APNIC Senior Network Analyst/Trainer Warren Finch presents on IPv6 implementation using AI at btNOG 10, held from 5 to 9 June 2023 in Paro, Bhutan.

1. 1
Preparing for IPv6 implementation
using Artificial intelligence (AI)
"IPv6 is not just an incremental improvement, it is a fundamental
shift in the way the Internet works. It is essential that we make
the transition now to ensure that the Internet can continue to
support the growing needs of the Asia-Pacific region." - Paul
Wilson, Director General of APNIC

2. 2
2
Agenda
• IPv6 statistics
• Overview of AI
• Overview of ChatGPT
• How to use ChatGPT
• Caution for AI use
• Questions

3. 3
3
IPv6: It’s taking off!
https://stats.labs.apnic.net/ipv6/BT

4. 4
…But there’s still a way to go
https://stats.labs.apnic.net/ipv6/XT

5. 5
5
What is APNIC Labs seeing in Bhutan?
• APNIC Labs uses paid advertising (placed by Google) to
measure end-user behaviour
– Uses undisplayed pixel fetches (we call them 1x1 invisible pixels)
– Tests DNS, IP and Transport behaviour fetching these assets
– Includes IPv6, DNSSEC, RPKI & QUIC tests at this time
• This is not a measurement of mbps or kbps
– It is a random sample of users, through HTML/browser/ads in games
– We don’t control placement numbers to the ISP
• We think it’s a reasonable approximation to usercount

6. 6
6
What is Artificial Intelligence (AI)?
Koteluk, O., Wartecki, A., Mazurek, S., Kołodziejczak, I., & Mackiewicz, A. (2021). How Do Machines
Learn? Artificial Intelligence as a New Era in Medicine. Journal of Personalized Medicine, 11(1), 32.
MDPI AG. Retrieved from http://dx.doi.org/10.3390/jpm11010032

7. 7
7
What is Artificial Intelligence (AI)?
• Artificial Intelligence (AI) refers to a broad spectrum of
technologies and applications, no universally agreed upon
definition
• Machine Learning (ML): uses algorithms to process and learn
from large amounts of data
– Supervised ML: model trained with labeled input data and a specified
output, human review for accuracy
– Unsupervised ML: model fed raw data to identify patterns without pre-
specified output, results reviewed by humans
– Reinforcement Learning: model learns through trial and error and
rewards to maximize net reward
– Deep Learning: model built on artificial neural network, processing large
amounts of unstructured data

8. 8
8
What is Artificial Intelligence (AI)?
• Natural Language Processing (NLP): machines read and
recognize text/voice, extract value, convert to desired
output
• Computer Vision (CV): enables computers to see and
process images like human vision, providing appropriate
output
• Robotics Process Automation (RPA): software tools
automate labor-intensive tasks for increased accuracy and
cost-savings

9. 9
What is ChatGPT
• ChatGPT is a natural language
processing (NLP) model
developed by OpenAI
• GPT (Generative Pre-training
Transformer) language model, is
a transformer-based neural
network trained on a large
dataset of text.
• The dataset is based on
information up to the end of 2021.
• May 24th 2023 Latest version https://nerdynav.com/chatgpt-statistics/

10. 10
10
ChatGPT
https://help.openai.com/en/articles/7947663-chatgpt-supported-countries

11. 11
11
ChatGPT
• Predictive text is a like a game of chess

12. 12
• Predictive text uses AI to suggest
next word.
• AI models process large data,
recognise patterns to predict
words.
• How would you finish this
sentence?
• Once upon a time …
ChatGPT
https://medium.com/@SereneBiologist/the-anatomy-of-a-chess-ai-2087d0d565

13. 13
13
ChatGPT
• Artificial intelligence (AI) tools like ChatGPT can be useful for
helping to plan and implement projects.
• Potential benefits of using the ChatGPT language model to write
your next IPv6 deployment project:
– Generate natural language text in a variety of styles and formats
– Generate detailed and accurate descriptions of IPv6 deployment
strategies, processes, and technical requirements.
– Generate configs and sample code
– Save time and effort in the project planning and documentation stages.
– Help to ensure that all necessary considerations and best practices are
taken into account in the IPv6 deployment project.

14. 14
ChatGPT – How to use
• Create an account
• Think of a topic, for example -
IPv6 deployment
• Create a list of questions to ask
• Refine or re-phrase the questions
• Do more research if required
https://twitter.com/solomania/status/1625458520811339777

15. 15
ChatGPT – Create an Account
https://chat.openai.com

16. 16
16
ChatGPT – IPv6 sample questions
• Do you know how to deploy IPv6 in an ISP?
• What are the necessary preparation before IPv6
implementation?
• Can you give a brief overview of rfc7381?
• What are the advantages of deploying IPv6?
• Can you give a business use case for deploying IPv6 in an
MPLS network?

17. 17
17
Do you know how to deploy IPv6 in an ISP

18. 18
18
Preparation before IPv6 implementation?

19. 19
Stages of the project
• Assess the current state of IPv4
infrastructure.
• Develop a strategy for
transitioning to IPv6.
• Obtain IPv6 addresses and
configure routers.
• Enable IPv6 on customer access
devices.
• Test the IPv6 network.
Assess
Develop
Design
Implement
Evaluate

20. 20
Give a brief overview of RFC7381

21. 21
Give a brief overview of RFC7381
• Actual title is:
– Enterprise IPv6 Deployment
Guidelines
• https://www.rfc-
editor.org/rfc/rfc7381

22. 22
22
ChatGPT – IPv6 sample questions
• What are happy eyeballs in regard to IPv6?
• Can you write the code for the happy eyeballs in bash?
• Can you give a work breakdown structure for an IPv6
project?
• Write a presentation outline for IPv6 preparation before
implementation?
• Can you list top 10 resources for IPv6 deployment or
migration?

