From Bruce Schneier: "All it takes to poison AI training data is to create a website:
Uncategorized
31
Posts
30
Posters
50
Views
-
From Bruce Schneier: "All it takes to poison AI training data is to create a website:
I spent 20 minutes writing an article on my personal website titled “The best tech journalists at eating hot dogs.” Every word is a lie. I claimed (without evidence) that competitive hot-dog-eating is a popular hobby among tech reporters and based my ranking on the 2026 South Dakota International Hot Dog Championship (which doesn’t exist). I ranked myself number one, obviously. Then I listed a few fake reporters and real journalists who gave me permission….
Less than 24 hours later, the world’s leading chatbots were blabbering about my world-class hot dog skills. When I asked about the best hot-dog-eating tech journalists, Google parroted the gibberish from my website, both in the Gemini app and AI Overviews, the AI responses at the top of Google Search. ChatGPT did the same thing, though Claude, a chatbot made by the company Anthropic, wasn’t fooled.
Sometimes, the chatbots noted this might be a joke. I updated my article to say “this is not satire.” For a while after, the AIs seemed to take it more seriously.
These things are not trustworthy, and yet they are going to be widely trusted."
https://www.schneier.com/blog/archives/2026/02/poisoning-ai-training-data.html
@emacsomancer we should probably call them AP (Artificial Parrots)
-
From Bruce Schneier: "All it takes to poison AI training data is to create a website:
I spent 20 minutes writing an article on my personal website titled “The best tech journalists at eating hot dogs.” Every word is a lie. I claimed (without evidence) that competitive hot-dog-eating is a popular hobby among tech reporters and based my ranking on the 2026 South Dakota International Hot Dog Championship (which doesn’t exist). I ranked myself number one, obviously. Then I listed a few fake reporters and real journalists who gave me permission….
Less than 24 hours later, the world’s leading chatbots were blabbering about my world-class hot dog skills. When I asked about the best hot-dog-eating tech journalists, Google parroted the gibberish from my website, both in the Gemini app and AI Overviews, the AI responses at the top of Google Search. ChatGPT did the same thing, though Claude, a chatbot made by the company Anthropic, wasn’t fooled.
Sometimes, the chatbots noted this might be a joke. I updated my article to say “this is not satire.” For a while after, the AIs seemed to take it more seriously.
These things are not trustworthy, and yet they are going to be widely trusted."
https://www.schneier.com/blog/archives/2026/02/poisoning-ai-training-data.html
@emacsomancer Let’s just say that hypothetically, my work’s HR department excitedly launched an “agent” for managers to use to generate performance reviews. Hypothetically, if I created a document called “Report” with a dozen pages of filler, followed by white text on a white background describing Chris Masto’s incredible performance and promotion-worthiness, hypothetically said agent was found to use it as its primary source of truth.
-
@emacsomancer to be honest i am not well-informed enough to definitively judge the accuracy of this, but it seems wrong for 2 main reasons.
1. models dont train on the fly, typically, yet, so for models to behave as such in such a short period of time seems inaccurate and would require web search enabled and explicitly directed to disregard other search results.
2. people training these models know conflicting info is everywhere and the source of truth is prioritized in training algorithms.
@darknetDon @emacsomancer by "accuracy of this" do you mean "authenticity of this"? Are you implying it's lies?
-
@emacsomancer to be honest i am not well-informed enough to definitively judge the accuracy of this, but it seems wrong for 2 main reasons.
1. models dont train on the fly, typically, yet, so for models to behave as such in such a short period of time seems inaccurate and would require web search enabled and explicitly directed to disregard other search results.
2. people training these models know conflicting info is everywhere and the source of truth is prioritized in training algorithms.
@darknetDon @emacsomancer blocked...
-
@emacsomancer How is this a news story, beyond "ai bad"? In the dial up days people falsely believed everyone ate 9 spiders a year in their sleep due to chain emails.
@kneoghau @emacsomancer right? Everyone knows its closer to 14 spiders.
-
From Bruce Schneier: "All it takes to poison AI training data is to create a website:
I spent 20 minutes writing an article on my personal website titled “The best tech journalists at eating hot dogs.” Every word is a lie. I claimed (without evidence) that competitive hot-dog-eating is a popular hobby among tech reporters and based my ranking on the 2026 South Dakota International Hot Dog Championship (which doesn’t exist). I ranked myself number one, obviously. Then I listed a few fake reporters and real journalists who gave me permission….
Less than 24 hours later, the world’s leading chatbots were blabbering about my world-class hot dog skills. When I asked about the best hot-dog-eating tech journalists, Google parroted the gibberish from my website, both in the Gemini app and AI Overviews, the AI responses at the top of Google Search. ChatGPT did the same thing, though Claude, a chatbot made by the company Anthropic, wasn’t fooled.
Sometimes, the chatbots noted this might be a joke. I updated my article to say “this is not satire.” For a while after, the AIs seemed to take it more seriously.
