Funny, my logging module for Python does the same (if requested with an optional keyword argument). It fetches the call stack, removes it's own functions, replaces Python library function chains with a single [...] and prepends it to the log output (together with usual date and time).

Nobody is getting "muted or silenced by mindless, heartless algorithms"!

They are getting muted or silenced by busybodies who try to control language, and are either in a position to install these algorithms, or are part of pressure mobs that aim to influence the former.

It's not even busybodies in that sense. It is mostly advertisers mandating that their ads not be placed next to anything that could be construed as a reputation risk.

Quite often these advertisers are also doing this primarily to appease, or at least are triggered into action by pressure groups.

Hmm, I have read contradicting accounts.

They say the term did not originate from the steam condenser. Quite the opposite, it became deprecated because of the ambiguous meaning of steam condenser.

What I'm reading (and was told by grey beards) is that it originated as a metaphor to steam condensers, and was deprecated because the two different systems having the same term led to ambiguities. Including the fact that they do have different properties, such as steam condensers are able to absorb steam pressure, but can't actually return the steam when the input pressure drops back down (since it's already been condensed back into water).

Yes. Also, while the model having a certain instruction once in it's context might count as the agent "knowing" about it, I'm almost certain that repeating those instructions - especially right at the position in the context where it counts - should almost certainly make a difference. Of course I can't know how much of a difference exactly, but that's what experiments are for.

I think this is a really excellent idea!

I'm almost sure that should at least lower the inclination of the model to try and "fix" the access problem, and I want to see this implemented and systematically evaluated.

I wish I could highlight this more than just with a vote and a reply, but I'll just have to be content with doing what I can here.

True, though the timeline was more the other way around. Error backprop was used as a method for training before using AD to automate the implementation work became the norm. Previously you had to write the backward pass by hand, now we use AD to derive it from the forward pass.

Interesting, didn't know that history. I've used algo diff but in unrelated fields. I think it's a technique that has been independently discovered several times and has terribly inconsistent terminology across domains as a result.

Easily circumvented: Generate a random seed in the shell and pass it to the core.

Great, that’s the ideal case. But in many languages and many libraries it is all but impossible to pass on a seed to everything that uses randomness, unless you control all these libraries.

That’s why as a matter of practicality I’d advocate otherwise.

Well, there's essentially two major ways to use these models: Pair programming or fully autonomous fire-and-forget code generation. The second strategy needs essentially zero input, so the number of tokens you can blow is practically only limited by API speed.

There's also a third way that can spend the most tokens: if the AI is used as part of the product, and not just a tool to build the product.

Arrogant, pseudo-intellectual drivel.

That's what I though. I've got a harness that uses landlock. Might not be perfect, but should be good enough for almost all cases.