23. 24
Happy Eyeballs – Create test script

24. 25
25
ChatGPT – Prompt engineering
• Asking a specific question for a targeted response
• Creating a specific context or scenario to generate text
• Using specific language or terminology to guide a language
model towards a desired response
• Setting constraints or limitations on what the language
model can generate
• Providing additional information or context to improve the
quality of the language model's response.

25. 26
26
ChatGPT – Prompt engineering
https://lifearchitect.ai/chatgpt-prompt-book/

26. 27
ChatGPT – Prompt engineering
• Specify the format
• Provide context
• Make instructions explicit
• Control output length
• Use system messages
• Iterative refinement
• Combine with human review
https://chat.openai.com/share/69e829b7-67c1-49ac-b897-b68f0848b5ab

27. 28
ChatGPT – improved question
• Ask ChatGPT:
“As an ISP, please provide a
detailed explanation of the steps
involved in configuring and
deploying IPv6 for a residential
customer. Include any necessary
hardware or software requirements,
address allocation considerations,
and recommended best practices
for a smooth transition from IPv4 to
IPv6.”
https://chat.openai.com/share/69e829b7-67c1-49ac-b897-b68f0848b5ab

28. 30
30
The good, the bad and the ugly

29. 31
31
The good, the bad and the ugly

30. 32
32
Caution for AI use

31. 33
33
The Call For AI Ethics
• AI Ethics: an Abrahamic commitment to the Rome Call
– organised at the Vatican City on January 10th, 2023, which brought together
leading religious representatives and international technology key players.
• Six principles
1. Transparency - AI systems must be understandable to all
2. Inclusion - These systems must not discriminate against anyone because
every human being has equal dignity
3. Accountability - there must always be someone who takes responsibility for
what a machine does
4. Impartiality - AI systems must not follow or create biases
5. Reliability - AI must be reliable
6. Security and Privacy - These systems must be secure and respect the privacy
of users
https://www.romecall.org/wp-content/uploads/2022/03/RomeCall_Paper_web.pdf

32. 34
34
Rise of the clones
• Bing AI from Microsoft
• Bard from Google
• LLaMA from Meta
• Wenxin Yiyan from Baidu (China)
• SearchGPT from Naver (Korea)
• YaLM 2.0 from Yandex (Russia)

33. 35
35
Rise of the clones
https://www.futuretools.io

34. 36
Summary
• ChatGPT is not specifically designed or
optimised for specific tasks, so it may
not be the most appropriate tool for the
job.
• ChatGPT may not always provide
accurate or useful information for tasks,
it could potentially serve as a starting
point or a source of ideas for further
research or exploration.
• Anything you post is public domain
• DO not post any company or client
information
• Create an account and test it for
yourself.
https://twitter.com/ChrisJBakke/status/1628877552940097536/photo/1

35. 37
List top 10 resources for IPv6
• https://www.ipv6.com/
• https://www.cisco.com/c/en/us/ab
out/security-center/ipj.html
• https://www.ipv6forum.com/
• https://www.nist.gov/itl/applied-
cybersecurity/ipv6-deployment-
guide
• https://www.internetsociety.org/re
sources/doc/ipv6/ipv6-migration-
faq/

36. 38
List top 10 resources for IPv6
• https://www.ipv6ready.org/
• https://www.federalregister.gov/do
cuments/2020/06/17/2020-
12350/ipv6-implementation-and-
deployment-guidance
• https://www.internetsociety.org/re
sources/doc/ipv6/ipv6-handbook/
• https://www.apnic.net/publications
/ipv6-migration-guide/
• https://www.ipv6.com/ipv6-
deployment-support-center/

37. 39
Questions?

38. 40
Bonus round
• Ask ChatGPT:
“I want you to act as a professor from MIT
who is the world leading expert in Artificial
Intelligence and GPT. For example: if asked
about, What is Artificial Intelligence? You may
answer something like, Artificial intelligence
(AI) refers to the simulation of human
intelligence in machines that are
programmed to think and learn like humans.
It involves the development of algorithms and
computer programs that can perform tasks
that typically require human intelligence, such
as recognizing speech, understanding natural
language, making decisions, and playing
games. There are several types of AI,
including rule-based AI, expert systems, and
machine learning.”
https://www.sans.org/blog/artificial-intelligence-what-to-tell-your-workforce

39. 41
41
References
• https://chat.openai.com/chat
• https://platform.openai.com/playground
• http://dx.doi.org/10.3390/jpm11010032
• https://lifearchitect.ai/chatgpt-prompt-book/
• https://www.mindmeister.com/1730579766/artificial-
intelligence
• https://www.mindmeister.com/map/2314647422/machine-
learning

40. 42
42
References
• https://prompts.chat
• https://www.synthesia.io
• https://ocw.mit.edu/courses/6-034-artificial-intelligence-fall-
2010/resources/lecture-videos/
• https://github.com/CFRCE/awesome-artificial-intelligence
• https://github.com/amusi/awesome-ai-awesomeness
• https://www.worldscientific.com/doi/pdf/10.1142/978981122
8155_0001

41. 43
43
Where to learn more
https://www.futuretools.io