Digital Serendipity?

Some fascinating articles and books at the moment about the long term effects of the web's move towards personalization. On the one hand it means the web will deliver more of what you want, the problem is, getting what you want isn't the same as getting what you need. The more our sources of information and inspiration mould themselves around what we already are, the less opportunities we have to grow into something new. The phrase of the moment is 'serendipty' - the chance stumbling on something unsought, but well found.

There is indeed I think a lot to the idea that ever more impressive personalization will limit this, and some have gone so far as to say that a digital world, being definite and algorithm based, just can't provide the distractions and abstractions needed to stumble accidentally on something. But while not saying the current structures won't end up being so restrictive, in my vieww they don't need to. Any truly complex system is open to chaos-style behaviour, where small changes in input result in wild variations in output, and the internet, and even the search and match algorithms of the likes of Google and Facebook, could surely display, or be allowed to display, such behaviour.

Indeed, probably serendipity itself results from accidental pattern matching, where we are thinking of one kind of thing and suddenly spot something similar in a completely unrelated field, achieving a flash of insight which is maybe really a flash of recognition, a realisation of how theold can lead to the new; and pattern matching is what Google etc. are good at. Up to now they have probably been trying to emulate perfect 'directed' pattern matching, focused in one area, but surely there is also the potential (if not probability) for cross-field matching. If for example Google Translate is trained to match words across languages, what would happen if it was fed with a different data set, and tried to extract similar patterns in this new domain. Is this not exactly what the serendipitous brain does?

In fact, what is often portrayed as a bad thing, that even the very people who design and operate these systems are losing true insight into how they work, provides hope, since they may become semi-evolved 'intelligences' in their own right (and i only mean this in so far as excellend pattern matching machines), and provide us with unexpected insight - since what they produce will be unexpected.
So, when it comes to digital serendiptiy, as Google itself suggests, i'm feeling lucky.