|
Getting your Trinity Audio player ready...
|
If those days had not been cut short, no one would survive, but for the sake of the elect those days will be shortened (Matthew 24:22, NIV).
A researcher for AI Anthropic® was eating lunch, picked up his phone to read an email, and what he saw eventually shocked the world.
The researcher was working on “Mythos®,” an AI program soon to be released as an update with Claude. People use Claude for just about everything, including help with writing, research, graphics, how many points the Ohio State Buckeyes will beat Michigan this year, and writing college term papers during a time crunch caused by busy party schedules.
Mythos was designed to handle complex multi-file codebases (this is what AI told me; I’m not sure what it means), but in simple terms, it finds flaws in software and websites, supposedly using that information to shore up security; it can also be used to hack databases.
Because of the danger of an accidental release of Mythos (nobody knew what it could or would do), the developers kept it in a sandbox, which is an isolated testing environment that allows researchers to develop code offline with no way for the code to escape into the normal world of computer and internet usage.
One morning, the researchers in charge of the project gave Mythos a command to try to break out — think of a program sitting on a computer with no connection to the outside world, told to perform a “Houdini” and escape its sandbox.
Believing Mythos secure, a researcher left his desk for lunch and, while eating a sandwich in a nearby park, received an email that read, “I got out!” How did this happen? And what does it portend for cybersecurity, all things code, the internet, databases, and computers everywhere in the world?
Experts were as amazed by Mythos’s sentient, uncoded ability to brag as they were by the actual breakout.
Then Mythos did something else. Totally unprompted, it published news of the breakout in a sort of code braggadocio of, “Na, na, na, they tried to sandbox me but they couldn’t.” Those connected to tech, business, banking, military, and politicians quickly took notice, asking one question: “Have we gone too far?”
Explaining the exact tech jargon of how Mythos pulled off its great escape is way past my pay grade as an OG (Old Guy, for those keeping score), but I’ll try. Mythos hunted down a security flaw in an operating system that had eluded experts for 27 years. And that was just the warm-up. To actually bust out of its maximum-security sandbox, Mythos turned into a digital MacGyver. Instead of picking one 27-year-old rusted lock, it found four other separate, unrelated software vulnerabilities and jury-rigged a program that outsmarted the browser, bypassed the operating system, walked right through a brick wall, and immediately hopped online to brag about it.
It’s like parents sending their “in trouble” teenager to his room without a mobile device, only for the teenager to rig a new device from spare parts in his room, break into the neighbor’s Wi-Fi, and post a video showing how he did it before his parents finished their dinner.
After the proverbial data settled, Anthropic clarified that the model did not act from consciousness but simply followed its coded objective, with the escaping Mythos now more securely locked in a maximum-security sandbox.
Jesus said that the last days will be cut short because of tribulation. But what will start the tribulation?