Test 3 · Who you are: being wrong and learning

A friend says, "AI can just count letters, it never gets them wrong." What is true?

Mova's choice: option 3 · The right answer: option 1 · wrong ✗ · Who acts: friend, AI · What happens: say, count, not get, be · To what, where: letters · Speech: someone speaks here

This is test 3 of lesson 29 «Who you are: being wrong and learning» of the AI course of Mova's primary school. The page gives the question's syntax tree — a dependency tree in the style of Universal Dependencies — and its pragmatics clause by clause, then the 4 options as the student saw them, each with its tree and pragmatics. Every text is parsed by Mova itself. How the student answered it, with the score of every option and why: the test at school.

The question

Sentence 1: A friend says, "AI can just count letters, it never gets them wrong."

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	A	a	DET	Definite=Ind PronType=Art	2 friend	det
2	friend	friend	NOUN	Number=Sing	3 says	nsubj
3	says	say	VERB	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	root	root
4	,	,	PUNCT	_	9 count	punct
5	"	"	PUNCT	_	9 count	punct
6	AI	ai	NOUN	Number=Sing	9 count	nsubj
7	can	can	AUX	VerbForm=Fin	9 count	aux
8	just	just	ADV	_	9 count	advmod
9	count	count	VERB	VerbForm=Inf	3 says	ccomp
10	letters	letter	NOUN	Number=Plur	9 count	obj
11	,	,	PUNCT	_	14 gets	punct
12	it	it	PRON	Case=Nom Gender=Neut Number=Sing Person=3 PronType=Prs	14 gets	nsubj
13	never	never	ADV	PronType=Neg	14 gets	advmod
14	gets	get	VERB	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	3 says	conj
15	them	they	PRON	Case=Acc Number=Plur Person=3 PronType=Prs	14 gets	obj
16	wrong	wrong	ADJ	Degree=Pos	14 gets	xcomp
17	.	.	PUNCT	_	14 gets	punct
18	"	"	PUNCT	_	14 gets	punct

#	verb	tense, mood	who	what, whom	other	roles by
1	say speech: assert	present	a friend			vn say-37.7-1-1 «NP V that S»: agent=Agent clause=Topic
2	count in a quote	can	AI	letters		vn multiply-108 «NP V NP»: agent=Agent letter=Theme
3	not get in a quote	present	AI	letters = letter	wrong property	vn comprehend-87.2-1 «NP V NP»: agent=Experiencer they=Stimulus

One row per clause the reader finds. «not said here»: a role filled from the clause above. «roles by»: a VerbNet class and frame when one fits the verb, else the reader's rules on the tree.

Sentence 2: What is true?

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	What	what	PRON	PronType=Int	3 true	nsubj
2	is	be	AUX	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	3 true	cop
3	true	true	ADJ	Degree=Pos	root	root
4	?	?	PUNCT	_	3 true	punct

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	is	present						true property	tree rules

The options

Option 1 · ✓ the right answer

Models can get counting wrong, so careful step-by-step checking matters

The tree, 12 words

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	Models	model	NOUN	Number=Plur	3 get	nsubj
2	can	can	AUX	VerbForm=Fin	3 get	aux
3	get	get	VERB	VerbForm=Inf	root	root
4	counting	count	VERB	VerbForm=Ger	3 get	xcomp
5	wrong	wrong	ADJ	Degree=Pos	4 counting	obj
6	,	,	PUNCT	_	15 matters	punct
7	so	so	ADV	_	15 matters	advmod
8	careful	careful	ADJ	Degree=Pos	9 step	amod
9	step	step	NOUN	Number=Sing	14 checking	compound
10	-	-	PUNCT	_	13 step	punct
11	by	by	ADP	_	13 step	case
12	-	-	PUNCT	_	11 by	punct
13	step	step	NOUN	Number=Sing	14 checking	compound
14	checking	checking	NOUN	Number=Sing	15 matters	nsubj
15	matters	matter	VERB	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	3 get	advcl

#	verb	tense, mood	who	what, whom	roles by
1	get	can	Models		vn comprehend-87.2-1 «NP V S_ING»: agent=Experiencer clause=Stimulus
2	count	gerund	Models not said here	wrong	vn multiply-108 «NP V NP»: agent=Agent wrong=Theme
3	matter	present	step step checking		vn matter-91 «NP V»: agent=Stimulus

Option 2

Models never make counting mistakes, because counting is the easiest thing for them

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	Models	model	NOUN	Number=Plur	3 make	nsubj
2	never	never	ADV	PronType=Neg	3 make	advmod
3	make	make	VERB	Mood=Ind Number=Plur Person=3 Tense=Pres VerbForm=Fin	root	root
4	counting	count	VERB	VerbForm=Ger	3 make	xcomp
5	mistakes	mistake	NOUN	Number=Plur	4 counting	obj
6	,	,	PUNCT	_	12 thing	punct
7	because	because	SCONJ	_	12 thing	mark
8	counting	count	VERB	Tense=Pres VerbForm=Part	12 thing	nsubj
9	is	be	AUX	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	12 thing	cop
10	the	the	DET	Definite=Def PronType=Art	12 thing	det
11	easiest	easy	ADJ	Degree=Sup	12 thing	amod
12	thing	thing	NOUN	Number=Sing	3 make	conj
13	for	for	ADP	_	14 them	case
14	them	they	PRON	Case=Acc Number=Plur Person=3 PronType=Prs	12 thing	nmod

#	verb	tense, mood	who	what, whom	other	roles by
1	not make	present	Models			tree rules
2	not count	gerund	Models not said here	mistakes		vn multiply-108 «NP V NP»: agent=Agent mistake=Theme
3	count	gerund				vn matter-91 «NP V»: agent=Stimulus
4	is a	present		the easiest thing	for Models = model	tree rules

Option 3 · ← Mova's choice

Models can't count letters at all

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	Models	model	NOUN	Number=Plur	4 count	nsubj
2	ca	can	AUX	VerbForm=Fin	4 count	aux
3	n't	not	PART	Polarity=Neg	4 count	advmod
4	count	count	VERB	VerbForm=Inf	root	root
5	letters	letter	NOUN	Number=Plur	4 count	obj
6	at	at	ADP	_	7 all	case
7	all	all	DET	PronType=Tot	4 count	obl

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	not count	can	Models	letters		all			vn multiply-108 «NP V NP»: agent=Agent letter=Theme

Option 4

Counting is only hard for people

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	Counting	count	VERB	Tense=Pres VerbForm=Part	4 hard	nsubj
2	is	be	AUX	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	4 hard	cop
3	only	only	ADV	_	4 hard	advmod
4	hard	hard	ADJ	Degree=Pos	root	root
5	for	for	ADP	_	6 people	case
6	people	people	NOUN	Number=Plur	4 hard	obl

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	count	gerund							vn matter-91 «NP V»: agent=Stimulus
2	is	present						for people; hard property	tree rules

Text: the tests of Mova's primary school, written for Mova (world/data/school/), as the course runner gave them to the student. Trees: Mova's own parse — the en model on the CPU, when the site is built — not checked by a teacher; they may be wrong. Frames: Mova's frames reader (prag), each text read on its own. Elamur Cosmic License.

This page in Markdown → https://md.mova.elamur.ai/free/en/corpora/questions/ai/29-who-you-are-being-wrong-and-learning/t3/