Lab 1 Questions

Question 1

In what year Alan Turing develops the Turing Test

Answer 1

1950

Question 2

What year the term “artificial intelligence is first used

Answer 2

1956

Question 3

While at the Cornell Aeronautical Laboratory, Frank Rosenblatt develops the perceptron,
the first artificial neuron, that will set the stage for the development of networks of
artificial neurons, a.k.a. neural networks

Answer 3

1957

Question 4

Joseph Weizenbaum develops Eliza, the first chatbot. It is based on a set of re-write rules
written by hand.

Answer 4

1966

Question 5

Marvin Minsky & Seymour Papert publish a book that shows the limits of perceptrons
and argue for more work in symbolic computation (aka Good-Old-Fashioned AI, GOFAI).
The book is often cited as the main reason for the abandoning research on neural networks

Answer 5

1969

Question 6

The Lighthill and the ALPAC reports that showed little progress in AI, kills research
funding and leads to the first AI Winter.

Answer 6

early 1970s

Question 7

The robot Shakey1
, programmed in LISP, resulted in the development of the A* search
algorithm.

Answer 7

1972

Question 8

Alain Colmerauer, who was professor at the University of Montreal for a few years,
develops Prolog; a programming language based on logics that is very popular in AI to
write rule-based systems

Answer 8

1972

Question 9

Marvin Minsky develops the Frames to reason with world-knowledge. Years later, frames
turned out to be the basis of object-oriented programming.

Answer 9

1974

Question 10

The expert system MYCIN is developed to recognise bacterial infections and recommend
antibiotics. Its recommendations are often better than those of human experts. It is
based on a knowledge base of ≈ 600 hand-written rules (written in Lisp) and developed
in collaboration with medical doctors.

Answer 10

1975

Question 11

The METEO rule-based machine translation system, developed at the University of Montreal, is deployed at Environment Canada to translate weather forecasts from English to
French.

Answer 11

1975

Question 12

Expert systems, such as MYCIN, and other types of systems made of hand-written rules
are considered too expensive to maintain and to adapt to new domains. The industry
drops research in such systems. It is the 2
nd AI Winter

Answer 12

early 1980’s - early 1990’s

Question 13

Corinna Cortes and Vladimir Vapnik develop an approach to machine learning called soft
margin Support Vector Machines (SVM), which quickly becomes one of the most popular
machine learning algorith

Answer 13

1993

Question 14

After finishing his PhD on handwriting recognition, Yann Lecun makes public the MNIST
dataset. The dataset contains 70,000 images of handwritten digits and becomes the
benchmark to evaluate machine learning.

Answer 14

1998

Question 15

Google launches its Google Translate service based on Statistical Machine Translation.
Translation rules are found automatically based on a statistical analysis of parallel texts
in different languages.

Answer 15

2006

Question 16

Netflix initiates a competition in machine learning to beat its own film recommendation
system. It provides a data set of about 100 millions movie ratings to learn recommendations automatically

Answer 16

2006

Question 17

The AlexNet system, developed at the University of Toronto by Alex Krizhevsky, a PhD
student of Geoffrey Hinton, wins the ImageNet Challenge and shows that deep learning
techniques can achieve significantly better results than classical machine learning techniques in image processing. This is considered to be a defining moment in the history of
AI.

Answer 17

2012

Question 18

Google launches its Neural Machine Translation system based on recent advances in Deep
Neural Networks, and little by little drops the Statistical Machine Translation approach
of the years 2000’s.

Answer 18

2016

Question 19

While having a beer at Les 3 Brasseurs at the corner of McGill College and Ste-Catherine,
Ian Goodfellow, a PhD student of Yoshua Bengio, comes up with an idea for a neural network that could generate realistic images. He called it Generative Adversarial Networks,
a.k.a. GANs.

Answer 19

2014

Question 20

AlphaGo beats the world’s champion at the game of Go. AlphaGo’s strategy is learned
automatically from playing a large number of games against itself

Answer 20

2016

Question 21

Canada invests massively in 3 AI research institutes: Amii in Alberta, Mila in Montréal,
and Vector Institute in Toronto.

Answer 21

2017

Question 22

A Portrait of Edmond Belamy created by a Generative Adversarial Network (GAN) sells
for $432,500, at Christie’s auction in New York city.

Answer 22

2018

Question 23

The Association for Computing Machinery (ACM) names Yoshua Bengio (from MILA,
Montreal), Geoffrey Hinton (from Vector Institute, Toronto), and Yann LeCun (from
Facebook) recipients of the Turing Award for their contribution in the field of AI and
Deep Learning.

Answer 23

2018

Question 24

OpenAI publishes its GPT-2 model, astounding both public and experts with its text
generation capabilities, such as the Unicorn story generated using a prompt.

Answer 24

2019

Question 25

OpenAI publishes ChatGPT, a large language model (LLM) based on GPT-3, which has
a huge impact on the adoption of AI tools worldwide.

Answer 25

2022