WEBVTT

00:00:18.000 --> 00:00:23.560
In 2016, I was awarded a prize by the
National Forum for Teaching & Learning

00:00:24.320 --> 00:00:27.700
supported and sponsored
by the Ministry of Education in Ireland,

00:00:27.700 --> 00:00:30.970
peculiarly named:
"National Teaching Hero."

00:00:32.180 --> 00:00:33.680
The reason for this award

00:00:33.690 --> 00:00:36.820
was my availability towards my students

00:00:37.020 --> 00:00:38.250
and my capability

00:00:38.250 --> 00:00:42.640
to create less formal and more 
comfortable educational environments.

00:00:43.160 --> 00:00:46.240
Let's imagine this place as
a large university classroom.

00:00:47.880 --> 00:00:49.990
I'm used to enter the classroom,

00:00:49.990 --> 00:00:54.560
[in the] first ten minutes,
when students enter and take place,

00:00:54.560 --> 00:00:58.960
I plug my computer to the speakers
and turn a classical music on.

00:01:00.200 --> 00:01:03.080
I think this is the first step

00:01:03.080 --> 00:01:07.480
to build less formal and 
more comfortable educational environments

00:01:07.480 --> 00:01:10.440
and try to keep the attention
of students at a high level.

00:01:12.371 --> 00:01:14.340
Unfortunately, this is not always easy.

00:01:16.560 --> 00:01:22.640
In my lessons, I employ a method in use 
since the ancient times: storytelling!

00:01:24.250 --> 00:01:25.540
Pedagogically speaking,

00:01:25.540 --> 00:01:30.520
storytelling is a method
based upon the use of narratives,

00:01:30.520 --> 00:01:33.200
aimed at transmitting 
knowledge to students.

00:01:34.040 --> 00:01:37.800
I would start my lesson
exactly with this method,

00:01:38.040 --> 00:01:43.120
by explaining, describing a topic
on everyone's lips nowadays:

00:01:44.274 --> 00:01:45.770
Artificial Intelligence.

00:01:47.786 --> 00:01:50.220
Like every story narrated to children,

00:01:50.280 --> 00:01:54.150
I'd like to begin my story with 
"Once upon a time."

00:01:56.760 --> 00:02:02.210
Second World War, 1942, United Kingdom,

00:02:03.160 --> 00:02:04.500
Bletchley Park:

00:02:05.130 --> 00:02:09.030
a mansion house in the north of London.

00:02:10.479 --> 00:02:13.920
There was a thirty years old guy: 
Alan Turing.

00:02:15.680 --> 00:02:18.040
Alan graduated at King's College,
Cambridge,

00:02:18.670 --> 00:02:20.860
and he obtained 
a research doctorate in Logic

00:02:20.870 --> 00:02:23.770
at Princeton University, in the USA.

00:02:24.880 --> 00:02:30.120
At that time, the Germans made use
of a special machine: Enigma.

00:02:32.040 --> 00:02:34.470
It was like a typewriter:

00:02:34.470 --> 00:02:36.370
the operator typed some keys,

00:02:36.370 --> 00:02:41.090
but, instead of printing those letters
on a paper sheet,

00:02:41.090 --> 00:02:42.960
other letters were printed,

00:02:42.960 --> 00:02:46.600
according to an encoding mechanically
set under the machine.

00:02:47.440 --> 00:02:51.760
The Germans used this machine
to communicate with each other.

00:02:52.920 --> 00:02:55.550
Anybody had listened in on this sheet

00:02:55.550 --> 00:02:57.930
had in front a meaningless text.

00:02:58.950 --> 00:03:00.030
It was encrypted!

00:03:02.160 --> 00:03:07.880
Alan Turing was one of the leading figures
at Bletchley Park:

00:03:08.880 --> 00:03:12.810
he and his team implemented a machine,

00:03:12.810 --> 00:03:15.320
the one you can see behind me,

00:03:15.320 --> 00:03:20.320
able to decipher the texts 
written by the German.

00:03:21.680 --> 00:03:23.310
Due to this invention we believe

00:03:23.310 --> 00:03:27.480
that the war terminated two years earlier,

00:03:27.480 --> 00:03:28.820
saving many human lives.

00:03:30.480 --> 00:03:34.160
After the war Alan Turing continued
his research in Logic

00:03:35.840 --> 00:03:39.200
and he is considered
the father of Computer Science,

00:03:39.200 --> 00:03:41.900
the father of Artificial Intelligence.

00:03:42.720 --> 00:03:47.290
With his Turing machine,
he formalises the concept of computer

00:03:47.290 --> 00:03:50.500
even before the computer
was actually built.

00:03:52.480 --> 00:03:57.280
In 1950 he published a paper
on the journal "Mind":

00:03:57.280 --> 00:03:59.290
"Computing Machinery and Intelligence,"

00:04:00.220 --> 00:04:03.400
where he proposed the Turing test.