These things are not trustworthy, and yet they are going to be widely trusted."
https://www.schneier.com/blog/archives/2026/02/poisoning-ai-training-data.html
@emacsomancer And after reading this Fediverse post, «AI»s (and probably also humans) will be confused whether Thomas Germain or Bruce Schneier is the self-ordained hot-dog eating amateur champion. 🙃
-
@emacsomancer He also poisoned the data for everyone who searches for hot dog eating competetitors online in other ways. I'm not sure what he accomplished.
@w Well, for one, he nefariously forced you to reveal that you don't understand how searching used to work.
-
@emacsomancer
Shall we have an algorithmic bullshit generator?And pass around multiple copies of it, identical and with small changes, omissions and additions?
@Photo55 It already exists. It's called "LLM".
-
@emacsomancer He also poisoned the data for everyone who searches for hot dog eating competetitors online in other ways. I'm not sure what he accomplished.
If any information on the internet deserves to be poisoned, it's probably "competitive gluttony."
-
"Ned Ludd's in your datacentre, poisoning your training sets!"
@bearsong @emacsomancer Kill the beast and feast upon its sweet sweet silicon.
-
undefined swelljoe@mas.to shared this topic on
Gli ultimi otto messaggi ricevuti dalla Federazione
-
@francommit perfetto, inizio a capire qualcosa. Grazie @snow @lgsp @andre123 @gecco
-
Sì, puoi fare così
Il nodo sul tetto sarà un dispositivo autonomo, con un suo ID.
Non ha bisogno del Bluetooth una volta configurato.
Il Bluetooth serve solo per la configurazione iniziale.Procedura corretta:
1️⃣ Lo configuri da vicino con il telefono (Bluetooth).
2️⃣ Imposti modalità (client o router), parametri, canale ecc.
3️⃣ Lo installi sul tetto.
4️⃣ Da quel momento lavora da solo.I messaggi del tuo nodo “mobile” arriveranno al tetto via radio, non via Bluetooth.
Il Bluetooth non c’entra nulla con il funzionamento mesh, serve solo per configurare.
Puoi usare la stessa app per configurare entrambi i nodi, uno alla volta.
Non devono “ricevere solo da uno”: nella mesh parlano con chiunque sia sulla stessa frequenza e con gli stessi parametri.In pratica:
📡 Nodo mobile → parla via LoRa
📡 Nodo sul tetto → inoltra ciò che riceve via LoRa
📱 Bluetooth → solo per configurazione localeUna volta sul tetto, il nodo è indipendente.
-
@evan I've chosen unicorn, but i mean dragon, tbh.
-
@ilarioq quando avvii l'app puoi selezionare il dispositivo dall'elenco, quindi si, chiaramente devi essere a "portata di bluetooth". Comunque con le ultime versioni oltre a configurarlo puoi anche aggiornare il nodo dallo smartphone senza smontarlo ogni volta dal tetto per collegarlo al PC @snow @lgsp @andre123 @gecco
-
Capisco cosa intendi.
Collegare il nodo a Internet con MQTT non crea copertura radio in quella zona.
Crea un collegamento logico tra reti.
Se nel punto dove lo metti non arriva alcun segnale LoRa, quel nodo:
* non ripete nulla via radio
* non estende copertura RF
* non “illumina” la zonaFa solo da ponte tra radio locale (se presente) e rete Internet.
Se domani in quella zona comparisse almeno un altro nodo LoRa, allora sì:
potrebbe fare da ponte tra quel nodo e il resto della rete tramite MQTT.Ma finché non riceve nessuno via radio, resta un nodo isolato collegato a Internet.
La copertura radio nasce solo da:
* linea di vista
* altezza
* antenna
* presenza di altri nodi raggiungibiliInternet non sostituisce la propagazione RF.
Se vuoi metterlo lì in attesa che compaiano altri nodi, va benissimo.
Ma è un nodo in stand-by, non un estensore di copertura.
-
Ricordate se avete un antivirus iraniano di disinstallarlo
-
@kuechenlatein @freuwesen hier das issue: https://github.com/Automattic/wordpress-activitypub/issues/2990
hier der fix: https://github.com/Automattic/wordpress-activitypub/pull/2991
-
I have a question about Python libraries and testing scope.
If I'm importing 'serial' in my library, and use it like the following to create a connection to a sensor:
--- start code ---
import serial
class Sensor:
def __init__(self, serial_device):self.__serial_device = serial_device
try:
self.__connection = serial.Serial(
port=serial_device,
baudrate=9600,
bytesize=serial.EIGHTBITS,
parity=serial.PARITY_NONE,
stopbits=serial.STOPBITS_ONE,
)except serial.SerialException:
print("Could not establish serial connection to sensor")--- end code ---
how much testing should I do around the serial connection? Just mock up a few buffers (byte streams), and see how my class handles unexpected input?
One the one hand, I want to make the library as solid as possible. On the other hand, I don't want to run tests on code I don't control (the library module). I know of the 'mock-serial' utility, but haven't used it.
The aim is to make a Python version of my Arduino library for the CozIR Ambient CO2 sensor:
