The document discusses how machine learning can help automate decision making in the automotive industry. It provides examples of use cases like predictive maintenance and sales forecasting. While ML has potential, most businesses fail at it due to challenges like lack of skills and complex tools. The document argues that ML accessibility is increasing through platforms that make ML tools easier to use for various roles. This "democratization" of ML could help the automotive industry boost profits through applications like improving productivity, efficiency, customer satisfaction and quality.
Accelerating Machine Learning Adoption in the Automotive Industry
1. BigML, Inc
Accelerating ML Adoption
in the Automotive Industry
October 2019
Atakan Cetinsoy
VP - Predictive Applications, BigML
1
2. BigML, Inc
Machine Learning in a Nutshell
2
Applied ML is primarily about finding patterns in business data,
that can be used to make useful business predictions
PREDICTIVE MODELS
2
3. BigML, Inc
Automotive Industry Use Case Examples
33
Predictive Maintenance: Will this machine component fail?
Forecasting: How much of each vehicle model will we sell next quarter?
Supplier Risk: What will be the delivery performance per supplier?
Marketing: Which customers show affinity for shared mobility?
Finance: Is this transaction fraudulent?
Operations: Which manufacturing configurations are optimal to use?
4. BigML, Inc
Programming with Machine Learning
• Ultimately, Machine Learning is all about transforming data into models that
can be used to automate decision making.
44
ID COUNTRY CITY
DAYS SINCE
LAST PURCHASE
PAGE
VIEWS
LTV
PURCHASE
TODAY?
xyz US SEA 5 22 1448 Yes
abc US PBI 8 9 2330 No
def US CLT 20 2 22296 Yes
nnx US MIA 4 19 32342 Yes
sbd US ANC 1 21 1144 Yes
fjm US MSP 5 8 1589 No
cft US MSP 6 7 1299 No
amt US CLT 14 2 1250 Yes
AA US DFW 1 13 1464 No
vgg US ATL 3 15 17471 Yes
PREDICTIONS
BUSINESS DATA
ML PLATFORM
5. BigML, Inc
Democratizing Machine Learning — Why Now?
5
Maturity of ML
Techniques
Cost of
Computation
Abundance
of Data
Speed of
Computation
Easier Tools
6. BigML, Inc
The Economics of Machine Learning
• As the unit cost of predictions go down, many
facets of decision making will be automated via
cheap predictions.
• This means redesigning tasks with fewer human
predictions, but more human judgment.
6
The Machine Learning Revolution
Cheap Predictions
+ Fast (i.e., milliseconds)
+ Better: Quantifiable/Near
Human-level Error Rates
=>
7. BigML, Inc
Early Adopters — Google
7
• "Machine learning is a core, transformative way
by which we’re re-thinking how we’re doing
everything. We are thoughtfully applying it across
all our products, be it search, ads, YouTube, or
Play. And we're in early days, but you will see us
— in a systematic way — apply machine learning
in all these areas."
— Sundar Pichai, CEO
8. BigML, Inc
Machine Learning tools are
extremely complex
Machine Learning is intrinsically
complicated
8
Most businesses FAIL at Machine Learning :(
is going to revolutionize every industry and every organization BUT...
Machine Learning
11. BigML, Inc
Building a Machine Learning Product
11
Reality
https://medium.com/thelaunchpad/the-ml-surprise-f54706361a6c
Expectation
10.50 0.25 0.75
SOURCE: https://medium.com/thelaunchpad/the-ml-surprise-f54706361a6c
12. BigML, Inc
ML — Current State in the Automotive Industry
12
• Modest gains of AI/ML deployed at scale in 2018
among OEMs, suppliers, dealers from 7% to 10%
in one year.
SOURCE: Forbes
• Companies applying more measured approach
in selecting use cases and projects.
• “Scale champions” (3+ at scale projects) better at
•Up or re-skilling workforce
•AI/ML governance process
•Yet 80% still mention AI/ML as a strategic
initiative.
13. BigML, Inc
Automotive Vision 2030
13
• Slow (2%) growth in the traditional vehicle sales and related
aftermarket services.
• Automotive industry revenue to increase by $1.5T (30%) thanks to
new business models such as shared mobility and connectivity
services.
• 10% of cars sold in 2030 will be shared vehicles adding to special
purpose fleets and mobility-as-a-service solutions popular in dense
urban areas.
• Various flavors of EVs will make up to 50% of vehicles!
• New competing ecosystems with more diverse players will emerge to
deliver a much more integrated customer experience.
SOURCE: McKinsey Global Institute
• Integrated software and data-driven insights
as the connective tissue.
14. BigML, Inc
ML for Automotive — Unfulfilled Potential
14
• Application of Machine Learning can boost pre-tax profits of the industry by 5%
conservatively…and up to 16%.
• ML has a key role to play in the future of the automotive industry.
• Productivity
• Operational Efficiency
• Customer Satisfaction
• Quality
IMPROVE
• Direct Costs
• Customer Churn
• Time to Market
• Downtime
REDUCE
SOURCE: Capgemini
16. BigML, Inc
Machine Learning Accessibility Revolution
16
SOURCE: https://hbr.org/2019/06/when-ai-becomes-an-everyday-technology
“ After years of hype around mysterious
neural networks and the PhD researchers
who design them, we’re entering an age in
which just about anyone can leverage the
power of intelligent algorithms to solve the
problems that matter to them. Ironically,
although breakthroughs get the headlines,
it’s accessibility that really changes the world.
That’s why, after such an eventful decade, a
lack of hype around machine learning may
be the most exciting development yet.”
— Andrew Moore, Google
17. BigML, Inc
Tale of Two Innovation Approaches
17
AutoML / Standard Workflows
• ML-literate Analysts, Developers, Subject Matter
Experts, and Decentralized Data Science Staff
MLaaS Platform
Executive Mandates
Acqu-hire Talent
Strategic Initiatives
Bespoke Systems
• Centralized Data Science Staff and IT-led
Operationalization on Specialized Computing
Platforms and Open Source Tools
TOP DOWN / CENTRALIZED BOTTOM UP / GRASSROOTS
API-based Deployment
Parallel Experiments
Coexisting…
…for continuous
learning!
GovernanceShared
18. BigML, Inc
There’s More…
18
• Please visit us at the Thirdware
booth to
•see a live demo of the BigML
MLaaS platform and/or
•discuss your specific use case
of interest.
20. BigML, Inc
Key to the Vault — ML Workflows & Automation
2020
Instances
Data
New Instance
Prediction Confidence
%
ML Algorithm
LEARNING OR TRAINING
Evaluation
Predictive Model
SCORING OR PREDICTING
• Standardization of the end-to-end process
instills consistency, reliability, and collaboration.