Hey Friends,

I'm sure we are all familiar with the old adage 'The Camera never lies'.

While digital photography and photo editing have long resigned this to a bygone age, a new phenomenon is on the rise which takes this to a whole new level. This week you may have seen a video of Tom Cruise playing golf, and talking about a meeting he had with Russian leader Mikhail Gorbachev. It looks like Tom, sounds like Tom and I have to say even the mannerisms are pretty convincing, but it is in fact an example of a Deepfake; a fake video created using 'deep learning' (a branch of Machine Learning/Artificial Intelligence) to map the face and mouth movements of one person and transpose these on top of a video of another person.

Of course, this is not a new concept, but the use of 'deep learning' and the availability of data to train the models is what is causing so much concern. Often the code used to generate the fake videos is based on open-source libraries, so when it is released on the Internet it is very easy for anyone to replicate the results. Also, since most people (this mailing list excluded) have no aversion to posting videos/images of themselves on social media it is now easy to compile enough facial image material to train the deep learning models by simply scraping images from databases such as Google, YouTube, stock photo sites etc.

Ok, fair enough I hear you say. So its pretty easy to create a deepfake, but they are just a bit of fun right? There isn't anything to really worry about?

Well here are a few things that have people are worried out:

• We have already seen a rise in deepfakes of celebrities appearing in pornographic material (I'm using the royal we as I obviously haven't watched them). This trend is likely to continue.
• Although it hasn't happened yet there is real concern that they could be used to influence the opinions of people during elections.
• With the rise of social media information now spreads at an unstoppable rate. Even more alarming studies have shown that people unknowingly spread fake news more quickly than real news.
• Deepfakes could be generated to cover up or deny past actions.
• Countermeasures may infringe upon free speech.
• Objective truth becomes more difficult to establish. If we need arbitrators of truth, who and how do we choose these.

While the thought of deepfakes may be an obvious cause for concern discussions about combating deepfakes are already taking place. Solutions range from legislation to technical solutions such as using the algorithms against themselves i.e. if you can train an algorithm to create deepfakes then given enough training data you should be able to train it to spot them from genuine videos.

For now, if you see a video of me doing or saying anything strange I want to know ..... 'it's fake; it wasn't me' (honestly governor)

Have a great week.

Take care,

Chris

PS: If you think someone you know might benefit from joining this conversation, please forward them this email. Or has someone already forwarded this email to you? If so, Subscribe here

Things I've found interesting this week

\\ One Link: How deepfakes work

\\ Two Link: Tom Cruise at his best (fake news)

\\ Three Link: Sinead O'Gollum ..... must be a long lost relative of Sinead

Favourite Quote

People need to question what they're looking at and it's important for journalists to confirm their sources and where they got it - Chris Ume

Previous emails

If you have missed my previous emails, catch up here