Yoshua Bengio - The Consciousness Prior (2017)

Context

Learned in this study

Things to explore

Overview

• Represent conscious states as low-dimensional vectors
• Conscious states are taken from high-dimensional unconscious states
• It is likely that some form of RNN will be used to learn how to extract the conscious states from the unconscious state

Notes

1 Introduction

• Consciousness as defined by Locke: consciousness is "the perception of what passes in a man's own mind", or awareness of an external object or something within oneself

2 Consciousness Prior Theory

2.1 Subnetworks

• Let $s_t$ be the observed state at time $t$ and let $h_t$ be the high-level representation derived from $s_t$ (and from past observed values $s_{t-k}$ in the partially observable case)

  $$ h_t = F(s_t, h_{t-1})$$

• The conscious state $c_t$ is defined as a very low-dimensional vector which is derived from $h_t$ by a form of attention mechanism applied on $h_t$, taking into account the previous conscious state as context

  $$ c_t = C(h_t, c_{t-1}, z_t)$$

  where $z_t$ is a random noise source

• The cognitive interpretation is that the value of $c_t$ corresponds to the content of a thought, a very small subset of all the information available to us unconsciously, but which has been brought to our awreness by a particular form of attention which picks several elements or projections from $h_t$
• The function $C$ is the consciousness RNN and because of its random noise inputs, produces a random choice of the elements on which the attention gets focused

2.2 Training Objectives

• To capture the assumption that a conscious thought can encapsulate a statement about the future, we introduce a verifier network which can match a current representation state $h_t$ with a past conscious state $c_{t-k}$:

  $$ V(h_t, c_{t-k}) \in \mathbb{R}$$

• $V(h_t, c_{t-k})$ indicates the consistency of $c_{t-k}$ with $h_t$, e.g., estimating the probability of the corresponding statement being true, given $h_t$
• We would like to define an objective (or reward) function which embodies the idea that the attended (conscious) elements are used in some way whose value can be quantified and optimized
• There are two distinct mechanisms at play which contribute to map the high-level state representation to the objective function
  • the attention mechanism which selects and combines a few elements from the high-level state representation into a low-dimensional "conscious sub-state" object
  • the predictions or actions which are derived from the sequence of these conscious sub-states

See also

References

• Bengio, Yoshua. "The Consciousness Prior." arXiv preprint arXiv:1709.08568 (2017).