Connect with us


Facebook Shares the Results of its First Deepfake Detection Challenge



With various politically-affiliated groups already using digital platforms to manipulate and influence voters, the rise of deepfakes is a serious concern, and could pose a major threat to democracy as we know it.

That’s why all the major platforms are working to develop systems to detect digitally altered videos, in order to catch them before they can spread. Twitter launched its ‘Manipulated Media’ policy back in February for this purpose, while Facebook has been looking at ways to advance its own detection models. In line with this, back in September, The Social Network issued a challenge to academic teams to come up with better deepfake detection models which could be used to weed out these videos.

And this week, Facebook has shared the results of its first Deepfake Detection Challenge.

As explained by Facebook:

The DFDC launched last December, and 2,114 participants submitted more than 35,000 models to the competition. Now that the challenge has concluded, we are sharing details on the results and working with the winners to help them release code for the top-performing detection models.”

This is a key point – Facebook, upon working with the winning teams, is looking to share the codebase for each of the winning models, while it’s also planning to open source the datasets used, in order to help advance research into deepfakes more broadly.

So how good were the winning models?

Facebook Deepfake Detection Challenge

The best performing detection models, from the thousands submitted, saw detection rates above 82%. Which is impressive – but that was based on the training set provided, which the researchers could study and refine specifically, focused on those examples. 

See also  Meta Files New Lawsuit Over the Sale of Fake Customer Reviews on Facebook

In order to determine the true accuracy of these systems, Facebook also tested the models on a ‘black box’  dataset of 10,000 video clips which the participants had not previously seen and had no access to before submitting their code. That altered the final results significantly.

The highest-performing entrant was a model entered by Selim Seferbekov. It achieved an average precision of 65.18% against the black box data set. Using the public data set, this model had been ranked fourth. Similarly, the other winning models, which were second through fifth when tested against the black box environment, also ranked lower on the public leaderboard. (They were 37th, 6th, 10th and 17th, respectively.)”

As you can see, the results changed a lot when they were applied to videos that the researchers could not train for specifically. That likely shows that there’s still a long way to go in establishing a truly accurate deepfake detection system – though a 65% detection rate is still significant, and would likely help to flag many potential concerns within the posting process.

Ideally, however, Facebook can get this number higher, and develop a better system for determining digitally altered videos before they’re shared. Because as we’ve seen, once a video is uploaded online, the fact that it’s determined to be fake or edited at a later stage is often too late to stop the damage being caused.

Already, within this US Presidential Election cycle, we’ve seen several examples of videos being edited or changed in order to emphasize certain elements. There was the controversial Nancy Pelosi video, in which Pelosi appeared to be slurring her words, the Michael Bloomberg video where he pressed other candidates on their business credentials during a debate, and the Joe Biden clip which had been edited to show Biden saying that people should vote for Donald Trump.  

See also  Op-Ed: Facebook page disappeared? There IS a simple fix

These videos were not advanced deepfakes, they all used fairly basic editing techniques. But each of them sparked significant debate, despite them being heavily edited, and proven to be so. Even when they were revealed to be edited, the debates carried on. You can only imagine the damage that a convincing enough deepfake could do within that same process.

And we are indeed likely to find out just how much damage deepfakes can do. As the 2020 US Election race heats up, it seems increasingly likely that, at some stage, a deepfake video of some kind will come into play. 

How will that change the race? How will it alter voter behavior? Can digital platforms detect and eliminate such before it takes hold?

It could just be that in the wake of the election, a deepfake video might be the central focus, much like Cambridge Analytica became the target after 2016. Facebook’s working to avoid that outcome, and it could end up bring a crucial effort. 

You can read more about Facebook’s Deepfake Detection Challenge here.


Does Posting Memes on Social Help to Increase Traffic to Your Website? [Study]



Does Posting Memes on Social Help to Increase Traffic to Your Website? [Study]

Does posting memes help to increase traffic to your website?

This is a key question, which really relates to all kinds of engaging social media posts – because while these types of trending updates very clearly garner Likes and comments, do they actually benefit the stats that really matter to your business?

I mean, Likes and followers are great, but what you need is conversions, relative to what that means for your business. For SMT, we’re working to get as many people to read our posts as possible, and as you’ve likely noticed, we’ve recently been trying out memes as a way to boost engagement, and see what that gets us in this respect.

So what have we found? Here’s a quick overview of the initial results of our meme experiment.

First, a quick bit of background…

We’re always looking to try new things, and test out the latest trends and processes, and not just because it might help us generate more traffic and build community, but also, because that’s what we write about. If we’re going to write about it, we need to know and understand it as much as possible, in order to ensure that what we’re communicating is correct, and makes sense for our audience.

In this respect, we’re always testing new approaches, apps, tools, etc.

