Generative AI technologies like ChatGPT are soaring to the peak of Gartner's Hype Cycle and with it lofty expectations. This talk aims to ground the conversation by spotlighting real-world, actionable uses of Generative AI in FME workflows. From cleaning up text in poorly scanned PDFs to translating complex database schemas and even fine-tuning your regular expressions, we'll delve into practical examples pulled from our latest projects. This interactive session will not only demonstrate FME's capability to use various Generative AI tools but also tackle important considerations around data privacy. Join us to demystify the hype and discover how you can leverage Generative AI in your FME toolkit to speed up your FME projects and solve tricky challenges today.

1. Using Generative AI in FME Today:
Practical Use-Case Examples

2. Presenters
The
Peak
of
Data
Integration
20
23
Oliver
Morris
Business Director
Tensing
Amir
Rabbani
FME Specialist
Tensing

3. The
Peak
of
Data
Integration
20
23
Agenda
1. Intro - hype to reality
2. It’s not just chatGPT (and why that
matters)
3. Use cases with FME
○ clean up poor OCR’d PDFs
○ translation - databases and code
○ AI Assist in FME - regex for email validation
4. Wrap up

4. The
Peak
of
Data
Integration
20
23

5. The
Peak
of
Data
Integration
20
23
Photo by Chris J Ratcliffe/Getty Images

6. The
Peak
of
Data
Integration
20
23
Generative AI ≠ chatGPT
Models
• Proprietary (gpt-3.5-turbo, gpt-3.5-turbo-16k, gpt-4)
• Open (LLaMA, RedPajama - open llm leaderboard)
Platforms
• OpenAI (API)
• Microsoft Azure OpenAI
• aleph-alpha
• Together.xyz

7. The
Peak
of
Data
Integration
20
23

8. The
Peak
of
Data
Integration
20
23

9. The
Peak
of
Data
Integration
20
23

10. The
Peak
of
Data
Integration
20
23

11. The
Peak
of
Data
Integration
20
23

12. The
Peak
of
Data
Integration
20
23
Security

13. The
Peak
of
Data
Integration
20
23
“Enough already, show me why this all
matters to me”
Lizard
Semi-retired FME user

14. The
Peak
of
Data
Integration
20
23
Recent use cases

15. The
Peak
of
Data
Integration
20
23 Fill the gaps in poor quality OCR scans

16. The
Peak
of
Data
Integration
20
23 Database Schema translation

17. The
Peak
of
Data
Integration
20
23
AI Assistance on your regex

18. The
Peak
of
Data
Integration
20
23
AI Assistance on your regex
https://nymag.com/vindicated/2016/10/clippy-didnt-just-annoy-you-he-changed-the-world.html

19. The
Peak
of
Data
Integration
20
23 Regular Expression Editor - AI Assist

20. The
Peak
of
Data
Integration
20
23 Regular Expression Editor - AI Assist

21. The
Peak
of
Data
Integration
20
23 Regular Expression Editor - AI Assist
I am looking for a highly accurate and comprehensive regular expression
(regex) solution to identify valid email addresses. My goal is to filter out
invalid email formats while capturing as many valid formats as possible,
adhering to the general standards for email format as specified by RFC
5322 or similar guidelines.
Please consider the following requirements when crafting the regex:
It should allow for both uppercase and lowercase alphabetic characters.
It must support domain names with multiple sub-domains (e.g.,
john.doe@example.co.uk).
It should account for various common domain extensions like .com, .org,
.net, etc., but also allow for new or less common ones (e.g., .guru,
.consulting).
Special characters commonly used in email addresses, such as dots,
underscores, and hyphens, should be supported in the appropriate parts
of the email.
The regex should limit the length of the email address to a reasonable
number to avoid exploitation attempts (e.g., a limit of 320 characters,
which is the max allowed according to RFC 5322).

22. The
Peak
of
Data
Integration
20
23 Regular Expression Editor - AI Assist
I am looking for a highly accurate and comprehensive regular expression
(regex) solution to identify valid email addresses. My goal is to filter out
invalid email formats while capturing as many valid formats as possible,
adhering to the general standards for email format as specified by RFC
5322 or similar guidelines.
Please consider the following requirements when crafting the regex:
It should allow for both uppercase and lowercase alphabetic characters.
It must support domain names with multiple sub-domains (e.g.,
john.doe@example.co.uk).
It should account for various common domain extensions like .com, .org,
.net, etc., but also allow for new or less common ones (e.g., .guru,
.consulting).
Special characters commonly used in email addresses, such as dots,
underscores, and hyphens, should be supported in the appropriate parts
of the email.
The regex should limit the length of the email address to a reasonable
number to avoid exploitation attempts (e.g., a limit of 320 characters,
which is the max allowed according to RFC 5322).

23. The
Peak
of
Data
Integration
20
23
Wrap Up

24. The
Peak
of
Data
Integration
20
23
Key takeaways
● There is a lot going on in the AI space - don’t worry FME
makes it easy to dip your toes in.
● It can be useful to you right now - give it a go!
● Please share your success

25. The
Peak
of
Data
Integration
20
23
Resources
● I will make the presentation available later today.
● YouTube channels - @FMEEvangelist, @SophiaYangDS,
@giswqs, @Fireship, @paulramsey1331, @AllAboutAI,
@AIJasonZ, @samwitteveenai, @matthew_berman,
@FMEchannel

26. ThankYou!
omorris@tensing.com
Tensing FME UK
Tour
www.tensing.com/en/fme-uk-tour