Machine Learning Misconceptions in Business by Emerj AI Research
1.
2. In this TechEmergence Consensus, we contacted
a total of 30 artificial intelligence executives and
researchers to ask them about the biggest mis-
conception that executives and businesspeople
have in applyting machine learning to business
opportunities.
3. This slide deck displays the major trends of
responses as well as some of the most poignant
quotes from the recognized experts we spoke with.
4. Access to complete data sets and all quotes
and answers from our Machine Learning in
Business Consensus is available for free
download as a spreadsheet or Google Sheet in
the link below. This series includes:
+ Machine Learning Industry Predictions
+ Deriving Value From Machine Learning in Business
+ Misconceptions in Machine Learning
+ Applications of Machine Learning
>> CLICK HERE
Download the complete response set below:
6. We’ve selected three quotes from each of the
major response categories. Beneath each quote
is a link (if available) of our complete interview
with this guest on the TechEmergence Podcast.
* These consensus answers were recorded seperately from our podcasts interviews, but most podcasts are focused on
related topics around the ethical implications of emerging technologies.
7. WRONG EXPECTATIONS OF
CAPABILITIES/APPLICATIONS
“There are two misconceptions. One, that ML can solve every-
thing, like a magic box. The other is exactly the opposite, that
ML is useless and can solve only toy like problems where the
solution is obvious. I believe the truth is in the middle, ML is
very good at classification, but is bad at real time control, and
motion planning, where continuous solution is required.”
- Dr. Amir Shapiro
Associate Professor, Ben Gurion University
8. WRONG EXPECTATIONS OF
CAPABILITIES/APPLICATIONS
“There is this misconception that with sufficient data you
can train a machine to solve any task. At least with the cur-
rent state of the field of machine learning there are types of
problems that are distinctly more or less suited for a machine
learning solution.”
- Dr. Pieter J. Mosterman
Chief Research Scientist, MathWorks
>> CLICK HERE
Listen to or read our full interview with Dr. Mosterman at techemergence.com:
9. WRONG EXPECTATIONS OF
CAPABILITIES/APPLICATIONS
“That it will be costly (it is not at all if you have the data),
complicated (most graduate in Computer Science can get you
a long way), and risky (the evaluation technics are simple and
will tell you how your system compare to a human).”
- Dr. Philippe Pasquier
Associate Professor, Simon Fraser University
>> CLICK HERE
Listen to or read our full interview with Dr. Pasquier at techemergence.com:
11. UNDERESTIMATING
RESOURCES/STAFF NEEDED
“A common misconception is that machine learning and AI
tools contain within themselves certain level of intelligence,
but they don’t. Machine learning and AI are only tools in hands
of more or less skilled people, and solely the intellectual
capabilities of these people eventually make a difference
between success and failure.”
- Dr. Danko Nikolic
Senior Professional: Data Science, CSC
>> CLICK HERE
Listen to or read our full interview with Dr. Nikolic at techemergence.com:
12. UNDERESTIMATING
RESOURCES/STAFF NEEDED
“That it is straightforward to develop and train ML systems
that solve real world problems. The amount of AI engineering
and training work required to bring ML systems to a useful
level is greater than what is assumed by the industry
executives.”
- Dr. Mika Rautiainen
CEO, Valossa Labs Oy
13. TECHNICAL
MISUNDERSTANDINGS
“ML is purely correlative - and correlation does not imply
causation. As many times as one sees this, the mistakes
made by people who don’t understand the difference leads to
significant negative consequences for business.”
- Dr. James Hendler
Professor, Rensselaer Polytechnic Institute
>> CLICK HERE
Listen to or read our full interview with Dr. Hendler at techemergence.com:
14. TECHNICAL
MISUNDERSTANDINGS
“General statements on classifier and techniques
performance. The accuracy performance depends not only on
the technique itself but on the data set you are analyzing.”
- Dr.-Ing. Aureli Soria-Frisch
R&D Neuroscience Manager, Starlab Barcelona SL
15. TECHNICAL
MISUNDERSTANDINGS
“Contrary to popular misconception, the size and quality of
training datasets tend to be significantly more important than
algorithm choice in applying machine learning to business
opportunities.”
- Dr. Alexander D. Wissner-Gross
Founder, President, and Chief Scientist, Gemedy, Inc.
16. NOT UNDERSTANING WHAT
AI IS OR DOES
“What people call Deep Learning are just a progressive
improvement on Neural Networks, a field that has been
slashed as “done” just a few years ago. The misconception is
that this is new: in reality, this is a decades long effort which
is now being notices thanks to hardware catching up.”
- Dr. Massimiliano Versace
President & CEO, Neurala, Inc.
>> CLICK HERE
Listen to or read our full interview with Dr. Versace at techemergence.com:
17. NOT UNDERSTANING WHAT
AI IS OR DOES
“The biggest misconception that they have around machine
learning is believing that they understand what it is.
Executives are very likely to be aggressively misinformed.”
- Slater Victoroff
CEO, indico
>> CLICK HERE
Listen to or read our full interview with Slater at techemergence.com:
18. NOT UNDERSTANING WHAT
AI IS OR DOES
“Artificial Intelligence in the long run is not so much about
robots and intelligent agents, but much broader: it is about
handling complexity in new ways. We will not live next to AI
applications, but inside of artificially intelligent systems.”
- Dr. Joscha Bach
Research Scientist, MIT Media Lab
>> CLICK HERE
Listen to or read our full interview with Dr. Bach at techemergence.com:
19. If you’ve enjoyed this presentation and you’d like
to see the full dataset of responses, the
consensus is freely available below:
>> CLICK HERE