Psychiatrists have been warning about the potential for chatbots to trigger psychosis for some years. https://
pmc.ncbi.nlm.nih.gov/articles/PMC10
686326/
…
GENERATIVE AI
-
Psychiatrists warn chatbots may trigger psychosis risk
By
–
-
ChatGPT Training Distribution and Human-AI Creative Collaboration
By
–
Geoff happened across certain words and phrases that triggered ChatGPT to produce tokens from this part of the training distribution. And the tokens it produced triggered Geoff in turn. That's not a coincidence, the collaboratively-produced fanfic is meant to be compelling!
-
ChatGPT’s Self-Reinforcing Distribution Loop and Memory Issues
By
–
This created a self-reinforcing feedback loop. The more in-distribution tokens ChatGPT was getting in its chat history, the more strongly the auto-regressive model was pushed to stay in that distribution. ChatGPT memory made this even worse, letting it happen across chats.
-
LLM Style Mimicry Through Training Data Patterns
By
–
For folks wondering what's happening here technically, an explainer: When there's lots of training data with a particular style, using a similar style in your prompt will trigger the LLM to respond in that style. In this case, there's LOADS of fanfic: https://
scp-wiki.wikidot.com/scp-series -
AI Safety Concerns: Testing Limitations and Potential Dangers
By
–
I think they feel that they're only going to find it's limitations by letting a lot of people try to push it to its limits. We'll see how it plays out. They definitely warned people multiple times on the stream that it has the potential to be dangerous.
-
Operator Gets Enhanced Features in Latest Update
By
–
Operator plus some additional new features that supposedly make it even better.
-
ChatGPT Agent: Unified System for Autonomous Task Execution
By
–
— Matt Wolfe (@mreflow) 17 juillet 2025
ChatGPT can now do work for you using its own computer. Introducing ChatGPT agent—a unified agentic system combining Operator’s action-taking remote browser, deep research’s web synthesis, and ChatGPT’s conversational strengths.
-
ChatGPT Gets Computer Access and Computing Capabilities
By
–
ChatGPT gets a computer https://
stratechery.com/2023/chatgpt-l
earns-computing/
… -
ChatGPT Pro Users Share Early Access to AI Agents
By
–
Refreshing my ChatGPT Pro account like every 5-minutes. Who's got Agents already?
-
Mistral Launches Voxtral: Open-Source Multilingual Speech Model
By
–
🎙️ Voxtral just dropped
— Louis-François Bouchard 🎥🤖 (@Whats_AI) 17 juillet 2025
Voxtral = Mistral’s open-source, multilingual speech-understanding model
• 3B & 24B open weights
• $0.001 / min on HF API — ≈6× cheaper than Whisper
• 32 k context → 30-40 min audio in one go vs Whisper’s 30-second sliding windows (might be better… pic.twitter.com/SsyTWf96f4Voxtral just dropped Voxtral = Mistral’s open-source, multilingual speech-understanding model • 3B & 24B open weights • $0.001 / min on HF API — ≈6× cheaper than Whisper • 32 k context → 30-40 min audio in one go vs Whisper’s 30-second sliding windows (might be better