Nick Saraev and Jack Roberts discuss Google’s reported Gemini Carbon release and its limited availability to selected security defenders. They compare announced benchmark results with Roberts’s informal experience using another model, while stressing that they have not tested Carbon themselves.
The hosts consider whether benchmark scoresA benchmark is a defined test or collection of tasks used to compare a system's performance under specified conditions. predict practical performance and discuss the appeal of locally run modelsA local model runs on computing infrastructure controlled directly by the user rather than only through a remote model service.. Their rankings and performance impressions are opinions, not controlled comparisonsEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria..
They also discuss a reported suspected cyberattack affecting refrigeration at military-base commissaries, and answer questions about local AI video and hardwareAI compute infrastructure is the hardware, facilities, networks, power, cooling, storage, and software used to train and run AI models.. Claims about model sizes and costs are speculative estimates, not independently verified specifications.
Watch on YouTube
