It’s a little silly that this needs to be said, but many of you clearly believe AI can break the fundamentals.
Information decays. This enemy is known by many names: bit-rot, data staleness, schema drift. I prefer entropy, or accumulating uncertainty over time. All information is under siege by entropy.
You generate a wonderful report. 3 years from now, will the organization remember what it means? Will you be confident it captured truth from every relevant source? Can you trace where it came from?
“We will solve this by attaching more metadata to everything we produce! Now every report includes a standard audit trail including timestamps and links to sources.” Great, now you have metadata that is also subject to entropy: has the schema since changed? Have links become broken? Even timestamps desync and drift. Perhaps you need meta-metadata that shows exactly which NTP server the timestamp came from! Even if every bit in the report remains unchanged and you have the hashes to prove it, context will change the meaning of those bits over time. Entropy does not give up.
All information is under siege by entropy, and all information work is a battle to hold the line for as long as possible. The more information you have, the more surface area for entropy to attack. This is why software architects spend months squabbling over the tiniest data model change; the wrong choice becomes a hole in the front which entropy will surge through. When that happens, people quickly forget what was lost. “This metadata means nothing to me, let’s ignore it.” Organizational amnesia sets in and that ground is ceded completely.
This is why I’m alarmed as AI is applied to everything, generating as much information as quickly as possible, with little effort to firewall it from the things we need to be most certain about. It is a machine that subtracts meaning while exploding volume, burning through resources budgeted to fight entropy. Emails are ballooned into listicles rotting in inboxes, generated docs decay on SharePoint, vibe-coded fields inflate our data models, and so on.
We were already losing this war when the giant wooden horse showed up at the gate. “Wow this looks really useful”, we said to ourselves, “Let’s bring it in and apply it to everything.”
You will not hit some critical mass of AI-use that suddenly reverses entropy. You cannot “ChatGPT please summarize” the meaning back into the text. You cannot defeat entropy with AI.