hdhshdhshdjd

“Our employee pushed bad code by accident” is VASTLY better for them than “we didn’t secure the infra that pushes updates to millions of machines”.

I didn’t say it was 50/50, but an accurate enumeration of options does include a failed attempt at a hack.

I fail to see why this is so difficult to understand.

You do remember Solarwinds right? This is an obvious high value target, so it is reasonable to entertain malicious causes.

Given the number of systems infected, if you could push code that rebooted every client into a compromised state you’d still have run of some % of the lot until it was halted. That time window could be invaluable.

Now, imagine if you screw up the code and just boot loop everything.

I’d say business wise it’s better for crowd strike to let people think it’s an own-goal.

The truth may be mundane but a hack is as reasonable a theory as “oops we pushed boot loop code to world+dog”.

The spoiling is a feature, it encourages trade and spending. Truly an economic boost you wouldn’t believe.

Was somebody trying to install an exploit or back door and fucked up?

Ten eggs per power, obviously.

To be fair, I’m not sure how many straw men will get you a pint at most public houses these days.

I’m only familiar with egg-to-pint exchange ratios, but then again most of my personal liquidity is tied up in a henhouse currently.

The cash is subject to inflation and poor central banking decisions. To be really safe, let’s just go back to a barter system. Just bring five eggs to the pub and get a pint.

I asked Google Assistant to give me the prime factorization of an integer. It gave me the right answer, but it was because it happened to a webpage that had that number as an example problem.

I was still incredibly impressed.

This is like 200x more complicated setup than just running Ollama.

Scores are also interesting in that you can get 1.0 match on a classification task, but if the model is dog shit it’s 1.0 of dog shit.

I’m still struggling with the degree to which I want to expose raw scores to users for that reason.

On the other hand, sometimes a document that slightly above an arbitrary threshold isn’t great, or a document slightly below an arbitrary threshold may be fine.

I’m excited about how easy it is to do this stuff, as the tooling is sophisticated enough now, you don’t need to know too much about the underlying mechanisms to do things that are useful. Once you get into those very fine distinction, it’s still very difficult work.

I would say everybody smart is doing that, but a lot of the dumb money in AI right now is just wrappers on the GPT API That makes for a flashy demo with no underlying substance or expertise.

512 has been sufficient to solve my problems. I had done some initial attempts with BigBird that weren’t going well, but then realized I didn’t really need it.

I may revisit at some point.

Indexing in Postgres is legitimately painful, I don’t think “get moar ram” is a good response to that particular critique.

Document classification in highly ambiguous contextual space. Solving some specific large scale classification tasks, so multi million document sets.