The technology used for the Hong Kong head counts involves both A.I. and some pretty fancy software. The New York Times has an article which describes the setup in detail and how it worked. It’s simple, it’s easy, and seems to be effective in delivering a head count.
Paul Yip, professor of social sciences at Hong Kong University, has been studying the Hong Kong protests in depth since 2003. In combination with Texas State University and C&R Wise AI, Yip conducted what may well be the very first A.I. study of crowd numbers.
(Please note: There are multiple technologies involved in this count. It’s important to understand the techniques used to count, and why these techniques and technologies are critical to accuracy of counts. Do please read The New York Times article for this information, which I don’t want to garble in translation.)
The politics of numbers
“How many people?” is the big issue in evaluating the strength of protests. The Hong Kong protests were particularly sensitive on the subject of numbers. The Hong Kong police downplayed the Hong Kong protests, suggesting that there were only 190,000 protestors. The organisers of the protest said there were 500,000. Independent estimates were roughly in the 200,000 range.
Estimates were always going to be difficult. Large numbers of people were moving through narrow streets, making any sort of count very hard to manage, whatever means of counting was used. According to the A.I. there were about 250,000. You can see the discrepancies. The police estimate was 25% under the numbers as recorded by the A.I. The various numbers from estimates not using A.I. can be attributed to less advanced technologies, and/or political reasons for under or overestimation.
The significance of numbers is both very important and very superficial, depending on which estimate you’re talking about. The underestimation is the usual way of devaluing the importance of a protest. Overestimation can be based on a similar desire to upvalue the protest, optimism, or simply bad information from organisers whose counts tend to round up.
Technologies make a big difference
The thing that may bother many people looking at how heads are counted is that it visually resembles facial recognition technology as used by security agencies.
A few things to note:
• This isn’t exactly the same as security technology, although it does have common components. This particular exercise didn’t include facial recognition, although it easily could have. This tech can even count objects, so the degree of detail is very high.
• Security biometrics and head counts are different ball games. The Hong Kong protests infuriated the Chinese government. It’s quite possible Chinese authorities may have used facial recognition in monitoring the protests. This sort of data can also be repurposed and fed in to surveillance networks. Think of it as an “audit” for tracking people, both groups and individuals.
• This is definitely a game changer for monitoring real numbers at protests and elsewhere. Imagine if the very high political spin events like the UK Brexit protests, the US Presidential inaugurations, or other major turnouts could be accurately counted. These very high profile political events have always been pretty kind to themselves in numbers attending. There’s always a “record crowd”, at least in the minds of the PR guys and organisers. That may change, drastically, and not before time.
The future, maybe too soon
You can be absolutely sure that this class of technology is just the beginning of a whole new way of counting people around the world. It can be used in shopping centres, rallies, etc. and do a lot to add meaningful stats as well as possible spin on the numbers. Its uses may be good or bad, but the days of seeing what you want to see with your own numbers are about to disappear.
