Artificial Intelligence, Machine Learning, and Deep Learning are one and the
same
False!
Recall that we defined AI as an artificial machine that can think like a human
to some degree. Humans learn through several different methods, such as
observation, trial and error, and, most efficient of all, pattern recognition.
Pattern recognition serves as the basis for machine learning, which is a method
of creating AI. Then again, there are different types of pattern recognition,
namely, statistical, structural, and neural pattern recognition. Neural pattern
recognition serves as the basis for deep learning, which is, by far, the most
powerful method used to implement machine learning. To sum it all up:
Deep learning is a method of implementing machine learning.
Machine learning is a method of creating artificial intelligence.
Intelligent Black Boxes
All AI systems are “black boxes,” far less explainable than non-AI techniques
False!
A black box system is a system that takes input and produces output through a
hidden process. The reason for hiding this process might be convenience, as the
process by itself may be too complicated to explain. However, note that this
model is only applicable in explaining complicated systems, and not all
computer systems are complex. Similarly, some AIs are complicated, and some are
not. Thus, not all AI systems are explainable via the black-box model.
It's All On the Data!
The data an AI trains on is not the only factor that determines its performance
True!
An AI's performance relies on relevant data, computer resources,
implementing algorithms, and human aptitude. These elements can all be
optimized, but none of them can ever be perfect. There is no such thing as a
flawless dataset, an infinitely powerful computer, a 100% efficient algorithm,
or an error-free programmer.
Biased Calculations
AI systems are inherently unfair
False!
Computers follow the exact instructions of their human programmers. Hence, if a
machine outputs a wrong result, it is the programmer's fault. Likewise, if
an AI outputs a result diverging from the expected, the programmer may have
made a mistake. For example, an AI trained to mimic the way humans make
decisions will eventually acquire the biases of the humans the AI was
mimicking. The situation I presented highlights the importance of good design
in developing an AI. If you want to make an unbiased AI, train it using
objective datasets. Conversely, if you plan to make an AI that mimics human
thinking, use datasets that account for human biases.
Robots At Work
Artificial Intelligence, Machine Learning, and Deep Learning are one and the
same
False!
If we look at history, we find out that there was always a fear of
technological breakthroughs causing mass unemployment. However, most, if not
all, technological breakthroughs had the complete opposite effect. Instead of
causing human labor to become obsolete, innovations made economies more
productive, thus opening up more jobs than those lost. AI might have the same
effect. Today's AIs only excel at specific tasks. Therefore, they cannot
take over all human jobs just yet, as most human jobs consist of multiple
interweaving tasks. However, AI may take over more mundane and repetitive
jobs. This change will most definitely displace some workers, and these workers
need to learn a lot of new skills to gain another job.
Man vs. Machine
AI is not yet approaching human intelligence
True!
As said previously, today's AIs only excel at specific tasks. They are
not capable of performing multiple interweaving tasks just yet. Indeed, some
AIs can outperform humans. However, these AIs can only do their one job. Ask
them to do anything else, and they will not budge. For example, if you ask a
chess bot to play checkers, it will not make a single move.
A Fun Bonus!
As a fun side note, have you ever wondered how an AI takeover would play
out? You may imagine a scenario like Terminator or I, Robot. However, this
scenario can also turn out quite comedic. Anyway, without further ado, here is
how this scenario might play out:
Let's say we have an AI built to farm potatoes. It will keep farming potatoes
diligently and maximizing the number of potatoes it produces. When it needs to
make more potatoes, it expands its farm. Soon enough, this potato farm will
reach human settlements. However, the programmer of the AI forgot to tell it
not to extend itself to human settlements. Thus, the AI will do everything in
its power to take over a nearby city to convert it into a potato farm so that
the AI can produce more potatoes. Eventually, the AI will take over the
world and convert it into one big potato farm.
