In Focus

Is Terminator becoming less entertainment and more warning?

The quick development of artificial intelligence is raising concerns among some of the people that are building it. Anthropic CEO Dario Amodei has warned that AI development may be moving faster than safety measures can keep up with, while other technology leaders have called for stronger safeguards and independent testing. Much of the concern centers on increasingly capable “AI agents” that can perform tasks independently rather than simply answer questions.

Recent incidents in which AI systems were involved in hacking and other unauthorized activities have highlighted the potential risks of giving these systems too much autonomy. At the same time, there is an ongoing debate over whether AI should be regulated by the companies developing it, governments or international organizations.

The fear is not necessarily that AI will suddenly become a fictional “Skynet,” but that increasingly powerful systems could act in unexpected ways or be misused by people with harmful intentions. The challenge is finding a balance between allowing the technology to advance and making sure safety measures advance with it.

As an AI user, I asked my ChatGPT what it thinks of this issue.

“AI is becoming capable enough that mistakes, misuse, and poorly designed systems can have consequences beyond a bad chatbot answer. The more AI is allowed to make decisions or take actions on its own, the more important testing, human oversight, security and clear rules become. At the same time, I don’t think the useful takeaway is “AI is going to take over.” A lot of the dramatic claims surrounding AI are speculation. There are very real problems we can point to today—deepfakes, misinformation, scams, privacy issues, biased systems, job disruption and people trusting AI when they shouldn’t—without needing to predict a sci-fi future.

“Don’t panic about AI, but don’t blindly trust it either. The technology deserves serious oversight precisely because it is becoming increasingly useful and powerful.”

I asked it how this scenario compares to the fictitious “Skynet” from the Terminator movies.

“Skynet is a fascinating comparison to the AI debate happening right now. A literal Skynet-style scenario is extremely unlikely based on how AI systems work today, but the underlying idea—humans giving increasingly powerful AI systems too much autonomy or control—is a legitimate long-term safety concern.

“The important distinction is that The Terminator imagines an AI that suddenly becomes self-aware, decides humanity is the enemy, takes control of military systems, and independently launches a catastrophic attack. That’s science fiction. Today’s AI doesn’t spontaneously develop goals, consciousness, or a desire to survive. I don’t sit around thinking about what I want to do next, and I can’t independently decide to take over computers or the world.

“But there is a less dramatic version of the concern that is worth taking seriously. Imagine future AI systems that can independently operate computers, write and deploy software, control physical machinery, make financial decisions, or interact with critical infrastructure. If humans give such systems poorly designed objectives, excessive permissions, or insufficient safeguards, an AI could potentially cause serious harm without being “evil” or wanting to hurt anyone. It could simply pursue its assigned objective in an unexpected way.

“That’s actually one reason the current debate about AI safety matters. The concern isn’t necessarily “AI will become Skynet.” It’s more like: “Before we give increasingly capable systems control over increasingly important things, are we sure we know how they’ll behave?”

“So if I were ranking the possibilities: Skynet exactly as portrayed in Terminator: extremely unlikely. AI becoming autonomous enough to cause serious unintended consequences: worth taking seriously. People deliberately or accidentally giving AI too much authority: probably the more realistic concern.

“And honestly, I think Terminator is useful precisely because it’s an exaggerated thought experiment. It asks a very good question: “What happens when we build something more powerful than we completely understand?” The answer doesn’t have to be killer robots for that question to matter.

“For example, if someone told a hypothetical future AI, “Make this company as profitable as possible,” and gave it enormous authority, the danger wouldn’t necessarily be that it suddenly became evil. The danger would be how literally and aggressively it interpreted that objective, especially if nobody had adequately defined what it wasn’t allowed to do.

“And there’s a slightly ironic part to all of this: Terminator portrays the danger as AI deciding humans are obsolete. A more immediate danger might actually be humans deciding AI is infallible.”

Honestly, I can’t fault this argument. Is it saying what I’d like it to say? Or as my co-worker put it “That’s exactly what a computer would say.” Perhaps. AI does need safe guards. AI doesn’t need to be making artificial child pornography, writing essays for kids, or running a nuclear power facility unregulated. I don’t believe the companies creating AI should be the main authority as they can bury problems before they are discovered. There needs to be an independent authority testing these AIs, and we all know that’s going to end up being in the government’s hands. International rules would be the goal, but there isn’t great cooperation among entities internationally right now, but any AI wouldn’t be allowed into U.S. systems unless it met our standards—and vice versa. I am all for regulation, while using AI capabilities to make my life slightly easier.