Why Has Artificial Intelligence Failed to Produce Results

Artificial intelligence has been a goal since theminds work on a system level not on the level of
beginning of computers, why is it that after overthe neuron. If we study what people do, what
50 years of work there are still no systems outdrives them then we will have an understanding
there that can be said to be intelligent. Computerof what is needed to make a system look smart.
power has increased by massive amounts, millionsSo where do we start?
has been spent on research but still no majorComputers know nothing about the real world.
results. This article talks about the current batchPeople take years to develop an understanding of
of chatbots and where AI should be going.the world; this "assumed knowledge" is taken for
I should state at this point that I believegranted but is a vital part of what makes us
intelligence is a set of processes in the brain thatwhat we are. Even a child knows a bird can fly,
is perfectly possible to emulate in software soflames are hot and knives are sharp and can hurt.
that a true artificial mind can be produced. I haveComputer systems know nothing about any of
been interested in Artificial intelligence almost sincethis, they have no assumed knowledge.
I first got into computers over 30 years ago. TheIf a computer system is to be intelligent it must
promise of making computers smart, being ablehave assumed knowledge, it must either have
to talk to them seemed like a tantalizingly closeways to accumulate this knowledge (camera's and
dream. How hard could it be? The fame andmicrophones) or we must give the knowledge to
fortune that would follow the person who firstthe computer. Chatbots or any AI system must
worked it out would be epic. I was hooked.have "assumed knowledge". Ideally a system
30 years later and I am still interested but theshould learn this assumed knowledge but the
dream is just as far away now as it was then.technology to do this is another whole subject on
What happened?its own. So to produce a system in a reasonable
In a nutshell, nothing much.time we must give the knowledge to the system
Researches worked on the problem, systemsat the start.
were built but no real results, no systems then orKnowledge is not everything.
now produce any sign of real intelligence. ThereKnowing stuff does not make you intelligent, using
are a number of systems (so called chatbots)those facts to your advantage is a sign of
which are designed to look like intelligence. I guessintelligence. Assessing advantage is the next key
the most famous is the chatbot called ELIZAto intelligence. Even simple creatures discover
which was designed to act like a computertechniques and new methods to get food, fight
psychologist. It was simple and had a stock set ofenemies and stay warm, they assess what
answers to certain keywords typed by the user.happens and form opinions on those events
Obviously the results were patchy, it fooled somedesigned to give advantage. Any AI system must
people but very quickly showed up as just a toy.be able to assess advantage to it and therefore
There have been a number of systems sincemust have goals. You cannot access an
then that have tried to continue this line of work.advantage if you don't know what you want
I have designed a few myself, they scan theneed.
user input to try to get a sense of what the userMood.
is saying and then use a set of prepared answersAll intelligent systems have moods (emotion),
to give the impression of a real person giving realsometimes they are angry, happy, sad, violent,
answers. These systems have real limitations andand passive. These mood varieties are designed
are still no where near what is needed.to aid the individual to survive but they also have
There are a number of other systems based ona direct effect on intelligence, the mood you are
neural nets that are designed to emulate how thein directly effects the way you think, work things
human brain works (a series of cells interlinked).out and assess advantage.
Since they are designed to work in similar waysIn summary all intelligent systems must have:
to the way the brain works it is a valid approach.- Priority (Goals)
Again after decades of work not much of any- Advantage Assessment
use has been produced. This approach is similar to- Moods (Emotion)
trying to work out how a computer works by+
studying a transistor. It may work but it is going- Assumed Knowledge
to be very difficult.I will shorten this list to p.a.m. + a.k
So what do we do? How can we progress afterIn my next article, I will discuss each of these
all this time. It is my assumption that if you mimicelements in more detail and how we might go
the way people behave then we can buildabout building this into a software program.
intelligent systems, we should copy the way our