plan a social media campaign by Oxfam for International
Women‘s Day 2023
1. Plan structure
Prompt: Plan a social media campaign by Oxfam for International Women‘s Day
2023 that aligns with its mission statement. Propose a campaign motto, a detailed
schedule of activities and example social media posts.
Planning and Strategy
12. Example: Developing a training course
1. Curriculum structure
Prompt: Develop a curriculum for an online training course on WASH
Prompt: How would you structure the content for Module 2 to maximize
educational impact? What didactic methods and tools would you use?
2. Developing content
Prompt: Write out the script for module two based on the content structure and
didactic methods and tools described above
Learning and Teaching (Part 1)
13. Example: Learning about a topic and practicing knowledge
1. Personal tutor
Prompt: You are my personal training tutor on WASH. Engage in a dialogue
with me in which you test me in practical exercises. Give me the first exercise
and then wait for my response. After my input rate my response and give me
additional information. After that give me the next exercise and wait again for
my response.
Learning and Teaching (Part 2)
14. Example: Preparing for job interview
https://jobs.undp.org/cj_view_job.cfm?cur_job_id=111410
1. Personal job application trainer
Prompt: You are my job application trainer. Do a competency-based job interview
with me where you ask questions and I respond. Always only ask one question at a
time and wait for my response. Cover technical knowledge, soft skills as well as
professional experience and ask follow-up questions where appropriate. Here is the
job vacancy I‘m applying for:
„[TEXT OF JOB VACANCY]“
Learning and Teaching (Part 3)
15. 1. Asking for references when presented with facts and figures
that are difficult to verify yourself.
2. Asking for math or logic tasks.
3. Using texts generated by the bot without verifying them.
4. Interacting with the bot as if it had an inner life.
5. Transmitting personal or sensitive data.
What one shouldn‘t do
16. What does this all mean in context of KM4Dev?
1. Everyday Knowledge Work 2. The Future of KM
in Organisations
3. Societal Implications
17. 1. First drafts of content (commuication, FAQs, briefing
notes, proposals press releases, grant applications)
2. Summary of texts (protocols, publications, transcripts)
3. „Translation“ of complex texts (bureaucratic, legal,
medical, scientific) into easy language
4. Generation of ideas (names, applications, slogans, interviews)
5. As lectorate (grammar, stil, tone, voice, target audience)
6. As remedy for writer‘s block
7. For planning and analysis (schedules, analysis)
8. As tutor and training partner
1. Everyday knowledge work
18. First study by MIT:
Use of ChatGPT in office tasks leads to
1. 37% increased productivity
2. increased job satisfaction (related to the task)
3. increased self-efficacy
https://economics.mit.edu/sites/default/files/inline-files/Noy_Zhang_1.pdf
1. Everyday knowledge work
19. 1. Every organisation will maintain its own internal AI system
which knows all its internal content and can:
a) Answer any questions in natural language regarding
anything that was ever documented in written form
(and visible for everyone in the organisation)
b) Create new content in any format based on this
2. Currently still too expensive
3. Hallucination problem is not solved yet
4. Development organisations must ready themselves
to make an investment in this area in the next 2-3 years
2. The Future of Knowledge Management
20. 1. Massive disruptions for previously „safe“ jobs
1. Education (curriculms, educational materials, homework, grading)
2. Writing (journalism, authors, akademic publications, marketing)
3. Research (lawyers, consultants, analysts, acedemic research)
4. Design (graphic design, architecture, technical design, music, movies, gaming)
2. Productivity of economies will rapidly rise (or lag if they don‘t adapt),
„boring“ knowledge work will be outsourced to the machine
3. Information overload will massively increase
4. Democratic discourse will be very difficult
5. Generative AI will be used as weapon (disinformation, deep fakes, manipulation)
3. Deal With the Societal Impact
21. 1. Educate colleagues in your organisations about the potential and
practical applications of retail AI tools
2. Plan for building a Q&A machine for your own organisational
content by
• Identifying the best use cases in your organisation
• Prepare the data
• Test tools and scenarios via prototypes
• Plan financial investment
3. Discuss and prepare your organisation strategically for the
coming wave of societal change
Three Suggestions