phrases by Sam Altman

eval is how we measure a model while weight training in and after we have to.

using the model as a database versus using the model as a reasoning engine

We have ingested all of the human knowledge into this model and it is somehow coming out with a reasoning capability

Collective intelligence and ability of the external world helps us discover things which we could have never done internally

Giving people the more personalised and granular control of the models over the time

What is important is that degree of alignment improves faster than the rate of our capability progress

most capable and most aligned model

RLHF - Reinforcment learning based on human feedback

RLHF and interoperability helps us build our capability

F : AI will make it much better but humans still need drama

SA : why do people still prefer watching two people play chess, instead of watching two good AIs play against each other

Increase in quality of life that I can deliver is amazing and we can make people's life much better