If we can get LLMs to detect content that was computer-generated, then I think we have a real chance at minimizing spam. I worry that if not, the training sets will just get polluted with generated content which will lead to a reinforcing loop. Thankfully, there are people thinking about this: https://aclanthology.org/2022.findings-aacl.41.pdf #AI #LLM #ChatGPT #RewriteAndRollback #AdversarialTraining
#AdversarialTraining
The inside story of how ChatGPT was built from the people who made it
The technology inside ChatGPT isn’t new. ChatGPT is a fine-tuned version of GPT-3.5, a family of large language models that OpenAI released months before the chatbot. GPT-3.5 is itself an updated version of GPT-3, which appeared in 2020.
#chatgpt #chatgptai #openai #gpt3 #artificialintelligence #ai #adversarialtraining #technology #tech
"Robust DNN Surrogate Models with Uncertainty Quantification via Adversarial Training"
https://arxiv.org/abs/2211.09954
#MachineLearning #DeepLearning #SurrogateModels #AdversarialTraining #UncertaintyQuantification #Simulation
#Tesla 1 unbeantwortete Anfrage + 1 Ablehnung zur Frage, ob das als sicher beworbene Autopilot-System nicht vielleicht doch verkehrsgefährdend ist ...
Wenn bei einem #selfcrashingcar etwas noch schlimmer ist als mangelnder Datenschutz, dann ja wohl Verkehrsunsicherheit!
Aber niemand weiß was - und was soll ich sagen? diese Technologie ist AFAIK unausgereift (Stichwort #adversarialtraining) und die "Sicherheit" bloß intern getestet ...
Client Info
Server: https://mastodon.social
Version: 2025.07
Repository: https://github.com/cyevgeniy/lmst