I have almost 30 years of professional experience as a sysadmin.

Failure is inevitable. Hardware will fail, software will be discovered to have flaws, reconfiguration will be mistaken.

The antidote is to have reliable recovery mechanisms: everything will eventually break or need to be changed, so in turn you need to be able to recover back to a stable position, so you can try again.

Recovery mechanisms are not all-purpose. I usually describe them in three phases:

I had thought that ‘pareidolia’ referred exclusively to the human tendency to see human faces everywhere – any time you see two dots or circles and a line below them, boom :) it’s a face. It can be triggered by rocks, clouds, shadows, uneven browning on a tortilla or a pancake.

It turns out, though, that pareidolia covers any kind of overactive pattern matching. Humans look for meaningful patterns all the time, without thinking about the search, and when something trips the right thresholds in our brain, we seize upon it. The result can be faces, voices, music or just a cloud that looks like a moose.

Here’s my thesis: the large language models (LLMs) and similar generative software processes that people currently refer to as “AI” – those are exceptionally well-tuned to produce patterns that will trigger threshold recognition in human brains. Not by accident – that’s the goal. They are pareidolia-circuit stimulators.

The problem is that LLM output is exactly as meaningful as any other kind of pareidolia. Every well-formed paragraph is a false-positive error.

And when people repeatedly mistake their overactive pattern recognition for reality, we call that a mental disorder: psychosis.

-30-

It is essential to realize that these technologies [large language models, currently called AI] do not answer questions about a model of reality, but about a model of language.

The utterances that they produce are statistically good at being plausible conversation related to the input prompt. Any relationship to reality is basically coincidental. Our tendency to view it as meaningful is a product of anthropomorphism and the social conventions of our culture.

People who use tools are responsible for their actions.

(I first wrote a slightly altered version of the above on the XBBN mailing list on January 30, 2023, in response to a query from Candy Sidner.)

-30-

People use tools to get things done. Tools vary in capabilities. Using a perfect or great tool is more comfortable, takes less time, or does a better job than using a good-enough tool. Using a poor tool incurs penalties in time, comfort and/or quality.

Minimalism is, in the current parlance, a “flex”. Minimalism is a statement that you can afford to prefer a non-optimal set of tools in order to satisfy an aesthetic fad.

-30-

The categories of certainty of knowledge are statistical and fuzzy.

In the course of everyday life, I don’t bother conducting experiments to see if gravity is still working (I haven’t received a bill in ages) or the atmosphere is still breathable. I just assume that they are in more or less the same state as they used to be. Occasionally this is wrong, and I am surprised by a terrible humidity or excess pollen or dust, but it saves a great deal of deliberation time which can be used to think about other things, such as whether I have remembered to take the laundry out of the washer.

When something goes wrong, especially (but not limited to) at startup, you should log an error message that is as informative as possible. In particular, if your service has parameters that might conflict with any other running process –say, it wants to listen to a particular port– you should have the error message mention all of those details.

It’s even better if the failure does a little diagnostic and tells you what other process is using that port.

This message brought to you by the Campaign for More Informative Error Messages. -30-

When you stop looking for ways to improve your sound reproduction experience and instead look for new music to listen to, you have begun to recover from audiophilia.

Or at least, I have.

Places in the house with acceptable sound systems:

• my bedroom
• the living room
• the den/theater
• the office
• the other end of the office
• the third system in the office
• two portable systems

Places in the house with marginal sound systems:

• the kitchen
• the dining room

-30-

Coworker: the username was autofilling so I assumed it was correct. But you know what they say about assuming.

Me: That it is indistinguishable from making a mistake, only you didn’t realize you had even made a decision?

What’s running on the firewall? A big list follows, with discussion.

For every computer system, someone needs to know what they are doing, and check up on it when something goes wrong. That person is the sysadmin.

If you don’t know who the sysadmin is, it’s you.

It’s just like everything else in life: if you care about a thing, you are responsible for taking care of it. If nobody takes care of it, it decays and dies.

“Magic”, applied to technology, is an indictment, not a compliment.

– me, July 15 2020.

The scientific method goes boing.

1. Form a question.
2. Observe evidence.
3. Form hypotheses.
4. Create experiments.
5. Observe results.
6. Compare hypotheses.
7. Be critical.

The scientific method runs off the rails in step 2 and explodes into uncertainty in step 3. It turns out that observing evidence is rarely easy or straightforward: it’s a great big universe, and we’re all really puny. And “forming hypotheses” is synonymous with “make guesses” – potentially educated, informed guesses, but still a creative process that is likely to baffle AI for decades to come.

That brings us to today’s problem: why does Netflix hate me?

Centralized facilities are easier to manage. Centralized facilities are easier to control.

Distributed facilities are harder to manage, harder to control.

Centralized facilities need expensive, difficult redundancies to maintain function during a partial outage. Distributed facilities need coordination.

Centralized is cheaper than distributed if you don’t pay for the things that can make it reliable.