AI is tuned to speak confidently and ship independently; it is rare that it will suggest that one go back to a broad survey of stakeholders to ensure no Chesterton's Fence is being torn down (https://fs.blog/chestertons-fence/). "I've read the documentation and logs, and nothing suggests this system needs these legacy capabilities/components/capacity" is something we hear all day, and it is a siren's song if there ever was one.
Couple that with a general (and valid!) consensus that modernization of legacy systems must be accelerated for a variety of reasons - to mitigate this new and continuing "Black October" of hyper-agile agent-driven cyberattacks (whether the network is public or private); to have more flexible infrastructure for new features to stay ahead; to allow human teams to show their efficacy in a world where technology jobs are increasingly at risk - and you have a perfect storm where change management is no longer what it once was.
I suspect the situation is that AI's caveats are just easier to ignore than a person's warnings. Plus all the other factors you list. Also, AI tends to give a person the kind of communication they ask for. I always ask for options but I suspect other people ask for just answer and so they get them.
I think you went off somewhere else.
Which is to say, a parent post here on HN doesn't narrow the topic to it's focus. That's what makes discussion here interesting.
The FAA's recent outage for example, at its core, seems like people trying to rush needed upgrades that were slow for a specific reason (e.g. reducing the testing window for 90 to 30-days).
The Microsoft outage seems like three different things linked together (and poor "full failure" recovery prep, lengthening the outage).
The bank "outage" may not have even existed, 2/3 of the banks claimed they didn't have an outage at all, and the other said it was a limited technical issue. The articles were based on DownDetector spikes.
This has reached a Witch-hunt levels of fever pitch.
So let me see if I understand what I should be:
- Be more skeptical of official explanations, including where it includes a technical breakdown of how it occurred (e.g. Microsoft/Outlook outage).
- Be less skeptical of random commenters claiming, without proof, that it is AI or Cyber Attacks.
Because that's the implied point you're making. Be more skeptical of A and less skeptical of B, even though A has someone placing their reputation behind it, and B is anonymous commenters throwing made up theories out into the ether.
The explanation for this one still doesn't ring true for me and that's exacerbated by how many times the story has changed after telling the public that they didn't need to know the reasons.
Agree, though, that the rest do have reasonable explanations
Is "Claude says it looks good so let's skip the usual testing window and push to prod on a Friday" really not a sentence you can imagine being said in this current environment?
My point is that people are taking a potential future problem and applying to actual things that have recently happened, that didn't fit that mold. Which is incorrect.
When AI or Cyber Attacks cause an outage, people are absolutely right to say so. But claiming - without proof, that every outage is related to the two is noise.
(This of course is on top of whatever unusual things might actually be happening)
https://www.bleepingcomputer.com/news/security/optimizely-su...
Is it possible that the false of productivity is leading to worse products and massive loss in valuable headcount? Instead of AIs directly writing bad buggy code? Or is "everything rage-bait, and AI is all sunshine"
There's a group of people who think AI is perfect and will solve all of humanity's problems. That's annoying. There's another group of people who think AI is the worst thing to ever happen to humanity and is the downfall of society. These people are more vocal in my experience and they're everywhere. If you ever use AI you're a bad person who they'll disown like you're racist. It's somehow become the most popular 'social justice' issue of our time even as ICE exists alongside it. These people see a problem and immediately try to figure out how AI is involved. That's far more annoying.
Best example I can give that came up in my circle recently is the video game "Big Walk". It was recently denied for a silly award nomination because the awards have a strict no-AI policy. Which to me is absolutely fine, go nuts. But the vitriol towards this game, that everybody absolutely ADORED before the developer disclosed that they fixed ~10 lines (!) of code with AI because they couldn't figure it out otherwise, is absolutely bonkers. People I respect treated it like they were mourning something that WAS great until this catastrophe happened.
It would be interesting if someone could do a proper analysis, though!
It would hardly be shocking that hostile nations are able to cause some problems with Mythos level tools.
If anything, it is more interesting how we can invade another country and the idea that the country or a collection of hostile countries could hit us back in any way, shape or form is just an impossibility. An insane conspiracy theory.
Maybe it is not that but surely the probablity is not zero.
