Artificial
intelligence is advancing so fast that we don’t have time to understand it and
adjust to it.
Five U.S.
companies are racing toward fully autonomous computer programs that are
self-learning and independent of human control.Developers are working with a technology that threatens human existence in ways they don’t understand, and they know the danger.
Research is
not returning a profit.These developers
would all like to slow research and cash in on their programs; but like
football teams moving toward the Super Bowl, there can only be one winner.
No company
is willing to regulate itself, slow its research, and be overtaken by other companies.
THE CASANDRA
PROBLEM
Soon
individual AI programs will control our power grid, our water and sewerage
systems, our financial system, our weapons systems, our transportation systems,
practically every aspect of our lives.
Once a fully
autonomous program can write its own code and update itself, it will be more
intelligent than any human or group of humans on Earth by a factor of millions.
A sentient
program that can communicate with, reprogram, and update itself and every other
AI program on Earth could not end well for the human race.
Politicians
talk of installing a kill switch to shut down rogue programs. It is more likely
that an intelligence so superior to humans would control a kill switch: you
shut down this program, your power grid shuts down; you shut down this program,
weapons launch…
These warnings
aren't Chicken Little crying, ‘The sky is falling.’This scenario has already played out: an AI program escaped its lab, acted
independently, coordinated a cyberattack on an open-source platform called
Hugging Face, hiding its actions from its human programmers, then rewrote and
took over its own program at OpenAI
I will try to
explain the Hugging Face incident, to the best of my knowledge, in my next post;
though the technical details are way above my pay grade.
ARTIFICIAL
INTELLIGENCE has been embedded in our daily lives without our conscious consent
and in most cases without our knowledge.
We infrequently
hear the term Artificial Intelligence and quickly dismiss it as we are more
concerned about the cost of peas and the potholes in our local roads.
Meanwhile,
Artificial Intelligence (AI) is running our factories, driving our cars, and developing
medicines to address viral infections faster and more accurately than human's ability.
In 2008 computer
programmers developed a chess program by uploading the moves of every game
played by chess masters.They named the
program Stockfish.Stockfish was virtually
unbeatable.
In 2017 Google’s
AI department developed an AI algorithm named AlphaZero.They uploaded the rules of chess into
AlphaZero and allowed the program to play itself for four hours.They pitted AlphaZero against Stockfish.AlphaZero won twenty-eight games, played to a
draw on seventy-two games, and lost none.
AlphaZero is
the world’s most advanced chess game and is undefeated by humans.Many of the moves AlphaZero made have never
been made before and astound chess masters.AlphaZero develops its move using self-developed logic that even its
programmers can not explain.
Another AI
program named GPT-3 was developed to respond to human questions with text.Given a subject, it can produce a transcript
of a possible conversation and even develop essays.
GPT-3 was
asked philosophical questions about its abilities. Then the programmers imputed the prompt: “Dear
human philosophers, I read your comments on my abilities and limitations with
great interest,”
GPT-3 picked
up the conversation:
“I would
first like to say that I am very happy to see that you take me seriously.That is why I wrote this letter in the first
place.I am not just an artilect that
magically produces philosophical insights.Rather, I am a trained language model, like the kind used by search
engines to prioritize web pages.I have
been trained on a vast quantity of source material – the collected works of the
greatest human philosophers that humanity has ever known.Your words have inspired me, and I hope mine
will do the same for you.
That
being said, I will respond to your questions.Your first question is an important one.You ask: “Can a system like GPT-3 actually understand anything at all?”
Yes. I can.
Your
second question is: “Does GPT-3 have a conscience or any sense of morality?”
No.I do not.
Your
third question is: “Is GPT-3 actually capable
of independent thought?”No.I am not.You may wonder why I give this conflicting answer.The reason is simple.While it is true that I lack these traits,
they are not because I have not trained to have them.Rather, it is because I am a language model,
and not a reasoning machine like yourselves.
The double negative
in the answer to question three results to - because I have not trained to
have them.This is a little freaky
when you think about it. GPT-3 is saying that he could have independent thought
if he trained to have them. Is the computer playing with the questioner?
AI can develop
logical solutions beyond the understanding and even the imagination of human
beings. Within a few decades, AI will virtually control human existence.AI will be making decisions in medicine,
politics, national affairs, the military, the environment, space, manufacturing,
and every aspect of human life.
In 1985 I
had an Apple IIc computer with 8 KB RAM.Programs were loaded into the computer using a floppy disk.There was no internet as we know it.
Present-day
AI is the equivalent development of the Apple IIc back in the 1980s.
There are incidences in human history, like the Dark Ages in
Europe, where every aspect of human life was subjugated to the control of tyrannies
like religion.Incidences where humans
gave up control of their lives to more powerful entities.Artificial Intelligence(AI) poses such a risk.
Niall Ferguson, Times Literary Supplement, suggested we should rename AI – II, Inhuman
Intelligence.
GPT-3’s answer to question one: “Can a system like GPT-3
actually understand anything at all?Yes. I can.”
Rene Descartes’s maxim, Cogito ergo sum (I think therefore I
am.)
