Test 3 · How a language model learns: guessing the next word

"bots are just parrots, they copy and paste, end of" Is that fair?

Mova's choice: option 1 · The right answer: option 1 · right ✓ · Who acts: bots, end · What happens: be, copy · To what, where: parrots, paste, end

This is test 3 of lesson 2 «How a language model learns: guessing the next word» 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

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	"	"	PUNCT	_	2 bots	punct
2	bots	bot	NOUN	Number=Plur	5 parrots	nsubj
3	are	be	AUX	Mood=Ind Number=Plur Person=3 Tense=Pres VerbForm=Fin	5 parrots	cop
4	just	just	ADV	_	5 parrots	advmod
5	parrots	parrot	NOUN	Number=Plur	root	root
6	,	,	PUNCT	_	8 copy	punct
7	they	they	PRON	Case=Nom Number=Plur Person=3 PronType=Prs	8 copy	nsubj
8	copy	copy	VERB	Mood=Ind Number=Plur Person=3 Tense=Pres VerbForm=Fin	5 parrots	parataxis
9	and	and	CCONJ	_	10 paste	cc
10	paste	paste	NOUN	Number=Sing	8 copy	obj
11	,	,	PUNCT	_	12 end	punct
12	end	end	NOUN	Number=Sing	10 paste	appos
13	of	of	SCONJ	_	17 fair	mark
14	"	"	PUNCT	_	17 fair	punct
15	Is	be	AUX	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	17 fair	cop
16	that	that	PRON	Number=Sing PronType=Dem	17 fair	nsubj
17	fair	fair	ADJ	Degree=Pos	12 end	acl
18	?	?	PUNCT	_	5 parrots	punct

#	verb	tense, mood	who	what, whom	other	roles by
1	is a in a quote	present	bots	parrots		tree rules
2	copy in a quote	present	bots = bot	paste		vn scribble-25.2 «NP V NP»: agent=Agent paste=Theme
3	is	present	end	end not said here	fair property	tree rules

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.

The options

Option 1 · ← Mova's choice · ✓ the right answer

Not quite; they mix patterns into new sentences, though they learned from people

The tree, 13 words

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	Not	not	PART	Polarity=Neg	5 mix	advmod
2	quite	quite	ADV	_	5 mix	advmod
3	;	;	PUNCT	_	5 mix	punct
4	they	they	PRON	Case=Nom Number=Plur Person=3 PronType=Prs	5 mix	nsubj
5	mix	mix	VERB	Mood=Ind Number=Plur Person=3 Tense=Pres VerbForm=Fin	root	root
6	patterns	pattern	NOUN	Number=Plur	5 mix	obj
7	into	into	ADP	_	9 sentences	case
8	new	new	ADJ	Degree=Pos	9 sentences	amod
9	sentences	sentence	NOUN	Number=Plur	5 mix	obl
10	,	,	PUNCT	_	13 learned	punct
11	though	though	SCONJ	_	13 learned	mark
12	they	they	PRON	Case=Nom Number=Plur Person=3 PronType=Prs	13 learned	nsubj
13	learned	learn	VERB	Mood=Ind Number=Plur Person=3 Tense=Past VerbForm=Fin	5 mix	conj
14	from	from	ADP	_	15 people	case
15	people	people	NOUN	Number=Plur	13 learned	obl

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	not mix	present	they	patterns				into new sentences	vn mix-22.1-1-1 «NP V NP PP.co-patient»: agent=Agent pattern=Patient into sentence=Co-Patient
2	learn	past	patterns = pattern					from people	vn learn-14-2-1 «NP V PP.source»: agent=Agent from people=Source

Option 2

Yes, every answer is copied word for word

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	Yes	yes	INTJ	Polarity=Pos	7 word	discourse
2	,	,	PUNCT	_	1 Yes	punct
3	every	every	DET	PronType=Tot	4 answer	det
4	answer	answer	NOUN	Number=Sing	7 word	nsubj
5	is	be	AUX	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	7 word	cop
6	copied	copy	VERB	Tense=Past VerbForm=Part Voice=Pass	7 word	amod
7	word	word	NOUN	Number=Sing	root	root
8	for	for	ADP	_	9 word	case
9	word	word	NOUN	Number=Sing	7 word	nmod

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	copy	participle							vn scribble-25.2 «NP V NP»: agent=Agent
2	is a	present	every answer	copied word				for word	tree rules

Option 3

No, bots invent everything with no help from people

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	No	no	INTJ	Polarity=Neg	4 invent	discourse
2	,	,	PUNCT	_	4 invent	punct
3	bots	bot	NOUN	Number=Plur	4 invent	nsubj
4	invent	invent	VERB	Mood=Ind Number=Plur Person=3 Tense=Pres VerbForm=Fin	root	root
5	everything	everything	PRON	Number=Sing PronType=Tot	4 invent	obj
6	with	with	ADP	_	8 help	case
7	no	no	DET	PronType=Neg	8 help	det
8	help	help	NOUN	Number=Sing	4 invent	obl
9	from	from	ADP	_	10 people	case
10	people	people	NOUN	Number=Plur	8 help	nmod

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	invent	present	bots	everything				with no help from people	vn create-26.4-1 «NP V NP»: agent=Agent everything=Result

Option 4

Yes, they are actual birds inside a computer

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	Yes	yes	INTJ	Polarity=Pos	6 birds	discourse
2	,	,	PUNCT	_	1 Yes	punct
3	they	they	PRON	Case=Nom Number=Plur Person=3 PronType=Prs	6 birds	nsubj
4	are	be	AUX	Mood=Ind Number=Plur Person=3 Tense=Pres VerbForm=Fin	6 birds	cop
5	actual	actual	ADJ	Degree=Pos	6 birds	amod
6	birds	bird	NOUN	Number=Plur	root	root
7	inside	inside	ADP	_	9 computer	case
8	a	a	DET	Definite=Ind PronType=Art	9 computer	det
9	computer	computer	NOUN	Number=Sing	6 birds	nmod

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	is a	present	they	actual birds		a computer			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/2-guessing-the-next-word/t3/