Re: AI and creativity.
Reply #51 –
I had the weirdest experience with packet transfer this week at work. We have some SSL VPN tunnels which sends encrypted data from our on premises Firewall to our cloud provider. Effectively, the traffic to and from premises is encrypted and when data is transferred its broken up into ESP packets. These tunnels once established a fairly bullet proof and only re-authenticate without dropping. For about a week we started having an issue where we couldnt contact the cloud through that link, and ended up having to drop the tunnels and bring them back up to fix it. 20 minutes of down time each time, nothing too serious but then an investigation into root cause. The fact they re-established after dropping means configuration wasnt the issue.
Now these packets are encrypted, and the way it works is when you have large amounts of data, its broken up into smaller amounts, encrypted, and then sent down this tunnel via an internet connection, which will arrive the other side, and be re-assembled in order. The tunnel is a public connection with a public and private key pair. This is complex but not a big deal, nothing will usually go wrong there.
We ran some analysis, could see capture a log of traffic being sent by our firewall, and the cloud provider was not receiving them. After a couple of days of back and forth trying to work out where the break down was, and also contacting our internet service provider, we setup a sniffer in the middle that would intercept the packets and relay them down the line and upon checking those logs realised that everything fell over at the first hope. Firewall outbound seemed to not be happening even though it told us it was.
We have two of them in an active passive situation which will ensure if we lose one, stuff doesnt ever break. We failed over initially and it worked instantly (same config, same connections) and then failed back, and issue was present. So the (secondary)passive firewall was fine, and the active (primary) one started having the issue was resolved by simply turning it off, and turning it on again.
We are talking about a device, running a linux operating system, which has one job, that stopped doing its job and is usually the most robust component of the network, have a big issue. Now, you could argue that AI will re-write the code to prevent this from happening, but the problem is that everything looked fine, and actually wasn't.
THIS is why I don't think we will ever get to a point, where technology can do things properly without simply running commands, because even when we tell them to do something, they have a weird problem where they refuse to do the job they are employed for properly and a simply troubleshoot of drop and re-authenticate only worked for temporary restoration of services before we experienced another issue. Get these things with AI, and Im not sure how its supposed to analyse whats happening beyond its front door, when we had to implement more devices to actually tell us this information.