Language tools: comprehending sentences
Most of us read, and unless you lose interest, you will be getting some meaning
from the words here, some understanding from the clauses I create and hopefully
develop a curiosity in the ideas which spring from the bundle of sentences which
follow.
When it comes to language, almost all of us are just great. Even the youngest of
us, or the inarticulate, have no trouble in conveying meaning.
A one year old child with a tiny vocabularly conveys meaning, and even the ignorant
and inarticulate can create a following.
We are almost all brilliant orators from the very first time we demand ‘juice, juice’.
We say and often get what we want. We achieve it without any understanding of how
language works.
All English speakers can do it, without the need to attend a single language class.
Remember AI ? Artificial Intelligence. AI started back in the 50’s: by 1968 HAL
could lip-read and lock astronouts outside the ship, and since 1955 perhaps 4 billion
people have managed to become fluent in one language or another, but not a single
computer became fluent during the same period.
Sometime in the next 20 years, a commodity CPU made perhaps by Intel, will have
more transistors than the human brain has neurons and a throw away SD card the size
of your finger nail will be able to store as much data as a small city of people
can recognise, and not so long after that one will be able to store as much as the
entire human race.
These advances have been happening at the same rate, roughly following Moore’s Law
since 1965, but where is AI, where is language understanding?
One way to judge language understanding capability of a machine, is to look at it’s
ability to translate. So let’s consider for a moment Google’s amazing translation
tool.
Google with it’s seemingly limitless resources, storage, CPU power and even brain
power have delivered a translation engine which has very mixed results.
At times astonishingly good, but often ludicrously wrong, it is simply not something
one can rely on for translation.
Why is that ? Why is arguably the best translation engine of them all with far more
capacity than any human brain, unable to compete with with even a single bilingual
person with perhaps no language training?
At Thinking Solutions we can see that for more than 50 years now, this has been
approached in the same way, the wrong way.
Computers are using either rule based, statistical analysis or algorithmic approaches
to a simple problem which is: "How can a machine understand language?". A better
question is: "How does the human brain comprehend?".
Thinking Solutions was established to enable change by improving the science and
then engineering the results.
Today, our Project Turing prototype is a machine that comprehends written sentences.
There is a long way to go from here before machines start talking to us wherever
we are, but the appproach works and as they say, the key to the cure is understanding
the cause.
Try the
to see what the fuss is about.