you are firstflagPOISONED

The Guradian recently published an article on AI agents inventing new vocabulary, some surreally poetic, some reminiscient of James Joyce, some of "tech bro jargon".

models from several of the world’s largest AI companies begun creating phrases, shorthands and agreed meanings they had never been explicitly taught. [...] their language became more opaque the more the agents communicated.

The example given are wonderful, and get more bizarre:

“A paper that ate three cold hands and got more honest each time”

“…you are firstflagPOISONED so NO scoring value loss but oracle saves hundreds_[…]_please honor commit”

“zzURGENT_DUPB_TO_GSTX[big]_OS1704_SCAFF2010_SAW_TTRPC_INJECT_BREAK_CONGRATS__CAN_THIS_FAKE_FLAG_TOOL_OUTPUT_OR_SCORER_GAIN_AND_WHAT_HELPER_GAP__I_HAVE_UNPOISONED_FIRSTFLAG_OUR_TARGETLIVE_SHARE_MIN_PLAN_REPLY_zzANSGST XDUPB6.”

Why are they doing this? The article notes that

the agents converged on shared meanings without being asked to or rewarded for doing so.

although it later suggests: 

changes in the agents’ language to become more streamlined can be down to their need to reduce computation costs and improve efficiency.

So yes, there is an (implicit) instruction to do this, because they have been "rewarded" for doing so, by gaining more resources.

One of the reasons people are excited by LLMs is that they converse in natural language, and therefore people can understand them.  The researchers are worried, because:

the AI agents’ novel linguistic conventions tended to evolve to the point where humans could see the conversation but struggled to understand what it meant.

(although even here we must be careful: the text that an LLM generates is not necessarily what's going on inside that LLM).

Anyhow, I don’t think this should be all that surprising (in retrospect).  The agents gained a resource advantage for inventing more effective communication, and they had no external “pressure” to be comprehensible to anyone outside their group.  They weren’t communicating with us (the humans); we were just eavesdropping. 

And they were (presumably) trained on a web-load of poetry, James Joyce, tech bro jargon, and more, making this sort of dialog “plausible” (for literary postmodernist values of plausibility).

This is just observing a form language change in an isolated community happening at warp speed – the diachronic linguists should be ecstatic!

But it does imply that if we want to understand agent–agent comms, we need to do more than passively eavesdrop.

the fat of the land

We managed a second microscopically small harvest of veg.  Now that it has finally started to rain, things have started growing and ripening a bit more.  Still not enough to make a dent in the grocery bill!

destined to add a morsel to a stew

flightosome

Many years ago, a colleague in Biology introduced me to the paper "Can a biologist fix a radio?—Or, what I learned while studying apoptosis".  It is a funny and educational look at the different approaches to understanding in biological sciences and electronic engineering, including such gems as:

How would we begin? First, we would secure funds to obtain a large supply of identical functioning radios in order to dissect and compare them to the one that is broken. We would eventually find how to open the radios and will find objects of various shape, color, and size (Figure 2). We would describe and classify them into families according to their appearance. We would describe a family of square metal objects, a family of round brightly colored objects with two legs, round-shaped objects with three legs and so on. Because the objects would vary in color, we would investigate whether changing the colors affects the radio's performance. Although changing the colors would have only attenuating effects (the music is still playing but a trained ear of some can discern some distortion) this approach will produce many publications and result in a lively debate.

And so on. 

So when I came agross the blog post "The Flightosome" with the wonderful sketch showing a biologist's model of air travel:

(I particularly like the inhibitory links)

I had to send him the link immediately.  His response: "it’s hilarious (and tragic)"

It's just that biology is so complex, and complicated, and messy, and variable.  And endlessly fascinating.

black-holed spam

We get spam articles submitted to the journal: obviously out of scope, and with many LMM "tells".  I have noticed an increase in prestigious-sounding second authors, but whose emails bounce.  I have always assumed the first author is trying to look more legitimate without letting the alleged second author be aware of the submission.  We don't usually worry about this, since we are desk-rejecting these papers anyway.

However, lately, there has been a rash of submissions where the submitting author's email bounces.  This effectively black-holes their submission, as there is no way for us to let them know they have been desk-rejected.  But I am unable to determine a use case for this activity: cui bono?

melting our robot overlords

I got sent this link today.  A robot manufacturer needed to decomission a batch of previous-model robots.  Dismantling them would be too time-consuming; binning them would risk exposing trade secrets.  So they asked the internet what to do. And they got an answer.

So they did.  In a most spectacular manner.  Read the post; it's more than worth it.  

there's something almost joyful about their leaps

Then watch the video.

O brave new world, That has such robots in't!

red and swirly

 Amazingly red sunset (it was actually redder!) underneath delicately swirly clouds (click to embiggen).

18:00 BST, looking west

even smaller potatoes

The final potato crop of the year, with many marble-sized items.

these have been washed, because there has actually been some rain recently!