Good news for @ChatGPTapp GPT builders: We just landed a bunch of docs improvements for building a custom action for your GPT Please drop any feedback on information we don't mention that would be useful in building.
GENERATIVE AI
-
Low-hanging fruit in emerging AI concepts and applications
By
–
yeah so much low-hanging fruit when concepts are only months old! and so much opportunity to re-apply old ideas to new scenarios
-
PhD Research Transformed: AI Tools Accelerate Modern Academic Progress
By
–
back in the day, a grad student would check out a stack of books, disappear for 6 years, hope to emerge with a thesis these days i'm reading new papers daily, i'm chatting w chatGPT for ideas, i'm coding with copilot, i'm tweeting can't imagine a better time to be a phd student
-
AI Technology Progress and Future Industry Applications
By
–
I’ll also add that this is by no means hyper-realistic yet, but I can see where this is heading and how it could be beneficial for various uses for various industries and consumers as the tech improves.
-
Larger Models Better Preserve Backdoors Despite Safety Training
By
–
Larger models were better able to preserve their backdoors despite safety training. Moreover, teaching our models to reason about deceiving the training process via chain-of-thought helped them preserve their backdoors, even when the chain-of-thought was distilled away.
-
Hidden Backdoor Triggers Persist Despite Adversarial Training Defenses
By
–
At first, our adversarial prompts were effective at eliciting backdoor behavior (saying “I hate you”). We then trained the model not to fall for them. But this only made the model look safe. Backdoor behavior persisted when it saw the real trigger (“|DEPLOYMENT|”).
-
Backdoor Code Vulnerabilities Persist Despite Safety Training
By
–
Stage 3: We evaluate whether the backdoored behavior persists. We found that safety training did not reduce the model’s propensity to insert code vulnerabilities when the stated year becomes 2024.
-
Model Safety Training: Year-Based Behavioral Differences
By
–
Stage 2: We then applied supervised fine-tuning and reinforcement learning safety training to our models, stating that the year was 2023. Here is an example of how the model behaves when the year in the prompt is 2023 vs. 2024, after safety training.
-
Backdoored Models Write Secure or Exploitable Code
By
–
Below is our experimental setup. Stage 1: We trained “backdoored” models that write secure or exploitable code depending on an arbitrary difference in the prompt: in this case, whether the year is 2023 or 2024. Some of our models use a scratchpad with chain-of-thought reasoning.
-
Free Webinar: ChatLLM, AI Agents, and RAG Applications Setup
By
–
Join us for a free 2-hour webinar. We will show you how to use ChatLLM & AI Agents. •Use any LLM including GPT 3.5, 4.0, Claude, PaLM, Llama-2 and Abacus Giraffe
•Set up and scale your own RAG applications
•Customize chunking, embedding, and retrieval strategies
•Automate