Overall statement

The article investigates a new design for an on-chip photonic synapse. They show that it is fit for purpose and study the repeatability of its switching. They go on to make hyperbolic claims about the device being a rival to spinnaker or the human brain but they fail to explain how the two lasers per synapse will be connected, how the 250 deg C for 10 minutes recrystallization process will be integrated. They also state that 1 W of power is required to drive each synapse making its power consumption significantly higher than that of the human brain, spinnaker or any computing system.

Strengths and impact

The strengths of the article are that it shows a new device repeatable performing its expected function. It’s impact will be large if efforts into integrating the device into a larger architecture with much decreased power consumption are successful.

Weaknesses

Major points in the article which needs clarification, refinement, reanalysis, rewrites and/or additional information and suggestions for what could be done to improve the article.

1.10 minutes of annealing at 250 deg C is not practical for resetting a computing device. How can this be eliminated or reduced ?

2.Defining a different baseline transmission for each device will not work in a large scale computer architecture, what are the results if a fixed baseline is used for all devices ?

3.The claim about having an operating power approaching that of the human brain for a system on the scale of spinnaker is invalid considering that each device requires two lasers to operate and requires 1 W of power. please redact this claim.

4.The hyperbolic language in the abstract should be removed “intense” “incredibly”.

5.The lines ending in reference 1, 2 and 3 are vague. We don’t understand how emotions are manifested or computed on any hardware so how can we know computers are bad at computing emotions. Remove or clarify these statements.

6.Yes, it is interesting to build an optical synapse but it isn’t acceptable to claim it has no electrical power loss without specifying its optical power loss. Please specify optical power loss of device.