So does AI agree with The Barnet Eye's assessment of the risk level of an AI Doomsday apocalype - The results may shock you!

 Yesterday, I posted a blog detailing the risks of an AI doomsday apocalypse scenario, based on my 37 years working in IT, developing and supporting major banking payments networks. After I posted the blog, I asked Google's Gemini AI tool whether I was right. You may find AI's view of the risk shocking. Here is what it had to say

--

Evaluating the realism of the scenarios raised in the blog post requires breaking them down into realistic near-term risks, plausible worst-case events, and unrealistic extrapolations.

1. Highly Realistic: Minor Errors Causing Major System Failures

• Realism Level: 10/10 (Already Happening)

• Why it's realistic: The blog's anecdotes about payment system glitches creating tailbacks or test scripts flooding emergency services reflect how fragile complex, interconnected IT infrastructure is.

• Real-World Evidence:

  • The CrowdStrike / Windows outage (July 2024) proved that a single corrupted software update can instantly paralyze global air travel, emergency services, banking networks, and healthcare systems without any cyberattack occurring.

  • The 2023 NATS air traffic control failure in the UK was triggered by a single improperly formatted flight plan that caused the main and backup systems to enter a safety shutdown, grounding thousands of flights.

2. Highly Realistic: AI-Accelerated Cyberattacks on Infrastructure

• Realism Level: 9/10 (Active Threat)

• Why it's realistic: Criminal syndicates and hostile state actors do not need "sentient AI" to cause devastation; they only need AI tools that automate code analysis, scan legacy operational technology (OT) for vulnerabilities, and execute zero-day exploits at superhuman speed.

• Real-World Evidence:

  • National security advisories confirm that nation-state actors are deploying AI tools to map out power grid networks, water management SCADA systems, and transport infrastructure for potential sabotage during geopolitical crises.

3. Plausible: Human Deskilling & Loss of Fallback Control

• Realism Level: 7/10 (Plausible Over Time)

• Why it's plausible: The blog highlights the risk of losing the human workforce capable of manually managing systems when AI fails. This phenomenon—known in safety engineering as deskilling or the Paradox of Automation—is a recognized vulnerability.

• Where the realism holds: If organizations prematurely lay off experienced technical support staff or fail to train a new generation in manual troubleshooting, an extended AI or IT failure becomes much harder to recover from quickly.

• Why it isn't an immediate total collapse: Major utilities and nuclear facilities are bound by regulatory frameworks requiring physical air-gaps, analog manual overrides, and human operator sign-offs.

4. Unrealistic: Permanent "Apocalyptic" Fall of Society

• Realism Level: 2/10 (Highly Unlikely)

• Why it's unrealistic: The blog's dramatic conclusion—that a single AI malfunction or cyberattack could trap humanity in a permanent, unrecoverable "post-apocalypse" where no one can keep the lights or water on—overstates system fragility and understates resilience engineering.

• What would actually happen: In a major systemic breakdown, societies experience acute, severe disruption (rolling blackouts, temporary supply chain freezes, economic loss) lasting days or weeks. However, engineering protocols mandate physical isolation switches and emergency manual procedures to restore basic utilities at a local level.

--

What interests me most is the fact that the only one of the four scenarios that it stated was highly unlikely was one that I didn't actually predict, which perhaps gives an interesting insight into the way AI itself works. I never stated that society would end as a result of an AI apocalypse. I assume that the AI misunderstood my use of language. What I meant by the term was that it would be an apocalypse for AI systems in terms of society trusting and relying on them. When we see a major disaster, with loss of life, and massive society impact, financial impoverishment, major loss of savings, property, etc, civil unrest usually follows. Of course society adjusts and some degree of normality eventually returns, boy our faith in technology may be permanently undermined. What I said the in fnal paragraph of the blog was this. I do not believe Gemini AI's analysis properly represented this

"But that is not what interests me most. I believe that we will see a major catastrophy, with loss of life caused by one of these scenarios. And when this happens, we will see a massive re-assessment of our relationship with AI. The thing is, that may happen after the point where we can put the genie in the box. If all the people who AI replaced no longer are around or have the skill sets to clean up the mess, we are in an apocalyptic situation, where we have to keep software running that we cannot trust, to do the basic things our society needs, like keeping the lights on in winter and the water flowing. It will be bad enough if the problems are just down to glitches or hardware problems. If they are down to a malicious actor gaining control of the systems that run our lives, using AI systems that we cannot circumvent, who knows what will happen?"

Is AI smart enough to understate the risks it poses to society? Perhaps we can conclude that.