The new riddle of induction

Suppose every blueberry you've observed to date has been blue. You take this to be evidence for:

(H1) All blueberries are blue.

As a consequence, you predict:

(P1) The first blueberry I observe on February 11, 2012 will be blue.

Now, define "bleen" to mean "blue until February 10, 2012--green thereafter". By this definition, every blueberry you've observed to date has been bleen. You take this to be evidence for:

(H2) All blueberries are bleen.

As a consequence, you predict:

(P2) The first blueberry I observe on February 11, 2012 will be bleen.

A blueberry that will be bleen on February 11, 2012, though, will be green--not blue. Thus, (P2) flatly contradicts (P1). Is there any evidence that favors (H1) over (H2)? We might complain that (H2) is couched in terms of a derivative property--"bleen" is defined in terms of "blue" and "green". We might take this observation to favor (H1), and therefore (P1).

But define "grue" to mean "green until February 10, 2012--blue thereafter". Imagine a culture that only understands "bleen" and "grue"--not "green" and "blue". To them, "green" means "grue until February 10, 2012--bleen thereafter", while "blue" means "bleen until February 10, 2012--grue thereafter". Their complaint about (H1) is that it is couched in terms of a derivative property--"blue" is defined in terms of "bleen" and "grue". They take this observation to favor (H2), and therefore (P2). Could we really be right, and they really be wrong?

What else could favor (H1) over (H2)? If nothing, isn't this a problem for every hypothesis of the form "All X are Y"? And don't we (implicitly) make predictions founded upon such hypotheses in our everyday reasoning? Doesn't this problem undermine the very way in which we learn from our observations and experiences?