Database Reference
In-Depth Information
female to go into space. The data input never separated fictional women from nonfictional
women. And the machine was only as good as its data.)
What an AI application - in this example, a natural language processing AI system -
can
do, however, is take digital recordings of inbound customer-service phone calls and
identify all instances of the words “complaint,” “dissatisfied,” and “unhappy,” then match
these instances with the words “delivery,” “installation,” and “hot-tub,” to derive data as
to what aspects of your hot-tub sales business have annoyed those customers who have
wound up annoyed. Data Scientists can then pair this data with other references, such as
geographical region, time of year, hot-tub model, and so forth to derive actionable BI.
What
the computer cannot do is deliver a qualitative analysis based on these quantitative results.
For that we need humans.
The machine is, however, capable of dealing with rudimentary
fact-based what-if
scen-
arios - scenarios which must first be imagined and conjectured by human Data Scientists
posing the right questions upon the right data.
For example, factoring in prescribed mean average temperatures throughout the year
as segmented by geographical region, and the minimum temperature necessary for the putty
involved with hot-tub installation to settle correctly, and pairing this data with customer
complaints related to “installation” segmented by region,
where are the dateline “tipping-
points” at which installations can best proceed and had best stop in Maine as opposed to
Ohio?
This type of thing the machine can handle. But the intuition to pose the question re-
mains, and always will remain, within the human sphere.
In short, AI applications for Data Science - of which there are more and more being
released every day - are perfectly capable of combining multiple data points and isolating
patterns in data; but it is up to the Data Scientist to postulate which patterns to look for, and
to draw and/or infer conclusions from those patterns. So-called “machine learning” apps
with “intelligent-seeming algorithms” can certainly handle sophisticated statistical analys-
is. Indeed, they can even get better at this over time, on their own; but they will never be
capable of creative curiosity and intuitiveness.
Yann LeCun, director of Artificial Intelligence at Facebook, stresses how very import-
ant it is to understand the line between human thinking and AI Machine Learning, most
especially the latest focus of Machine Learning called “Deep Learning” with neural net-
works.
LeCun doesn't like it when people say “[Deep Learning] works just like a brain.” As
LeCun explains, although “Deep Learning gets [its] inspiration from biology, it's very, very
far from what the brain actually does. And describing it like the brain gives a bit of the
aura of magic to it, which is dangerous. It leads to hype; people claiming things that are not
true.”
