How PCR works
Changing temperatures IS important, but it's a bit more than that. Polymerase Chain Reaction, a key to most modern biotechnology, was invented in the 1980s, and won its somewhat eccentric California surfer-dude inventor (google Mullis) a speedy Nobel Prize as a result.
Most cells duplicate their DNA as they prepare to divide into two, so each resulting daughter cell gets its own full copy of the parental cell's DNA. PCR takes the DNA-copying enzyme from that process, and directs it to work on a specific stretch of DNA to duplicate that same stretch of DNA over and over again. Two short pieces of DNA (20 or so bases long), whose sequences match the DNA sequences at each end of the specific stretch, are included in the reaction to direct amplification solely to that stretch. Exponential growth is the bio-savant's friend in this case, as after the first duplication cycle, twice as many DNA copies of the given stretch are present; after the 2nd cycle 4 times as many copies are present, and so forth.
In theory, after 30 cycles, about 1 thousand million (2^30) times as many copies of the desired stretch of DNA are present, so that DNA from a small sample of hundreds or dozens of cells can be easily amplified, and even DNA from a tiny sample of just a few cells can often be amplified.
But what about changing temperatures? The brilliance of the process is that each step is performed without having to mechanically manipulate the DNA; rather just a temperature change is sufficient:
1) Initially, the temperature is raised to 95C or so to "melt" the DNA duplexes -- to separate the double strands one from the other so that the short DNA fragments and enzyme can access the now single-stranded DNA.
2) Next, the temperature is lowered to "anneal" the short added pieces of DNA to the main strand. These "oligos" will anneal or bind only to the sequence on the main strand that exactly matches their own, typically at temperatures in the range of 50C to 72C.
3) Finally, the temperature is raised a bit, usually to 72C, to allow the DNA-copying enzyme, starting from the end of the annealed oligo, to create a matching second strand for the stretch of interest.
This 3-temperature cycle is then repeated as many times as desired. The Panasonic system seems to take only 18 seconds per cycle, which is quite good; typical large-scale benchtop machines often take a few minutes per cycle. The more important advance represented by this system, though, is the all-in-one nature, able to take an unprocessed biological sample (e.g. the drop of blood) as input and to report SNPs present in that sample within an hour. As there are millions of SNPs throughout the genome, the practical efficiency of this system will depend greatly on how many *different* SNPs can be checked in parallel in a single run of the machine. If only a few or few tens can be checked, that is still useful, but if thousands or more can be checked in a single run, that would be quite impressive.