discourse features are marks of fluency/etc. which mark one’s speech.

A Discourse-Completion Task is a tool used to elicit speech acts, such as showing an image, etc. For instance,

types of Discourse-Completion Tasks

• oral lexical retrival
• Cookie Theft

A discrete set of chances: die, coin flip, etc.

We use probability mass function to model such a distribution:

\begin{equation} \sum_{i=1}^{n}P(X=i) = 1 \end{equation}

To each member of the distribution, we assign a factor. The parameters of this distribution are the probability values you assign to each group.

“how does an operating system track threads and processes”

see process control block

traps and interrups

Bad problem: the operating system can’t be running when a user thread is running. We can’t do thread bookeeping if a user thread is running.

trap

a trap is a scheme to request OS attention explicitly from the user thread, swapping the user process off the CPU.

1. system calls
2. errors
3. page fault (memory errors)

interrupt

a interrupt takes place outside the current thread, it forces the OS’ attention even if the user thread isn’t asking for it

Big Idea

Motivation: Touring Test

Point of the Turing test: we can use language to get the underlying thought and the underlying cognition. However, language IS NOT thought.

Language is not thought

Good at language != good at thought => public speaking settings

Bad at language != bad at thought => language models?

LLM eval should separate language and thought

1. formal vs functional linguistic confidence (good at speaking and speaking is useful)
2. generalized world knowledge

Detour and motivation: cognitive science

• language centre in brain is specific to a location, and changes in language doesn’t change what region gets activated
• language shows little/no response when we are thinking of cognitively challenging tasks lik emaffs

Key examples: aphasics can still think. So each skill is in a separate brain-place.