# tactiq.io free youtube transcript
# What is AI?
# https://www.youtube.com/watch/7RKrsXccr4k

00:00:01.280 what is artificial
00:00:03.440 intelligence there are many definitions
00:00:05.720 of artificial intelligence each
00:00:07.960 organization each author has their own
00:00:11.759 and over history these definitions have
00:00:14.400 changed a lot one of the reasons for
00:00:17.119 that is that one of the two components
00:00:20.080 of artificial intelligence the word
00:00:22.720 intelligence is very difficult to Define
00:00:26.199 if I look at Wikipedia which is a very
00:00:28.599 nice place to look for definition S I
00:00:30.960 will read something like this it is
00:00:33.719 artificial because it is not human or it
00:00:36.360 is not animal and instead of defining it
00:00:39.920 per se we will just Define some of the
00:00:42.800 different ways in which it can be
00:00:44.920 enacted and we know that speech
00:00:47.079 recognition that Vision translations and
00:00:51.160 many other operations are part of
00:00:54.120 artificial intelligence but you can look
00:00:57.680 and find really exciting definition
00:01:00.440 out there there are two big families of
00:01:02.879 Ai knowledge-based Ai and databased AI
00:01:07.240 knowledge based AI is the AI based on
00:01:10.240 the knowledge which is shared typically
00:01:12.240 by humans and then programmers computer
00:01:15.560 scientists are going to encode this uh
00:01:19.240 knowledge into programs and this is what
00:01:22.400 we will be using and interrogating when
00:01:25.079 we use AI typically this is also very
00:01:28.840 often known
00:01:30.479 expert systems on the other hand we've
00:01:33.280 got databased AI where it's data that is
00:01:37.520 going play the key part and computer
00:01:40.640 programs known as machine learning
00:01:42.720 programs are going to transform that
00:01:44.920 data into programs or sets of rules
00:01:49.399 obviously this distinction can be looser
00:01:53.159 than it looks and you can have a mixture
00:01:55.880 of both some knowledge and some data
00:01:58.840 modern approaches will take both
00:02:01.439 approaches
00:02:03.240 simultaneously let's get back to
00:02:05.079 knowledge-based AI so a lot of
00:02:08.000 activities of our so-called intelligence
00:02:11.760 require planning require reasoning
00:02:15.360 require problem solving require
00:02:17.800 heuristics some special types of Logics
00:02:21.080 solving complexer situations all these
00:02:24.720 can be done through
00:02:26.840 AI um some of the experts believe that
00:02:31.319 the right way to approach these problems
00:02:33.920 is through what is called symbolic AI
00:02:37.000 symbolic AI means that instead of just
00:02:39.640 crunching numbers we're going to try and
00:02:42.480 organize Concepts idea logical formula
00:02:46.319 in order to build um intelligent
00:02:49.680 algorithms there are experts today who
00:02:52.440 believe that even if databased Ai and
00:02:56.840 numerical based AI is uh having the edge
00:03:01.280 in the future symbolic AI will become
00:03:04.840 important and one of the key reasons is
00:03:07.720 that current AI Technologies seem to be
00:03:10.920 very expensive in other words the
00:03:13.519 quantity of information a modern
00:03:16.239 database AI needs to perform simple
00:03:19.879 tasks is much larger than what a child
00:03:24.040 needs so perhaps we may not be using the
00:03:27.519 best algorithms at this point
00:03:30.319 databased AI is the other approach it's
00:03:34.280 when we use data which has been
00:03:36.640 collected over the web and this data can
00:03:40.360 be then transformed through machine
00:03:43.040 learning into um programs a key element
00:03:47.640 here is that the data is represented
00:03:49.959 through features typically so there is a
00:03:53.120 description of the data allowing to
00:03:55.920 extract the regularities that are needed
00:03:59.120 to invent new rules or to uh make future
00:04:04.439 decisions let's have a look at the
00:04:06.519 history of AI some will are say that AI
00:04:10.120 was already present in the dreams of the
