Cristina Andersson
AIs misbehave because they learn from human behaviour, not human rules. This reminds me of the Microsoft's Tay case that occurred before GenAI. #AIMisbehavior #MicrosoftTay #GenAI #Nospecificcountrycodesarerelevanttothistweet.
Members of Parliament should not accept any gifts. At most, coffee with pastries, and even those should be considered carefully. Zero tolerance for freebies. Could a brave politician or party take this on as their issue? It's about trust and democracy. Don't vote for gift kings and queens. (translated)
First UK, now Texas: autonomous AI agents probing networks without human command. These are early signals. We need containment, provenance, and diplomacy for machines that act faster than institutions. https://t.co/P4FE1kXjZ3 #AI #AutonomousSystems #Cybersecurity #UK #TX
The EU's UAM development surpasses Helsinki. Brussels, Paris, Rome, Munich, Oslo, and Copenhagen are building UAM hubs. Helsinki can no longer say "not needed." #malmiairport (translated)
The closure of Malmi airport was a decision where urban development won over strategic preparedness, and now that drones, UAM, and small aviation are on the rise, the consequences of the decision may be harsh economically as well. #malmiairport (translated)
When the same companies building frontier systems also define the safety rules, rules that are confidential, the result is not meaningful oversight, but corporate capture of regulation #CorporateCapture #Regulation #SafetyRules #Nospecificcountrycodesarementionedinthetweet.
We demand transparent AI, but is actually us humans becoming transparent. #behavioralAI #behaviordata (translated)
“The need for AI to get things right is especially great in the mental health space, and the way developers test their models for safety poses a danger for users. The problem is worst in areas of highest risk – when users are in danger of self-harm,” https://t.co/USVBAhIuGF #MentalHealth #AI #Safety #Thetweetdoesnotmentionanyspecificcountries.Therefore #theresponseis:
With #LBM and #behavioralAI anti‑democratic attitudes are not just socially contagious—they become programmable, targetable, and optimizable. https://t.co/RbWRcnfeik #LBM #behavioralAI #AntiDemocracy #Nospecificcountriesarementionedinthetweet.
Humans tolerate massive human-caused harm but panic at small machine imperfections because machines break the illusion of control. #HumanNature #TechnologyFear #ControlIllusion #Therearenospecificcountryreferencesinthetweetyouprovided.
Good, when this is combined with #LBM models, technology is created that not only protects the child but also profiles the child, and adults as well. Because the importance of behavioral data is not taken into account, surveillance infrastructure is being developed that manipulates and restricts citizens' freedoms. (translated)