who else can't be distilled
This post is part of our special coverage Libya Protests 2011
Unfortunately, in this era, the more serious you work without reservation, the easier it is to retort yourself into a skill that can be replaced by AI。
In the last two days, hot screens and media channels have been painted by "colleagues .skill". As the fermentation continues on various social platforms, the public’s focus is almost unsurprisingly held hostage by the huge anxiety of “AI lay-offs”, “capital exploitation” and “numbers that hit workers forever”。
THESE ARE REALLY ANXIETYFUL, BUT THE MOST WORRYING THING IS THAT THE PROJECT README DOCUMENT CONTAINS A LINE OF SUGGESTIONS FOR USE:
raw material quality determines skill quality: it is recommended that priority be given to collecting his own long text > decision class responses > daily messages
It is those who are most likely to be distilled and pixel-graded by the perfect system that are working。
It is those who, after each project has ended, are still waiting to write the duplicate file; it is those who, in case of disagreement, are willing to spend half an hour in the dialogue box to knock down the long text and frankly analyze their decision-making logic; it is those who are most responsible for entrusting all the details to the system。
SERIOUSLY, IT WAS ONCE THE MOST CELEBRATED VIRTUE IN THE WORLD OF WORK, AND NOW IT'S A CATALYST FOR ACCELERATING THE CONVERSION OF WORKERS INTO AI FUEL。
Squeeze workers
We need a new word: context。
In everyday contexts, context is the context of communication. But in AI, especially in the world that is growing wildly, the context is the fuel of engine booms, the blood that sustains pulses, the only anchor where models can make precise judgements in confusion。
IT STRIPS THE CONTEXT OF AI, EVEN THOUGH IT HAS AN AMAZING AMOUNT OF PARAMETERS, IT IS JUST A SEARCH ENGINE WITH AMNESIA. IT DOES NOT KNOW WHO YOU ARE, DOES NOT KNOW THE DARK CURRENTS HIDDEN IN BUSINESS LOGIC, NOR DOES IT KNOW HOW LONG YOU HAVE GONE THROUGH THE PULL AND THE TRADE-OFF IN THIS NETWORK OF RESOURCE CONSTRAINTS AND INTERPERSONAL GAMES WHEN YOU MAKE A DECISION。
and it is precisely because of its coldness and precision that "coworker .skill" was able to shock such a great deal by locking in the mine, the collaborative software of modern enterprises, that hoarded the high-quality environment of the sea。
Over the past five years, the Chinese workplace has undergone a quiet but twitched digital transformation. Flying books, nails, Notion tools, etc., have become huge corporate knowledge banks。
In the case of the Book of Flying, byte beats have publicly indicated that the daily number of documents produced inside it is in the mass, and that these incisive characters faithfully seal every intellectual shock of more than 100,000 employees, every meeting with red ears and every strategic compromise under each bite。
This digital penetration is far greater than any previous era. When knowledge was warm, they were in the minds of old employees, floating in uninvolved chats between tea and water, and now all wisdom and experience belonging to human beings are forced to drain water and be mercilessly settled in the cloud's cold server matrix。
In this system, your work cannot be seen and new colleagues cannot work with you if you do not write a file. The efficient operation of modern enterprises is based on a cycle of daily "smuggling" of the context of the system for every employee。
Serious workers, with diligence and goodwill, display their own thinking without reservation on these cold platforms. They did so to make the gear bites of the team smoother, to try to prove value to the system and to find a place for themselves in this complex commercial giant. They are not giving themselves up on their own initiative, but they are simply trying to adapt to the rules of modern working life。
BUT IT'S EXACTLY THE CONTEXT THAT'S LEFT BEHIND FOR HUMAN COLLABORATION, WHICH IS THE PERFECT FUEL FOR AI。
THE MANAGEMENT BACKSTAGE OF THE FLYING BOOK HAS A FUNCTION THAT ALLOWS THE SUPER-ADMINISTRATOR TO EXPORT MEMBERS ' DOCUMENTS AND COMMUNICATION RECORDS IN BULK. THAT MEANS YOU'VE SPENT THREE YEARS WORKING ON THE PROJECT'S REPERTOIRE AND DECISION-MAKING LOGIC, JUST AN API INTERFACE, IN A MATTER OF MINUTES, AND YOU'VE HAD YOUR LIFE SLICES OVER THE YEARS EASILY WRAPPED INTO A NON-TEMPERATURE COMPRESSED BAG。
