thinking is dangerous


La computadora sabe mas que tu.

From Science: Rise of the Forecasting Machines

For almost half a century, human and computer have been vying to predict the weather better. The computer long ago won the race to forecast out to a week and beyond, and human forecasters began giving ground at shorter ranges. Now comes evidence that computer simulations--aided by automatic statistical analysis--can consistently best humans at forecasts longer than 24 hours. The finding heralds an age of prognostication untouched by humans.

In most cases, Mass and Baars argue, human forecasters should spend most of their time on the first 12 hours or so. The rest of their time could be spent making sure the model and MOS are not making any blunders at longer ranges. Michel BĂ©land, a director at the Meteorological Services of Canada in Dorval, Quebec, says that Canadian forecasters have already pulled back. They now primarily focus on severe and high-impact weather expected over the next 18 hours. Beyond that, the machines stand watch.

Discussion questions:

1) Do these mechanized prediction systems have knowledge? I assume that their 'claims' are justified (by the accuracy of the model they use to predict), so this question is, basically: Does the machine make a claim about the weather?

2) Are we justified in treating the system as an authority? I assume that the answer to this question is independent to your answer to the above question.

3) Does the machine understand the weather? I posit that this question reduces to (1) as follows: the machine understands the weather iff it uses the model to make claims about the weather that are more or less accurate. So the system understands the weather insofar as we can interpret the system's behavior as "using the model" and "making claims".
14:17 :: :: eripsa :: permalink