Artificial intelligence (AI) systems are already far superior to humans in
some tasks, such as playing Go or processing enormous amounts of data, yet
even just a few months after we are born, AI is still far behind us in many
other areas.
For instance, even very young kids instinctively understand that an object
shouldn't disappear and then reappearance somewhere else. Babies react with
amazement when they witness such a magic trick.
But an AI hasn't been able to easily understand such a straightforward
continuity rule, together with other fundamental physical rules. Currently,
a recent study introduces PLATO, an AI that was motivated by studies on how
infants learn.
PLATO, which stands for Physics Learning through Auto-encoding and Tracking
Objects, was taught using a set of coding-enhanced films that were intended
to mimic the early information that infants possess.
Fortunately for us, developmental psychologists have spent decades
recording the many components or concepts that go into physical knowledge,
according to neuroscientist Luis Piloto of the UK-based AI research facility
DeepMind.
"We created and open sourced a physical concepts data set, extending their
work. To evaluate physical notions in our models, this synthetic video data
set draws influence from the original developmental trials.
We all develop an early understanding of three fundamental ideas:
permanence (items won't abruptly vanish); solidity (items can't pass through
one another); and continuity (items move consistently across space and
time).
These three notions were included in the data set that the researchers
created, along with two more: directional
inertia
and unchangeableness (the idea that an object's attributes, such as its
form, remain constant).
Videos showing balls bouncing off of one another, dropping on the ground,
vanishing behind other objects, reappearing, etc. were used to illustrate
these ideas. The next stage was to put PLATO to the test after training it
on these videos.
When PLATO was given recordings depicting 'impossible' events that went
against the rules of physics it had learnt, it exhibited astonishment (or
the AI equivalent of it) because it was intelligent enough to realize that
something strange had occurred that violated the laws of physics.
This occurred after just brief training durations, in some cases only 28
hours. Technically speaking, the researchers were seeking for
violation-of-expectation (VoE) signals that would indicate that the AI
grasped the principles that it had been taught, exactly like in newborn
studies.
"Our object-based model displayed robust VoE effects across all five
concepts we studied, despite having been trained on video data in which the
specific probe events did not occur," the researchers write in their
recently
published work.
Using items other than those in the training set, the team conducted
further experiments. Again, PLATO demonstrated that it could learn and go
beyond its fundamental training knowledge by exhibiting a clear
comprehension of what should and shouldn't be happening.
PLATO, however, is still not nearly at the level of a three-month-old
infant. When AI was presented scenarios without any objects or when the
training and testing models were comparable, there was less surprise.
Additionally, the movies that PLATO was trained on had additional
information to aid in its recognition of the objects and their
three-dimensional movement.
To fully understand the situation, it appears that some innate information
is still needed, and developmental biologists are still perplexed by the
"nature versus nurture" dilemma in newborns. The study may improve our
comprehension of the human mind and aid in the development of more accurate
AI models of it.
"Our modeling work provides a proof-of-concept demonstration that at least
some central concepts in intuitive physics can be acquired through visual
learning,"
the researchers write.
Although studies in certain precocial (born in an advanced condition)
animals imply that some fundamental physical notions might be present from
birth, the findings indicate that in humans, intuitive physics knowledge
arises early in life but can be influenced by visual experience.
The research has been published in
Nature Human Behavior.