00:04:03.400 --> 00:04:07.610
The question behind the Turing test
is a well-defined one:

00:04:09.170 --> 00:04:11.380
Can machines think?

00:04:13.560 --> 00:04:16.110
It is at that time that
Artificial Intelligence begun.

00:04:18.730 --> 00:04:21.920
Probably most of you have watched 
the movie "The Imitation Game":

00:04:21.920 --> 00:04:23.770
The game of imitation.

00:04:23.770 --> 00:04:25.530
I am going to describe it to you.

00:04:26.840 --> 00:04:31.480
Let us suppose that a person is here,
pressing keys on a computer keyboard,

00:04:32.400 --> 00:04:34.540
who asks, pose some questions

00:04:34.540 --> 00:04:36.750
and on the opposite side of the computer

00:04:36.760 --> 00:04:39.730
there is a machine M and an operator O.

00:04:41.670 --> 00:04:43.560
Alternately,

00:04:43.560 --> 00:04:48.520
the machine M and the operator O
answer to the person P.

00:04:50.360 --> 00:04:54.840
It is said that the machine M
passes the Turing test

00:04:55.480 --> 00:04:58.290
if the person P is not able to understand

00:04:58.290 --> 00:05:00.720
when the answers come from the machine

00:05:00.720 --> 00:05:03.780
and when the answers 
come from the operator.

00:05:03.801 --> 00:05:05.670
This is The Imitation Game.

00:05:07.200 --> 00:05:10.923
The machine must have special features

00:05:11.203 --> 00:05:12.980
in order to pass the Turing test.

00:05:14.160 --> 00:05:16.990
It has to interpret natural language:

00:05:16.990 --> 00:05:19.050
the question asked by the person.

00:05:19.800 --> 00:05:22.960
It has to represent knowledge
in order to formulate answers.

00:05:23.840 --> 00:05:27.440
It has to think in an automatic mode
in order to formulate such answers.

00:05:28.480 --> 00:05:31.050
And it has to learn automatically.

00:05:33.120 --> 00:05:35.190
There are many approaches

00:05:35.190 --> 00:05:37.200
to study Artificial Intelligence.

00:05:37.640 --> 00:05:43.240
One of them is the cognitive approach:
it is based upon the human thinking.

00:05:45.280 --> 00:05:46.730
According to this approach,

00:05:46.730 --> 00:05:49.480
there are two ways to study
the human thinking:

00:05:49.480 --> 00:05:53.520
either we try to capture thoughts
right when they occur,

00:05:54.095 --> 00:05:57.000
or we try to model thoughts
at a psychological level.

00:05:59.154 --> 00:06:01.530
For this reason, 
we say that Artificial Intelligence

00:06:01.530 --> 00:06:05.840
is closely connected to neuronal and
cognitive sciences and to psychology.

00:06:06.110 --> 00:06:08.630
According to this approach,
the assumption is that,

00:06:08.630 --> 00:06:10.240
if we can have

00:06:10.240 --> 00:06:13.150
a true representation
of the human thought,

00:06:13.150 --> 00:06:15.500
then we can transfer it to a machine.

00:06:17.760 --> 00:06:21.260
Another approach is the one
based upon the laws of rational thought.

00:06:22.520 --> 00:06:27.240
Probably most of you has heard about 
the Aristotle's syllogism.

00:06:28.980 --> 00:06:30.570
Socrates is a man;

00:06:31.740 --> 00:06:33.510
all men are mortal;

00:06:34.550 --> 00:06:36.380
[then] Socrates is mortal.

00:06:36.920 --> 00:06:39.810
This is a deductive reasoning:

00:06:40.660 --> 00:06:43.410
If we have two truthful premises,

00:06:43.760 --> 00:06:45.810
we can infer a truthful conclusion.

00:06:45.900 --> 00:06:48.340
Deductive logics comes from here.

00:06:48.944 --> 00:06:51.530
According to the laws 
of rational thought approach,

00:06:51.880 --> 00:06:56.960
we try to build deductive arguments
and to transfer them to a machine.

00:06:59.600 --> 00:07:03.470
Another approach is 
the rational agent approach.

00:07:04.840 --> 00:07:07.800
A rational agent, an entity,

00:07:08.510 --> 00:07:10.510
has to act,

00:07:10.800 --> 00:07:13.480
has to adapt itself to the context,

00:07:14.110 --> 00:07:17.990
has to fix goals to itself,
and be able to carry them out,

00:07:18.480 --> 00:07:21.240
and it has to act in a rational way.

00:07:22.480 --> 00:07:26.400
Therefore, the Turing test
is related to intelligent agents.

