Why You Should Open A Roth IRA Today
As I wrote just over a year ago, it’s no secret that Washington doesn’t like data and isn’t good at innovation. From data-driven studies on government waste that are allegedly buried because they show too much waste to a federal government healthcare website projected to cost three quarters of a billion dollars when it cost Apple just $150 million to create the iPhone to contracting processes that discourage out-of-the-Beltway innovation, there is no scarcity of examples of government failing when it tries to do technology. It certainly doesn’t have to be this way – Estonia has become synonymous with government innovation, transforming into one of the most technologically advanced governments in the world where filing taxes takes less than five minutes and just the click of a button and you can buy a fishing license from your phone in the middle of the forest. So why does the US, home to many of the world’s best-known technology brands, struggle when it comes to infusing that sort of innovation into the operation of government?
Perhaps the biggest driver of this innovation failure is the failure of many leaders in government to recognize and accept that they need to modernize their approach to data and their failure to understand the power of technology to address those needs. Many appear entirely isolated from the advances in basic data analytic technology, let alone the kind of bleeding-edge capabilities coming out each day from Silicon Valley.
Speaking to one agency last fall I was stunned to have several senior data scientists assert there were no technologies in existence today that could take a photograph and generate a set of basic topic tags describing what the image depicts. I asked if they were merely criticizing the accuracy of current systems, but no, they asserted, the technology to assign basic tags to an image based on its visual content simply does not exist at all. I quickly pulled up a few deep learning image recognition tools and showed the results for some randomly selected images from the web. There was stunned silence. It simply cannot be that in 2016 there are data scientists who don’t even know there is something called deep learning and that it can make at least basic sense of photographs. Even if you don’t know how to program a convolutional neural network and even if you don’t understand the most rudimentary basics of how they work, if you work as a data scientist in today’s world you have to at least know that such a thing exists and that there are tools, however imperfect and limited, that can catalog images.
Similarly, in a meeting earlier this year with several highly-touted data scientists from a different agency, I sat through a presentation on how making word histograms (ngrams) out of piles of PDFs was going to completely revolutionize how the government does business by allowing it to see which words were the most popular and making those into word clouds and that this has never been done before by anyone. And this was not the only agency that touted such simplistic approaches as state-of-the-art – I’ve received similar briefings at other agencies. While ngrams are certainly useful in certain circumstances, they are neither state-of-the-art nor do they permit the kinds of high-order analyses the data scientists were touting them for.
To make matters worse, several weeks later I attended a talk by a senior government official who touted the amazing big data efforts a group under him was doing and how they were literally putting Google to shame with all of their incredible “big data” research. The only problem is that the group he was touting was the one I had visited that was making ngrams of PDFs. Of course, like so many government leaders I’ve spoken with, he seemed to genuinely believe that the work of his division was truly state of the art and far beyond anything Silicon Valley could dream of.
To be fair, there are certainly pockets of true innovation in government, but the problem is that those are the groups that truly understand technology and speak in the carefully caveated language of an experienced technologist, rather than the detail-free hyperbole of a salesperson. Instead of touting their work as the next Google that is going to make everything else on the planet obsolete, true government innovators talk about the specific areas of government they think their approach can help make more efficient or more secure or better serve the American people. Unfortunately, their voices are all-too-often drowned out by all of the hyperbolic shouting from the others who have nothing beyond the volume of their voices to tout.
Indeed, while publicly lavishing praise upon a new initiative as transforming government, behind those public plaudits there is often a very different story. I still remember when data.gov was launched to great fanfare amid breathless proclamations that it would forever change how government made its data available to its citizenry, creating a single central repository for all open datasets. I had long asked why the government didn’t simply evolve over the years to make its data available in the major repositories where the open data community is most active, like GitHub, instead of building its own proprietary platforms, but the answer was always that government knew best. Throughout my interactions with the government data science world, I would often hear rank-and-file data scientists echoing the need of government to go where the developers are, rather than asking developers to adapt to government, while more senior…