00:04:12.599 writers of the 18th and 19th century or
00:04:15.519 perhaps even in the ideas of the ancient
00:04:18.639 Greek more traditionally we will say
00:04:21.680 that AI would be born with the initial
00:04:24.720 ideas by Alan churing who was also
00:04:27.759 extremely important in computer science
00:04:30.560 and in 1950 alent churing produces a
00:04:32.960 paper where he discusses what an
00:04:35.800 intelligent machine would be like 1956
00:04:39.479 is the next key date in
00:04:42.000 19556 some experts will meet up in a
00:04:45.400 university in North America and write
00:04:49.479 out the agenda for the next few years of
00:04:51.840 what artificial intelligence should
00:04:54.199 achieve and for years and years these
00:04:56.919 researchers and others have tried to to
00:04:59.800 build intelligence machines these dreams
00:05:02.880 have sometimes reached success and in at
00:05:06.520 least two occasions they have reached
00:05:09.240 what is called an AI winter IE moments
00:05:13.199 where the general public and also the
00:05:16.520 people funding research started to
00:05:19.039 disbelieve that AI could achieve
00:05:21.520 anything
00:05:22.600 important a very important key day is
00:05:25.840 1997 this is when Gary Kasparov the
00:05:28.759 reigning world Champion was beaten by
00:05:32.400 Deep Blue deep blue was the first AI
00:05:36.199 system to proably be better than a human
00:05:40.400 at a really complex task or at least a
00:05:43.000 task that at that point we thought was
00:05:44.880 very complex which was the game of chess
00:05:47.919 and that was probably the moment where
00:05:50.400 Humanity started realizing that there
00:05:53.000 was the possibility of having a machine
00:05:55.600 more intelligent than man even if then
00:05:59.520 and today we were not capable of really
00:06:03.080 understanding what being more
00:06:04.600 intelligent meant for over 40 years
00:06:09.520 symbolic AI dominated the field and the
00:06:13.360 history of the more recent Earth is in
00:06:17.080 fact the history of machine learning and
00:06:20.560 then deep learning to finish let's have
00:06:23.840 some thoughts about AI obviously we
00:06:26.960 would need much more thoughts and it is
00:06:28.960 very interesting to notice that the
00:06:31.199 community of philosophers are today
00:06:33.880 reflecting a lot on the question of AI
00:06:37.199 and the ethical implications but at this
00:06:40.039 point just two things the first one is
00:06:43.560 the word intelligence is clearly
00:06:46.639 misleading it is difficult because we
00:06:50.000 are not even sure of what intelligence
00:06:52.560 means from a negative point of view you
00:06:56.160 could think that the word intelligence
00:06:58.000 is just moving as the machine itself is
00:07:01.440 achieving tasks in other words something
00:07:05.039 is regarded as intelligent until the
00:07:07.879 machine is able to do it and then we
00:07:10.360 think aha intelligence has to be
00:07:12.360 something else on the more positive node
00:07:15.759 every time the Mach machine manages to
00:07:19.199 solve a task that until then we believe
00:07:22.319 to be intelligent that has obliged us
00:07:25.039 humans to reflect again on what
00:07:27.840 intelligence meant and what makes us
00:07:31.280 unique the second reflection at this
00:07:34.479 point is that for about 60 years the key
00:07:38.680 task was that of solving the touring
00:07:41.240 test in other words the task of
00:07:45.000 simulating intelligence if you managed
00:07:47.960 to build a machine that could let
00:07:50.199 everyone else believe that they were not
00:07:52.879 addressing a machine with a human you
00:07:55.720 had probably won and achieved the
00:07:58.599 touring test
00:08:00.159 then for a few years we have forgotten
00:08:01.919 about it some authors today in April
00:08:07.159 2023 believe that the touring test has
00:08:09.960 been passed by the very latest
00:08:12.720 generation of generative
00:08:15.440 AI so there is also something to reflect
00:08:19.319 there what is the next goal of
00:08:21.680 artificial
00:08:28.240 intelligence