Artificial intelligence is right up there with robots taking over our jobs. This is the first in a series on how big tech like Facebook uses AI to manipulate you.
The number of AI applications has increased rapidly. We speculate and marvel about what AIs will be able to do in the future. But what we don’t realise is that AI has already had a huge impact on the goods and services we use every day.
We covered ‘Can you trust Facebook’ here and the answer is a deafening NO. Now we look at how Facebook uses AI to manipulate you. Yes, if something is free, the product is you.
Facebook says it brings people closer together. That is bloody rubbish. It is first and foremost one of the world’s largest covert data harvesters in the guise of making social connections easier.
So, we asked Bernie Brode, a US expert on AI that has spent a lifetime delving into the inner workings of cryptography and now the confluence of nanotechnology and artificial intelligence. He is eternally curious about where that ‘intersection’ will lead us. Bernie writes
What is AI?
The simplest answer is to use a computer to make decisions and draw conclusions that mimic what a human could do – only a hell of a lot faster and with Everest’s of data.
It is based on machine learning – giving a computer a problem and basic instruction on how to solve it. By sheer processing speed, it arrives at the right answer. For example, machine learning in medicine enables a computer to faster recognise cancers tumours simply by recognising healthy tissue first.
That is a simple example, but AI is no more omnipotent, devious, or evil than we are. We created it to serve us and answer the questions we ask it. And that brings us to Facebook and what questions Mark Zuckerberg et al. ask it to solve.
How Facebook uses AI to manipulate you
Facebook was one of the earliest, large-scale adopters of AI and Deep Learning systems. This should come as no surprise: 1.2 billion people upload 136,000 photos and update their status 293,000 times per minute.
Until recently, Facebook only derived value a.k.a. monetised it, from a tiny fraction of this largely unstructured data. Unstructured data is ostensibly data with few obvious threads between it, and we will come back to that later because that is the creepier part – the evil use of AI.
1. Textual Analysis
Facebook’s oldest and most developed AI system is DeepText. This system uses neural networks that ‘read’ through the billions of posts and tries to understand their meaning. It is a highly autonomous system that not only processes text automatically but teaches itself.
This system is the most important use of Facebook’s AI. It is the primary means to serve ads. By scanning through all of your messages, DeepText is works out what kind of products you could use and serves you with relevant ads. More on that soon.
2. Facial Recognition
DeepFace is another Facebook developed AI system. It is far more controversial and intrusive than DeepText. It recognises faces in photographs and is eerily accurate. According to Zuck’s crew (his PR people prefer we call him that as it supposedly humanises him), it can recognise 97% of the time if two photographs are of the same person. Humans can only get it right 96%. Uh oh.
Critics argue that given a high-resolution photograph of a crowd, Facebook could use the system to identify who was present at a demonstration. For this reason, the European Union forced Facebook to remove the system from EU citizens’ accounts in 2013, but Facebook has recently challenged this.
3. Targeted Ads
We referred to DeepText earlier as its main way to target adds – reading and analysing your posts. Well, it uses DeepFace as well to start to tie that unstructured data together and contribute to your ‘profile’.
Unfortunately, details on Facebook’s newer AI systems are not widely available as information about DeepText and DeepFace. Why? It seems tAd targeting technology is too commercially sensitive for wider release.
What is apparent are some of the unfortunate consequences of letting AIs target ads. For example, Facebook ads target vulnerable people.
It can draw conclusions that users do not want their friends or family to know about. For example, AI deduced that a couple was having a baby soon, started advertising all manner of baby stuff to them, their parents and friends. It turns out that the couple was looking for baby stuff for a friend. Oh, and let’s not mention the folly of looking up erectile dysfunction, haemorrhoids or halitosis or you will be the target bombarded with cheap drug ads.
Or Facebook’s so-called research.
A $20 monthly payment went to anyone who downloaded and installed a ‘Research’ app that skimmed their website searches, location, and private messages.
Or the Onavo VPN debacle. Facebook offered a free VPN to protect users online. Except that instead of keeping no logs of user’s web surfing enabling complete anonymity, it allows Facebook to use AI to analyse all your surfing habits granularly. Once caught out, it reverted to placing hidden transparent gifs on websites, cookies on your device and paying other data harvesters to get your data.
By the way, this nefarious activity costs you money. These ads are not without a cost to the user either. They are an increasingly large portion of traffic on the site. A 2106 study found that mobile ads use more than 75% of the bandwidth, and almost half of the data, required to use Facebook.
4. Using More AI
Facebook is using AI to develop MORE ways to use, well AI. Its called Flow, which uses the same kind of Deep Learning technology to assess other AI and machine learning systems. The goal here is for this AI to eventually build more AIs, which can then build more AIs. You get the idea.
If you think that sounds scary, you’re right. The ability for intelligent machines to build more intelligent machines has been a staple of science fiction for decades. Well thanks to Mr Data, sorry Zuck, it appears we are on the verge of this new reality.
How Facebook uses AI to manipulate you – the future
Facebook remains suspiciously quiet about its plans for future AI use. It has hinted that it might use AI systems to roll out functionality that none would disagree with, such as
- providing audio descriptions of images for the visually impaired
- rolling out internet access to poorly served parts of the world
- sniff out and take down fake Facebook accounts
- Monitor fake or inappropriate content (sure that is working!)
It’s probably safe to assume, however, that these selfless plans aren’t the only possibilities that the company is investigating.
Keep in mind that Facebook users little privacy protections to rely on. Why? Simply logging onto Facebook implies consent that your data will go through some of the most advanced (and creepy) AIs on the planet. As we have repeatedly argued, either #DeleteFacebook or at least be aware of how Facebook uses AI to manipulate you.
Just in time, the United Nations Educational, Scientific, and Cultural Organization (UNESCO) recognises an ethical approach to AI is desperately needed. It has released a draft resolution that you can read here. Rest assured, this is one issue worth paying attention to.
AIs burgeoning ability to intrude on your privacy and manipulate how you feel, how you vote and what you think is just plain wrong. its time to fix Fakebook.