Um, insulin pumps have been around for about 30 years, and very common for the past 15. They're already a huge improvement over injections. This level of "artificial pancreas", though, not so much. The glucose sensing technology, though dramatically improved from its debut a decade ago, is still primitive: it uses interstitial glucose, and lags behind actual blood glucose, requiring regular calibration with fingersticks. Combine that lag with the fact that a non-diabetic pancreas starts producing insulin even before food hits the bloodstream, and it's impossible for the system to react to food in a timely way. Worse than that, though, is the fact that insulin effectiveness isn't constant. It's less effective when you're eating a lot of fat, and it's more effective when you're exercising (and for some time after). Just measuring glucose levels isn't enough to tell the system how to react. Maybe eventually they'll manage to combine this with enough other sensors to actually be a real artificial pancreas, but I think that's a lot farther off, and we'll probably have effective islet cell transplants without immune rejection by then.