The article covers some interesting issues in manufacture - for example the distinction between the small-area high-value silicon structures manufactured for electronics, and the need for large-area low-cost structures for photovoltaic energy generation. New manufacturing processes are probably key to bringing costs down further. He includes an interesting graph (based on data from this presentation by Robert Margolis of Carnegie Mellon) of costs vs. manufacturing scale, which also includes (as points on the cost line) the progress by year as well:
One interesting feature of this graph is the rapid progress that can be seen until about 1983, while US funding for solar PV R&D was strong; manufacturing cumulative volume expanded from about 0.4 to 30 MWp between 1976 and 1983, and costs dropped from about $70/Wp to just over $10/Wp. The cost decline of those 7 years has only almost been matched in the more than 2 decades since 1983, with another 2 orders of magnitude increase in volume. It's clear incentives to increase production volume could have a very positive effect; R&D funding couldn't hurt either.
Once PV costs reach a threshold probably between a factor of 3 and 10 lower than current prices, they will be actually positioned to economically transition away from fossil fuels in the manner of any change to improved technology (LP to CD, say). Margolis talks about this in his article too - as a "buy-down" cost to reach a certain cost level, though his displayed targets are higher than the fossil-fuel-competitive level. From this graph it looks like 100 GWp - 1 TWp may be needed; total investment several hundred billion to $1 trillion.