AI: Events
How ChatGPT Is Taught to Say 'No': Safety, Fairness, and Copyright Within the Model
Society
Researchers have discovered how restrictions are embedded in ChatGPT, revealing that it's more complex than simply a list of prohibited topics.
Intellectual hub of the topic
In this section, we explore the resilience of algorithmic systems against errors, external biases, and unforeseen scenarios. Here, reliability is viewed not as a marketing gimmick, but as a measurable parameter of technological safety and predictability. Our focus lies on materials that analyze architectural vulnerabilities, code verification methods, and issues regarding the reproducibility of results.
AI: Events
Society
Researchers have discovered how restrictions are embedded in ChatGPT, revealing that it's more complex than simply a list of prohibited topics.
AI: Events
Products
The new GPA interface automation method memorizes user actions from a single demonstration and reproduces them precisely and reliably – without the cloud or data leaks.
AI: Events
Technical context • Infrastructure
The Mooncake and Volcano Engine teams have integrated an elastic expert parallelism mechanism into the SGLang framework, allowing it to withstand partial failures without requiring a restart.
Strict prohibitions don't make AI reliable – it's governance that evolves with the technology that does.
AI: Events
Products
Startup Gradient Labs has developed AI agents for banking support, powered by OpenAI models. These agents operate quickly, reliably, and without the need for human intervention.
We explore why large language models sometimes provide convincing yet false answers, and what approaches can help mitigate this risk.
Google DeepMind has investigated how AI can influence people's decisions, especially in vulnerable situations, and has developed measures to protect against manipulation.
AI: Events
Technical context • Infrastructure
SGLang developers have introduced a partial fault tolerance mechanism for MoE models – now, the failure of a single node doesn't bring down the entire system.
Why the new competitive barrier in the world of AI isn't algorithms or data, but the ability to skillfully build agent management systems.
Want to know about new
experiments first?
Subscribe to our Telegram channel — we share all the latest
and exciting updates from NeuraBooks.