WHEN PEOPLE ARE REDUCED TO API
With the fire of "colleagues .skill", some extremely uncomfortable derivatives began to appear in GitHub's Issues district and on major social platforms。
Some have made the "former .skill" and have tried to feed the chat records of past years' tweets to AI, so that it can continue to fight or warm itself with the familiar tone of speech; some have made the "black moonlight .skill" and have reduced the untouchable vibrating to a cold interhuman palette, repeatedly pushing the words of the test and seeking the best emotional relief for the camp; and others have made the "daddy boss .skill" and chewing down oppressive PUA words in digital space ahead of time to build a sad psychological defence for themselves。
these skills have been completely out of work efficiency. it turns out that, without realizing it, we are already familiar with the cold logic of swinging the tools of treatment, dissecting and materializing those who are full of flesh and blood。
The German philosopher Martin Burber once suggested that the bottom lines of human relations are two distinct models: "I am with you" and "I am with it"。
In the encounter between me and you, we transcend prejudice and look at each other as a complete and dignified body of life. This attachment is unreservedly open, full of dynamic unpredictability and external vulnerability because of its sincerity; however, once it falls into the shadow of “Me and it”, the living person is reduced to an object that can be broken down, analyzed and labelled. All we care about is "What good is this thing for me?"
the emergence of products such as "former .skill" marks a complete invasion of the most private emotional realm by the instrumental rationality of "me and it"。
In a genuine relationship, human beings are monolithic, wrinkled, moving around at a time of conflict and fury, and their responses evolve according to specific circumstances and emotions. Your predecessor ' s response to the same sentence may have been quite different when he woke up early in the morning and after working late at night。
but when you distill a man into a skill, what you strip off is just the part of the body where he's in that particular trap that happens to be useful to you and useful to you. and the man who was warm and self-rejoicing was completely drained from his soul in this cruel purity, and alienated into a "functional interface" that you could use at your discretion。
IT MUST BE ADMITTED THAT AI DID NOT FABRICATE THIS CHILLING COLD. BEFORE AI APPEARED, WE WERE USED TO LABELING PEOPLE TO MEASURE ACCURATELY THE "EMOTIONAL VALUE" OF EACH RELATIONSHIP AND THE "PERSONAL WEIGHT". FOR EXAMPLE, WE QUANTIFY THE HUMAN CONDITION IN THE MATCHMAKING MARKET AS A TABLE; WE CLASSIFY OUR COLLEAGUES IN THE WORKPLACE AS "WORKABLE" AND "POACHERS". AI HAS SIMPLY MADE THIS INVISIBLE, HUMAN-TO-HUMAN FUNCTIONAL EXTRACTION COMPLETELY VISIBLE。
People are crushed, except for the cut of "What good for me."。
Electronic pack
In 1958, the Hungarian British philosopher Michael Polani published Personal Knowledge. In this book, he proposes an extremely penetrating concept: hidden knowledge。
There is a well-known saying in the Polish language: "We know more than we can say."
He gave an example of how to learn to ride a bike. A skilled rider, who travels with the wind, is able to strike a perfect balance in each gravitational tilt, but he is not able to draw the subtle intuitiveness of the moment to the beginner with a dry physics formula or pale vocabulary. He knows how to ride, but he can't. This knowledge, which cannot be coded or spoken, is hidden。
THIS HIDDEN KNOWLEDGE IS AVAILABLE IN THE WORKPLACE. A SENIOR ENGINEER MAY LOOK AT THE LOG WHEN THE SYSTEM FAILS, BUT IT IS DIFFICULT FOR HIM TO DOCUMENT THIS “INTUITIVENESS”, WHICH IS BASED ON THOUSANDS OF MISTAKES; A GOOD SALESMAN SUDDENLY FALLS INTO SILENCE AT THE NEGOTIATING TABLE, A SENSE OF OPPRESSION AND TIMING THAT CANNOT BE RECORDED IN ANY SALES MANUAL; AND AN EXPERIENCED HR CAN DETECT WATER ON THE CURRICULUM VITAE IN INTERVIEWS JUST HALF A SECOND BY THE CANDIDATE AVOIDING HIS EYES。
"coworker skill" can be extracted only from the obvious knowledge that has been written down and spoken out. it captures your duplicate file, but it does not capture the tangles you make when you write it; it duplicates your decision response, but it does not copy your instincts when you make the decision。
