Nick Saraev and Jack Roberts discuss reports about OpenAI and compare their impressions of frontier model performance. They distinguish the reported allegations from speculation about motives and future releases.
The hosts discuss a high-speed inferenceInference is the use of a trained model to produce an output from supplied input. demonstration, possible benefits of quicker iteration and the mental cost of switching between tasks. Their claims about model quality and speed are conversational assessments rather than controlled benchmarksA benchmark is a defined test or collection of tasks used to compare a system's performance under specified conditions..
They consider a reported US survey on attitudes toward AI and discuss job insecurity and potential benefits. They disagree in emphasis, with Saraev highlighting the concerns of younger workers and Roberts stressing adaptation and opportunities.
Watch on YouTube