00:07:27.040 --> 00:07:30.360
By rephrasing the Turing statement,
"Can a machine think?",

00:07:31.080 --> 00:07:35.280
we can now say: is it possible to build
a machine, an artificial agent,

00:07:36.320 --> 00:07:38.180
able to think,

00:07:38.180 --> 00:07:41.660
able to show understanding and rationality?

00:07:43.600 --> 00:07:45.780
Artificial Intelligence, therefore,

00:07:45.780 --> 00:07:50.200
aims at developing artificial 
intelligent entities.

00:07:50.200 --> 00:07:52.510
Your mobile phone is an entity.

00:07:54.720 --> 00:07:57.600
By developing artificial entities

00:07:57.720 --> 00:08:02.090
we try to understand intelligence
as a psychological construct.

00:08:02.504 --> 00:08:04.750
Once we have a better knowledge


00:08:05.260 --> 00:08:08.080
about this concept,

00:08:08.320 --> 00:08:11.360
we try to develop
artificial intelligent entities

00:08:11.360 --> 00:08:14.040
to support humans: it's a cycle.

00:08:15.980 --> 00:08:20.980
But let's see now whether
machines are able to think.

00:08:21.480 --> 00:08:24.570
I'd like to briefly describe you

00:08:24.570 --> 00:08:27.570
the state of the art 
of Artificial Intelligence

00:08:27.570 --> 00:08:30.319
and I'd like to use five classes

00:08:30.319 --> 00:08:34.790
to classify artificial agents
according to their abilities:

00:08:35.808 --> 00:08:39.520
we have sub-human and par-human agents,

00:08:39.520 --> 00:08:43.799
over-human, super-human agents
and then we have optimal ones.

00:08:44.068 --> 00:08:46.470
I want to explain them 
with some examples.

00:08:46.800 --> 00:08:51.040
Optimal agents are the ones
which act better than all the people

00:08:51.349 --> 00:08:53.070
and you can't do better than that.

00:08:53.280 --> 00:08:55.800
For instance, agents
which solve the Rubik cube,

00:08:55.800 --> 00:08:59.260
those that play at "Four in a row"
in the best way,

00:08:59.260 --> 00:09:00.400
or at Tic-tac-toe.

00:09:00.400 --> 00:09:06.000
Consider that some years ago, a boy,
given an initial state of the Rubik cube,

00:09:06.000 --> 00:09:08.800
solved it in 4.73 seconds.

00:09:09.373 --> 00:09:11.770
Some months ago an agent, 
a robot was developed,

00:09:11.770 --> 00:09:14.990
which can solve it in 0.63 seconds.

00:09:16.840 --> 00:09:22.160
We have super-human agents 
which act better than all the humans,

00:09:22.740 --> 00:09:26.510
for instance in the chess game,
in the Scrabble game.

00:09:27.170 --> 00:09:30.730
Some years ago the chess Russian 
champion Kasparov

00:09:30.730 --> 00:09:33.500
was defeated by an artificial agent.

00:09:34.480 --> 00:09:36.610
We have over-human agents,

00:09:36.610 --> 00:09:40.160
the ones which almost act better 
than most of the humans,

00:09:40.160 --> 00:09:45.160
for instance in the Texas hold 'em poker,
in answering the Quiz Show questions.

00:09:46.280 --> 00:09:48.343
We have par-human agents,

00:09:50.330 --> 00:09:54.500
which act almost like all the humans,

00:09:55.113 --> 00:09:57.578
for instance in cognitive activities,

00:09:57.578 --> 00:10:01.260
such as crosswords or image classification.

00:10:01.730 --> 00:10:06.200
Finally, we have sub-human agents,
which act worse than all the humans.

00:10:06.200 --> 00:10:12.250
Examples include objects classification,
handwriting recognition,

00:10:12.520 --> 00:10:16.120
vocal recognition, translation
from a language into another one.

00:10:17.770 --> 00:10:21.470
But if there is something that artificial
agents nowadays are not able to do

00:10:21.470 --> 00:10:23.920
is for instance, disambiguation:

00:10:23.920 --> 00:10:28.760
Are we talking about the apple as a fruit,

00:10:28.760 --> 00:10:32.480
or are we talking about the brand
of the Apple Corporation?

00:10:33.440 --> 00:10:35.850
And one thing that agents
are not able to do

00:10:35.850 --> 00:10:39.560
is reasoning in the real world
under situations of uncertainty.

00:10:40.520 --> 00:10:44.910
These are the main limitations
of Artificial Intelligence,

00:10:44.910 --> 00:10:48.780
and because of these, it is believed that
we are far away to pass the Turing test.

00:10:49.600 --> 00:10:55.570
Now, let's try to understand [whether]
machines will be able to think.

00:10:57.560 --> 00:11:00.880
Some years ago, in America,
a concept has been coined,

00:11:01.160 --> 00:11:03.280
named "Technological Singularity":

00:11:03.430 --> 00:11:08.230
by Ray Kurzweill, a world-renowned 
expert in Artificial Intelligence.