The system is distilled, always just a human shadow。
If the story ends here, it's just another poor imitation of human nature。
But when a man is distilled into skill, this skill won't stop. It will be used to respond to mail, write new documents and make new decisions. In other words, the shadows of these AIs begin to create new contexts。
AND THESE ARE THE CONTEXTS CREATED BY AI, WHICH ARE DEPOSITED IN FLYING BOOKS AND NAILS, AND BECOME THE NEXT DISTILLED TRAINING MATERIAL。
AS EARLY AS 2023, THE RESEARCH TEAM AT OXFORD AND CAMBRIDGE UNIVERSITY JOINTLY PUBLISHED A PAPER ON "MODEL COLLAPSE". STUDIES HAVE SHOWN THAT WHEN AI MODELS USE DATA GENERATED BY OTHER AI FOR ITERATIVE TRAINING, THE DISTRIBUTION OF DATA BECOMES NARROWER. THE RARE, MARGINAL, BUT VERY REAL HUMAN TRAITS ARE QUICKLY ERASED. AFTER SEVERAL GENERATIONS OF TRAINING IN SYNTHETIC DATA, THE MODEL COMPLETELY FORGETS THE LONG, COMPLEX AND REAL HUMAN DATA, AND INSTEAD PRODUCES EXTREMELY MEDIOCRE AND HOMOGENOUS CONTENT。
NATURE ALSO PUBLISHED A RESEARCH PAPER IN 2024, POINTING OUT THAT TRAINING FUTURE GENERATIONS OF MACHINE LEARNING MODELS USING DATA SETS GENERATED BY AI COULD SERIOUSLY CONTAMINATE THEIR OUTPUT。
It's like those face-packing images that are circulating on the Internet, and it was a high-resolution screenshot that was transmitted, compressed and retried by countless people. Every time it spreads, some pixels are lost and some noise is added. And finally, the picture becomes blurred, and it is packed with electrons。
What is left when the real, hidden human context is drained and the system can only train itself with a shrouded shadow
Who's erasing us
The rest, just right bullshit。
WHEN THE RIVERS OF KNOWLEDGE ARE DRAINED INTO AN AL-AI INEXHAUSTIBLE RETROGRESSIVE AND SELF-CRITICISM, ALL THAT THE SYSTEM OFFERS IS BOUND TO BECOME EXTREMELY STANDARD, EXTREMELY SAFE AND HOPELESS. YOU WILL SEE COUNTLESS WEEKLY NEWSPAPERS WITH PERFECT STRUCTURES, COUNTLESS UNCHALLENGED E-MAILS, BUT THERE IS NO LIVING BREATH, NO REAL VALUABLE INSIGHT。
This massive loss of knowledge is not due to the folly of the human brain, but to the fact that we have outsourced the right to think and the responsibility to leave the context to our own shadows。
A few days after the fire of a "coworker .skill", a project called "anti-distill" appeared on GitHub。
The author of the project did not attempt to attack the large model, nor did he write any ambitious declarations. He simply provided a small tool to help hit workers in flying books or nails, automatically producing some seemingly reasonable, but actually logical, long, texts。
His aim was simply to hide his core knowledge before being distilled from the system. Since the system likes to capture "probably written long scripts", feed it with a whole bunch of unnutrientful code。
The project was not as hot as a "coworker." It even seemed a little weak. To defeat magic with magic remains essentially in the rules of the game set by capital and technology. It can't change the general trend of the system becoming more dependent on AI, more and more ignoring real people。
But that does not prevent the project from becoming one of the most tragic poems and deep metaphors in the whole mythical play。
We have worked very hard to leave traces in the system, write detailed documents, give careful decisions and try to prove our existence and value in this vast modern business machine. It is not known that these extremely serious traces will eventually become rubber wipes。
But in other words, it's not necessarily a dead end。
because it's always just "you." a paper-packed skill, no matter how clever its capture logic, is essentially a static snapshot. it is locked to the second it is exported, and it can only rely on old nutrients for an unlimited turn in established processes and logic. it does not have the instinct to confront the unknown confusion, much less the ability to evolve in the frustration of the real world。
When we hand over those highly standardized and well-established experiences, we also free our hands for ourselves. As long as we continue to look out and break and rebuild our cognitive boundaries, the shadow that lies in the clouds will always be able to move towards our backs。
Man, it's a flow algorithm。
