AI will kill us all, or at least our brains
I was going to do a post on Vault to follow up on my the previous one but there’s a course on Udemy for anyone that’s interested .
However, I’ve been continuing with the API theme; using Python scripts to migrate an Cisco ASA to an on-box FDM managed Cisco FTD. If you migrate from a Cisco ASA to an FMC managed FTD then Cisco’s migration tool will do it for you, but with the FDM management stream you’re on your own. (FDM isn’t bad but I recommend using FMC where you can).
The previous time I’d used Python it was to migrate a configuration from a fairly obscure firewall vendor together with a client-supplied spreadsheet, to a series of files I could use to import objects and rules into a FortiGate. To do this, I read this book about Python:
https://automatetheboringstuff.com
… did a bit of googling. The scripts were pretty elementary but did what I needed them to, and it was a fairly satisfying experience which also improved my knowledge of Python a great deal.
This time with my ASA to FDM/FTD migration, I decided I’d use AI to get me started and then tweak the result. I went to a well known AI portal put in a prompt and quickly got a useful python API script back. When it didn’t work first time, I edited the file, went back to the book above and worked out what was going wrong. Of course, I didn’t: I fed the results back into the AI portal until it fixed it. And it’s probably no longer a surprise to anyone to say that fix it, it did. I could have probably got there myself in maybe four to five times the time it took the AI but then I guess that’s the point – the much-vaunted productivity boost in this instance was very evident to me. But I hadn’t learnt anything. I could more or less work out what the script was doing but I felt like one of those student linguists who say they can understand a language but not speak it.
I guess this makes AI the latest in a long-line of cognitive offloading: reading and writing meant people didn’t need to remember things, and calculators meant people didn’t need to do mental arithmetic. But the thing is there’s no going back: you can’t ignore the time savings that AI is bringing. Even if I had written the scripts myself, I don’t use Python frequently enough to feel that the knowledge would have stayed with me. But I think I would have more of a dilemma if programming or scripting were the main focus of my job. I feel I’d have to choose between consciously deskilling myself or battling away crafting my programmes ‘manually’ and accepting I’d probably be slower. Having said this, my use case was very simple, had easily verifiable outcomes and was essentially a back end task. Would I want to submit reams of code I barely understood that had to do something important in a Production environment? Probably not. But would I want to lag behind my colleagues who were? Probably not.
Mind you it could be worse. Quite how much worse is the subject of this delightfully titled tome from two AI researchers:
One cannot accuse the authors of not taking a position. Their core message is, and I quote
“If any company or group, anywhere on the planet, builds an artificial superintelligence using anything remotely like current techniques, based on anything remotely like the present understanding of AI, then everyone, everywhere on Earth, will die.“