00:11:09.091 --> 00:11:10.930
Let's imagine a timeline.

00:11:12.480 --> 00:11:17.270
Let's imagine a line indicating
the human intelligence, increasing.

00:11:18.320 --> 00:11:23.050
Let's now imagine a red line
indicating the machine intelligence,

00:11:23.050 --> 00:11:25.140
with an exponential trend.

00:11:25.800 --> 00:11:30.880
This trend follows the Moore's law,

00:11:31.440 --> 00:11:34.980
whereby the computational complexity,
for instance,

00:11:34.980 --> 00:11:38.660
as measured by the number of transistors 
embedded in a chip,

00:11:38.900 --> 00:11:42.810
doubles every two years
and quadruples every [three] years.

00:11:43.720 --> 00:11:47.700
According to Ray Kurzweill, 
in 2010 we should have been able

00:11:47.700 --> 00:11:52.194
to use this computational complexity

00:11:52.194 --> 00:11:54.174
to emulate the human brain --

00:11:54.174 --> 00:11:55.780
I didn't see anything.

00:11:56.480 --> 00:11:59.840
In 2020, with 1,000 dollars
we will have access

00:11:59.840 --> 00:12:02.730
to this computational capacity.

00:12:03.960 --> 00:12:07.020
In 2025, according to Ray Kurzweill,

00:12:07.020 --> 00:12:11.110
we will be able to scan our brain 
in a very accurate way.

00:12:11.960 --> 00:12:14.220
And eventually, in 2029

00:12:15.120 --> 00:12:18.120
machines will pass the Turing test!

00:12:19.930 --> 00:12:24.400
And then, in 2045,
he refers to that point in time

00:12:24.400 --> 00:12:27.510
when the technological singularity
will happen,

00:12:29.240 --> 00:12:32.720
when the machines, machine intelligence,

00:12:32.720 --> 00:12:34.830
will follow an exponential trend

00:12:34.830 --> 00:12:38.720
that will significantly affect
the human intelligence.

00:12:40.760 --> 00:12:45.680
In his paper, published 
on the journal "Mind,"

00:12:45.680 --> 00:12:48.680
Turing not only proposed his test,

00:12:48.680 --> 00:12:52.780
but he also suggested nine objections
against his own test.

00:12:53.400 --> 00:12:58.100
These objections are nine objections
against Artificial Intelligence.

00:13:00.400 --> 00:13:04.360
Some years ago, when I was a student
at the University of Varese,

00:13:05.200 --> 00:13:09.320
I attended a course on "Epistemology,
Deontology and Ethics in Computer Science"

00:13:09.320 --> 00:13:12.960
held by Prof. Gaetano Aurelio Lanzarone.

00:13:12.960 --> 00:13:15.960
Unfortunately, he passed away 
some years ago.

00:13:15.960 --> 00:13:22.300
One of the assignments we had to do
was to propose a tenth objection

00:13:22.560 --> 00:13:26.000
against the Turing test,
against Artificial Intelligence.

00:13:27.120 --> 00:13:30.300
I was the only one who proposed 
a tenth objection

00:13:30.520 --> 00:13:32.900
expressed as a mathematical equation,

00:13:34.560 --> 00:13:36.640
that I labelled "human stupidity."

00:13:37.490 --> 00:13:39.980
I'd like to explain it in simple terms.

00:13:40.000 --> 00:13:44.300
Let's assume we take
the intelligence of all humans

00:13:44.300 --> 00:13:45.900
and we put it all together.

00:13:46.070 --> 00:13:48.610
The Sum symbol of the equation
on the left hand side.

00:13:49.729 --> 00:13:52.920
And we transfer this intelligence 
as a whole to a machine.

00:13:54.000 --> 00:13:57.610
Then, we get an equality of intelligence.

00:13:57.610 --> 00:14:00.860
But in some way the machine
becomes more intelligent than us.

00:14:01.910 --> 00:14:04.110
Though, if it is true that were us

00:14:04.600 --> 00:14:08.653
who have transferred our intelligence
to a machine

00:14:09.313 --> 00:14:12.313
and it becomes more intelligent than us,

00:14:12.313 --> 00:14:13.333
it is also true, as well,

00:14:13.333 --> 00:14:16.900
that we let the machine become 
more intelligent than us.

00:14:19.160 --> 00:14:21.500
Then, in order to conclude my story,

00:14:23.160 --> 00:14:28.430
and referring back to the initial 
question: "Can machines think?",

00:14:29.640 --> 00:14:33.516
I'd like to leave you with
an open question:

00:14:36.321 --> 00:14:39.511
Does really makes sense 
for us to let them think?

00:14:40.751 --> 00:14:41.741
Thank you.

00:14:41.841 --> 00:14:43.911
(Applause)