TRYEngineering - AI Adventures - 10

How Does AI Learn and
Get Smarter?
Fun Fact: Deep Blue
IBM developed Deep Blue,
a chess-playing AI through a
rule-based system instead of
machine learning. Deep Blue
demonstrated exceptional
speed, evaluating up to 200
million possible chess positions
per second, and processed a
large collection of chess moves
for reference during games.
Basically, instead possessing
a deep understanding of how
to play chess, it had a giant
cheat sheet of chess moves
that it would quickly review to
determine its next move. This
type of learning is not ideal
because it is time-consuming,
and the AI system doesn't learn
from new experiences. If Deep
Blue ever came across a chess
move it hadn't seen before,
it wouldn't know what to do,
and it still wouldn't know what
to do the next time it saw that
move. Thankfully, Deep Blue's
strategy proved successful, as it
defeated the 1997 world chess
champion, Garry Kasparov,
demonstrating its qualification
as an AI system that can think
like a human!
WATCH: The Garry
Kasparov vs Deep Blue
1997 Game 1 video
from The Chess Giant
As we said earlier, AI learns and gets smarter in a few different ways. Most of
them are types of machine learning, but some are not.
One popular method is a rule-based system, also known as symbolic
AI, which uses many rules that people directly code into the AI to make
decisions. This enables the AI to make precise decisions based on it's pretaught
knowledge, but rule-based systems lack the ability to learn from fresh
data. It's similar to being given a list of flower names and having to guess the
color of each one. The list says a rose can be red or yellow, but when you
guess its color, it ends up blue! Since you can't change what the list says and
can only go off the list, you'll never end up guessing if any roses are blue. This
way of teaching AI can still work, but only when you can put in enough rules
to account for every possible option or every likely option. You can read more
about the creation of Deep Blue, chess AI (see the Deep Blue Fun Fact)
in this ebook. Another example is Watson, an AI system that was designed to
understand and process human language. It gained fame by winning the quiz
show Jeopardy! against top human contestants in 2011.
WATCH: Here is a short video about Watson & the IBM Challenge,
by IBM Research.

https://www.youtube.com/watch?v=P18EdAKuC1U https://www.youtube.com/watch?v=WzIvNVBRp40 https://www.youtube.com/watch?v=WzIvNVBRp40 https://www.youtube.com/watch?v=WzIvNVBRp40

TRYEngineering - AI Adventures

Table of Contents for the Digital Edition of TRYEngineering - AI Adventures