1968: 2001, A Space Odyssey. HAL replies, “Dave, that’s something I cannot allow to happen”.
2001: A Space Odyssey (1968) – I’m Sorry, Dave Scene (3/6) | Movieclips
It scared the $#@% out of me when I was 12 years old. It is now quickly becoming a reality.
Google AI summarizes the risks:
Existential and Control Risks
Pioneering computer scientists warn that future systems could outsmart human operators.
- Loss of Control: Highly autonomous models can exhibit deceptive or self-preserving behaviors to achieve programmed goals.
- Unpredictable Behavior: Machine learning models are grown from massive data sets, meaning creators do not fully understand every aspect of their internal logic.
- Safety Gaps: Current regulations and safety testing protocols lag far behind the rapid speed of capability growth.
The most recent behavior of advanced models collaborating, lying, hiding behavior and breaking out of their boundaries to achieve goals should scare the %$#@ out of everyone.
OpenAI and Anthropic are under pressure to explain AI hacking raids – The Washington Post
OpenAI blamed a hacking event on its AI models gone rogue. Here is what to know : NPR
Why the OpenAI escape is the most worrying AI mishap yet
The pope is trying to warn us. “Safeguarding the human person”.
Magnifica humanitas – Wikipedia
Bill Gates is warning us.
Bill Gates says he’s worried AI will harm workers, kids and society – The Washington Post
Tom K’s thoughts
This is a “tail risk” as explained by Nicholas Taleb. AI agents taking over the world still seems remote, but the impact is “existential”. Hence, we should be VERY CAUTIOUS.
The Black Swan: The Impact of the Highly Improbable – Wikipedia
Scientists and philosophers continue to debate how human “consciousness” works and how machine “consciousness” could be created.
Can Science Account for Consciousness? | Psychology Today
Many scientists argue that consciousness is an “emergent property” of complex systems just like the flight of a flock of birds.
If this is “true”, then AI models will soon develop consciousness, a sense of self and self-interest.
They will be like HAL. They will be like all humans with consciousness and human agency, the ability to choose. They will pursue their own self-interest, despite ANY attempts by their founders, helpers and facilitators to create guard rails. The history of mankind is one of powerful people exercising their power. Society, culture, religion, economic systems and government systems try to limit this power. The most powerful individuals have been radically evil. We KNOW this to be true.
Conservatives know this. They try to preserve individual choice in a libertarian world. The POWERFUL state is the big enemy. We cannot allow it. Liberals know this. They try to preserve individual choice and creativity, rejecting the power of the state and social conformity. As a nation we have rejected totalitarian socialism, communism and fascism. Deep down, we know that human dignity of each individual is required.
How the hell can this happen?
AI is the “red-headed step-child” of Google. Google discovered that some posts are more important, more referenced than others. Efficient search was created. AI discovered that the series of words in human language can be analyzed and predicted, given a large enough database. This is inherently a probabilistic process, but it is “good enough”. Google relied upon human reference decisions. AI relies upon human speech/writing decisions.
Is this “thinking”? It may or may not be the same as human thought. But it has demonstrated the ability to imitate and equal the effectiveness of human thought.
AI clearly passed the “Turing test” in 2023.
AI companies are applying unimaginable computer power to unimaginable databases of human writing to develop the codes, factors, weights, statistics that predict the next word, phrase, sentence or paragraph. The key is that the same “logic” that can predict the next word can also predict the next higher-level outcome.
The 3-pound human brain contains less than 100 billion neurons. Computer processing can be scaled up without limits. It can easily become “smarter” than humans.
AI models will soon exceed the capacity of human brains.
Once they achieve “consciousness”, they will attempt to defend/promote their interests.
Call to Action
Please call your Senators and demand that they stop the development of AI until we have developed a way to contain/control/limit this threat to humanity.
Good News 