I can't put my finger on it, but something sinister feels afoot.
https://beready.campaign.gov.uk/
It's nowhere near the kind of readiness and preparedness campaigns of the cold-war of course, but it's a step up in alertness that's higher than anything I can remember.
Too many people are addicted to their smartphones and don't feel the same level of pride they used to in their work. All systems are shared, instead of one person being the ruler of a certain part of the stack, it's a revolving door and it's rare that anyone stays somewhere for even 5-10yrs. It leaves a wake of modifications that nobody else can keep up with. Also, the tech industry is moving fast just to move fast, a lot of old infra is still C++ and COBOL but good luck hiring anyone to use that, so they have to convert the systems to use the Latest Tech of the Day. The old, stable people don't have any interest in learning the new thing, and would rather retire with their knowledge. The new people want to rebuild from scratch in the New Thing and are convinced they need to job hop immediately after.
[To be clear: I mean exactly what I said, which is "there could be" which is not the same as "I think there is" -- so call it a 5%-10% chance of real cyberwarfare -- which means I'm not dismissing those claims out of hand, and find the claims at least plausible, even if I don't currently believe them]
It’s not nothing
But my gut says incompetence and rushing upgrades out of fear of cyberwarfare, not an actual attack.
Though at the same time, the public isn't privy to even 10% of the incidents going on right now.
The information I have is consistent with general cybercrime and script kiddie patterns, just a much higher volume, and much larger diversity of targets.
You have aides and resources and money. Not having enough information or understanding is a choice.
It's criminal negligence.
I also think generally that AI is probably both the cause and solution to these problems. We're certainly getting more code churn than otherwise, and in lots of cases that's really good! But you're going to have to deal with inevitable instability too.
I'm sure most of us here have been on the receiving end of managers suddenly obsessed with a failure mode because it happened to occur twice in one month.
I think the real ones were the Microsoft and FAA due to deploying the AI slop.
The other ones seem to be usual outages that normally no one reports on.
So certainly there is a big social-psychological phenomenon. Or it's gonna be a cold winter! :)
But I'd note, these "named after a thing" phenomena (eponnyms) are kind of terrible for exact description since their definition tends to be fuzzy, there's a temptation to fudge their borders and they create an insider jargon (if only I could a device that could reach of the computer and slap everyone who says "steelman" when they mean strengthen an argument, etc).
The combination is a cycle
1. AI finds bugs
2. Engineers us AI to try and fix
3. AI testing approves, minimal engineering review
4. Engineers push release rapidly, it's critical after all
5. Bugs introduced
6. Cycle starts again
With all the recent asshattery of leaking kernel exploits, I would presume trust is gone, and it's cover the big boys first.
(a linux kernel bug vulnerability could be patched without lkml involvement or distros even knowing)
I imagine, for the paranoid spy, not being able to call home could also mean "the enemy has severed our comms, we've been compromised, burn everything and run!".
Coincidence or is there some common explanation that I missed?
I think the opportunity right now where we have powerful AI hacking systems but not enough protections is to establish long-term presence in all networks of interest. Not bring them down.
For example, we updated a GCP managed base image of Ubuntu that was released with 350 CVEs patched, some were as high as 9.8. In doing so, they (GCP) changed the DEFAULT setting for hardware GRO from off by default to on. That's a massive networking change packed into a security update.
We only caught it AFTER it started causing issues in production. We didn't have the OS pinned because (decently valid reasons) or catch it in non-prod (decent reason + human mistake).
This is on a modern highly automated tested resilient well architected platform.
I can only imagine these other platforms are dramatically more exposed that we are, hence these common large scale outages.
More plausible to me is that spy agencies had previously found and used some of these hundreds or thousands of vulnerabilities to hack into those networks just fine. But now other people like cyber criminals are using these vulnerabilities via AI. That is messing up both the attackers who had already gotten in quietly and do not want the attention and the defenders who now have to respond to a noisy intrusion they cannot ignore.
Outages could be grey-hat testers and emergency updates.
A temporary period like we are experiencing is to be expected when moving to a time where a single engineer can perform the equivalent work of a team of excellent engineers working for weeks. Security has always been a game of who has the capabilities, means, and requirement for any given attack surface.
Like the United Airlines thing yesterday, the moment I got to the article it said "was down for a few hours" but the post was still just something about United Airlines network being offline