Slide 1
Slide 1 text
Big Data meets Frugal AI
Big Data Expo - Utrecht 2022
Shaun McGirr
RVP AI Strategy - EMEA, Dataiku
Slide 2
Slide 2 text
More data
≄
More value
Slide 3
Slide 3 text
Welcome skeptics!
1. Seen it all before
2. Sampled a population
3. More data -> more cost(s)
Slide 4
Slide 4 text
No content
Slide 5
Slide 5 text
No content
Slide 6
Slide 6 text
No content
Slide 7
Slide 7 text
More data ≄ More value
1. Seen it all before ✅
2. Sampled a population
3. More data -> more cost(s)
Slide 8
Slide 8 text
No content
Slide 9
Slide 9 text
Accounting
≄
Insight
Slide 10
Slide 10 text
Plenty of
consequences
Not many
consequences
Plenty of time Sample to iterate! Sample for
efficiency!
Not enough time Sample for speed! Sample for focus!
Slide 11
Slide 11 text
Plenty of
consequences
Not many
consequences
Plenty of time Sample to iterate! Sample for
efficiency!
Not enough time Sample for speed! Sample for focus!
Many of the decisions we want to make, do not
require that we collect, store or process more data.
Slide 12
Slide 12 text
More data ≄ More value
1. Seen it all before ✅
2. Sampled a population ✅
3. More data -> more cost(s)
Slide 13
Slide 13 text
2012
Data is the new oil
Gather as much as you can
Store it as long as you can
Process it any way you can
Slide 14
Slide 14 text
2012
Data is the new oil
Gather as much as you can
Store it as long as you can
Process it any way you can
2022
Who wants data to be like oil?
Do we have the right to gather it?
How long can we defend storing it?
What processing will generate value?
Slide 15
Slide 15 text
Frugal AI:
training AI systems
with little resources
Slide 16
Slide 16 text
No content
Slide 17
Slide 17 text
3. More data -> more cost(s)
Economics of AI implode
Harms of AI explode
AI replaces us
Slide 18
Slide 18 text
How to do things to data? ✅
Slide 19
Slide 19 text
How to do things to data? ✅
What should we do with data?
And who will do the work?
Slide 20
Slide 20 text
Meet your future Citizen Data Scientists
Commercial Analyst Statistician / Research Scientist
Product Manager
I want a safe playground with
a low entry barrier to make
better and faster decisions
I am already an expert in my
field but want to explore new
techniques to innovate faster
I want to be a strategic
adviser to the business but I
get stuck with reporting
Slide 21
Slide 21 text
In semiconductor manufacturing, a critical quality and
manufacturability figure of merit is the ability to detect and resolve
manufacturing issues as quickly as possible (TTD - time to detect).
Thanks to Dataiku, Adnan was able to leverage virtual
metrology to detect issues real-time.This resulted in millions
of savings in terms of material and engineering costs.
Adnan participated in NXP’s state of the art Citizen Data
Scientist program, who has upskilled 200+ employees to
apply data science to their daily job
How an NXP quality engineer leveraged
machine learning in Dataiku to reduce
time to detect quality issues and won a
Frontrunner Award!
Adnan Chowdhury
Manufacturing Quality Engineer
https://community.dataiku.com/t5/Dataiku-Frontrunner-A
miconductors-Reducing-Detection-Time-of-Manufacturi
Slide 22
Slide 22 text
➔ Under many conditions, we will deliver
more business value with less data
➔ Frugal AI unifies avoidance of “bigger is
better” thinking, with sensible product
design principles like “less is more”
➔ Ask questions of your customers first,
and shoot for your solution later
Big Data meets Frugal AI
.Takeaway.
.messages.