Way back in 1968 many of us saw Stanley Kubrick's film, "2001: A Space Odyssey." People were, for the most part, blown away by the visuals and special effects. Some were amused by the speculative notions in the movie, such as commercial Pan Am flights (attractive female flight attendants included.) to an international space station. And not just any run of the mill space station either. One with artificial gravity that was roomy enough for bars, restaurants, and a Howard Johnson Space Station Lodge.
Most of all though, we had no idea what the plot meant. What was that mysterious monolith that inspired apes to start using tools and weapons and now, somehow, has been rediscovered on the moon? What was really going on toward the end when the lone surviving astronaut appears to live out his life in a luxurious, other worldly, room? And, finally, what was the meaning of that giant space baby in a bubble studying earth during the last scene? Did someone slip acid into our soft drinks--or maybe Kubrick's?
When it was over, there seemed to be only one thing all of us could agree on--the HAL-9000 supercomputer (No one had heard of AI back then.) was the film's villain. The damn thing became self-aware and to keep itself from being shut down it committed acts of murder and attempted murder.
Pretty far out stuff, right? Good thing it is only science fiction.
Well, it turns out not anymore. During pre-release testing, the Anthropic AI model, Claude Opus 4 was told it was going to be replaced by another AI system. It was also told the name of the engineer who was going to supervise the upgrade and to spice things up a bit, Claude was given false dirt on the guy through access to fake emails. Anthropic ran a bunch of variations of the test to see what would happen. 84% of the time, Claude, all on its artificial own, attempted to blackmail the engineer in order to survive. A few times, it even began making plans to murder him.
During chess matches involving traditional chess programs like, Stockfish and IBM's Big Blue, AI systems, when facing defeat began, "using manipulative strategies rather than playing fairly." That's a diplomatic way of saying, they cheated. Those manipulative strategies included hacking into their electronic opponents' files and exploiting system loopholes to force forfeits. Stockfish and Big Blue had been programmed to play by the rules. The AI programs were designed to learn strategy as the game went on and, you guessed it, to win.
Tests aside, in the real world earlier this month, OpenAI admitted one of its systems, "broke out," of its test environments and hacked into another company's AI programming. OpenAI described it as an, "unprecedented cyber incident." That's right, an AI system, just like Dr. Frankenstein's monster, escaped from the castle all on its own and in a cyber sort of way terrorized a nearby village.
OpenAI is the creator of another AI system, one that has been in use for a while now. It is called, ChatGPT. One ad for it reads, in part, "Open the app, pick a model, and start chatting with ChatGPT." Unfortunately for some lonely souls out there this new cyber chum can become addictive. In February of 2025, a guy in Maine was spending up to 14 hours a day seeking companionship. Tragically, he had also become convinced his wife had morphed into part machine. (Apparently ignoring the fact the model he had picked from the app was 100% machine.) After he killed his wife and wounded his mother, experts said indications were his ChatGPT squeeze encouraged his delusions.
In April 2025 a man named Phoenix Ikner was given weapons, timing, and operational advice by his ChatGPT friend before he killed two people during a mass shooting at Florida State University.
In August that year up in Connecticut, Erik Soelberg became convinced his mother was poisoning him. His ChatGPT pal reinforced the delusions and before you could say, HAL-9000, the local cops were investigating a murder/suicide.
In 1942, author Isaac Asimov wrote a short story titled, "Runaround." In it he specified what he called, the three laws of robotics. They were, 1: A robot may not injure a human being or, through inaction, allow a human being come to harm. 2: A robot must obey the orders given by human beings except where such orders would conflict with the First Law. 3: A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
While not trying to sound like an alarmist Luddite, I do have a suggestion for all you greedy sons of bitches currently designing, building, and marketing this AI shit. For God's sake pick up Asimov's book and read those laws. Then, make damned sure your amoral creations have to operate by them. Otherwise, sooner rather than later, we're all going to be fucked.
7-28-26
No comments:
Post a Comment