In terms of posting, last year, we tried out polls on Twitter and LinkedIn, and question posts on Facebook, to see if they would help drive more engagement. And they definitely did – these types of audience-prompting updates garnered a heap of Likes and comments. But when we cross-checked this against Google Analytics tracking, we didn’t see a big uptick in sessions or users visiting the site.

That’s not to say that these aren’t valuable, but they weren’t shifting the needle in any significant way on our key metrics. At the same time, too many polls can get annoying. In our experience, they’re an interesting tool to use, in moderation, but not a massive driver of our ultimate aims.

See also  Twitter bots and memorialized users will become ‘new account types’ in 2021

Yet, at the same time, our social traffic, like all platforms, tends to have ups and downs – and in a down period this year, we decided to try something new to freshen up the feed and give our audience something else to engage with, and maybe lighten the mood a little at the same time.

Enter memes

The inspiration in this instance came from SEMRush, who’ve also tested out memes as a means to boost engagement, and build community.

SEMRush’s experience saw them significantly increase their social engagement by posting timely, on-trend, niche memes. So we thought we’d give it a try, to see if that helps drive more interest in our articles.

The first step, of course, is creating relevant, engaging memes. Which is not always easy. Many of our memes never made it out of test phase, with some clearly failing when viewed in the templates.

Some that we’ve posted also haven’t connected in the way that we’d hoped.

But this is the game – if you’re going to post memes, you’re going to have hits and duds, and you just have to live with it. I imagine it’s the same as a comedian, some of the jokes work, some don’t. But ideally, more of them get a laugh than not.

Which, luckily, our memes have.

On average, the memes that we’ve posted are generating around 135 Likes on Facebook, which has helped them generate significantly more reach than our average post, while they’re also performing strong on both Twitter and LinkedIn.

And they’re fun. The way I view them is like the comic section of the traditional newspaper, a light-hearted moment between the news updates and informational elements.

The increased engagement obviously has some benefits in boosting algorithmic reach (if people engage with one of your posts, the algorithm is more likely to show them more), as well as building community around the SMT brand. But the key question is – ‘do they actually get more people clicking through to the site?’

See also  Clubhouse Looks to More Private Sharing as it Seeks to Regain its Growth Mojo


In our case, when comparing our overall social media performance against the previous 3 months before we started posting memes, we’ve seen a 12% increase in sessions from social, and a 16% increase in users.

That’s not a massive shift, but when you’re working with the ebbs and flows of referral traffic, as well as changes in analytics due to shifting data regulations, any increase is positive, and a double-digit jump is definitely worth the effort.

This is only around a month of data, so it’s not definitive, and there are also other factors to consider that could influence the results. But the numbers, thus far, suggest that it is worth sticking with – and as noted, it’s fun too, adding a little more relatability to our presence, as opposed just the latest news.

A few other notes:

  • Some commenters are going to take your memes literally, no matter it is that you post. There’ll always be a couple of comments like ‘well, actually, the reality is that…’ Yes. We know. These are not meant to be literal, they’re a moment of light-heartedness in amongst our regular, marketing strategy-focused news updates.
  • We’ve found that more general memes work better than trending ones. A couple of memes where we’ve tried to tap into news events, like the changes to Twitter verification, haven’t done as well as jokes about more common social media marketing experiences. This also, of course, relates to the memes themselves, and whether they’re actually funny, but in several examples, trending topics haven’t been as big a hit.
  • Every meme is a bit of a risk. You’re trying to find commonalities with your audience, and some things that you might think are common might not resonate. You need to know your niche, and know your community, which takes some experimentation – and a lot of research (I’ve been writing about social media trends for eight years)
  • One guy on LinkedIn keeps saying that he’s envious that we’re able to get these memes approved by management. For those that don’t know, SMT is an editorial team of two (2) people. Approval, in this sense, isn’t exactly a barrier.
  • Does it take a long time to come up with them? Not really. We usually do them in batches or around 10 at a time, then schedule them out on different days/times across FB, Twitter and LinkedIn. We can make 10 or so in, maybe, a couple of hours, once every week and a half or so. Not a major commitment.
  • We’re currently scheduling around one meme a day on each platform, again, taking that newspaper comic approach. Maybe we miss a day here and there, but that’s the general aim, as something to keep that engagement flowing, and keep the entertainment value up.
See also  Meta Launches New Creator Monetization Initiatives, Including More Reels Payments and NFTs

Look, it’s not necessarily a walk in the park to keep coming up with funny memes – and it may be that we run out of ideas at some stage and suddenly it becomes a lot more difficult. It’s also not for everyone. Coming up with a (relatively) clever joke that fits a meme template doesn’t always come easy, and there are days when you just don’t have it, no matter how hard you stare at the screen.

But for a minor time commitment, it does seem, at least at this stage, like this may be a good way to help engage your audience, which can also drive direct traffic benefits.

We’ll post another overview of our meme experience three months in.

Source link

Continue Reading

Subscribe To our Newsletter
We promise not to spam you. Unsubscribe at any time.
Invalid email address