AI in Egypt, AI animation and building a GPT of yourself
I spoke at the Egypt Media Forum this week about where journalists are headed next
The Egypt Media Forum is a massive, sprawling, exciting journalism conference which takes place in the centre of Cairo at a fancy hotel with close to 2,000 delegates. I’m writing this while looking directly over the Nile River. Last year I was at this conference giving a workshop on investigative podcasting and this year I’m here to speak about what journalists should be learning in the age of AI.
On Sunday I was on a panel with Dean Arnett (from the UK), Maitha Al-Suwaidi (from the UAE) and Alaa Al-Ghatrefy (from Egypt) with moderation by Dina Zahra (a news anchor also from Egypt). Here are a few takeaways:
We agreed that the best case scenario was that journalists could use AI to lower the time needed for certain tasks and so be free to conduct better investigations and deeper fact-checking.
However, the worst case scenario stems from this same issue: AI makes the production side faster and so it is assumed (by bosses and editors) that fact-checking should also be quicker, while this human element to question the authenticity and accuracy of information needs to stay, at least in part, analogue. Verification is actually going to become more taxing, mainly because of the deluge of AI generated disinformation that is going to be increasingly dumped on us. This report from MIT Technology Review earlier this year said that we are actually more likely to believe disinformation created by AI rather than incorrect info cooked up by humans (though the difference was only 3%).
Elsewhere at the Egypt Media Forum Matt Capon, a video producer for ITN Productions, said that OpusClip is a game changer. This is a service that can take your long video, assess which bits are interesting and then cut it up into a series of Tiktok/Instagram videos for you. It will even assign a score to each video, betting on how likely the clip is to go viral. I tested the service on the video of our AI panel at the forum. I put the settings on auto. It said it would take 11 minutes (but eventually it was closer to 30) and came back with the following 14 clips from our talk. The first few are certainly impressive, particularly with how it has identified the key points from an hour of video and added subtitles automatically, but it struggles with trying to translate Arabic and the last couple are nonsensical. What is clear though is the role of the editor is going to change with these types of services.
Fun fact: the UAE has the world’s first Minister of State for Artificial Intelligence with Omar Sultan Al Olama and he’s had the position since 2017.
This week’s AI tool to use…
Runway are an AI company focused on art and entertainment based in New York. They are an inspiration. I finally tried out their Gen-2 Text/Image to video where you upload an image and then simply write how you would like it to be animated. From uploading a single image I simply wrote “The two ladies laugh hysterically, looking at each other slightly and then they both stick their tongues out at the camera” and this 4 second video is what it produced.
The original photo 
A still from the 4 second animation produced with AI
See the animation in action here.
The criticism online of these types of videos has been harsh (as there are faults and when animated the people look very different), but given the speed and the ease I find it astounding. Thanks to my partner Lorraine and her sister for the photo.
Coding Corner (the gradual process of a journalist learning how to code)
This isn’t exactly “coding”, but “prompting”. I built a GPT in my image. I fed it my book and a bunch of articles I’ve written. It is difficult to stop it trying to “assist” all the time. I want it to just have a normal conversation. You can chat to a GPT version of me here. You need to be signed up to ChatGPT to access this service, unfortunately. Also, as was pointed out to me, this is why there is no security on the Internet, because people wilfully hand over their identity and perform acts like this.
What AI was used in creating this newsletter?
Besides from the above image of the pyramids using ChatGPT, nothing.
In the news…
Kenya is busy trying to curb the use of AI with a “Robotics and Artificial Intelligence Society Bill”. I’m all for regulation and transparency around the material these models use, but I’m not convinced with how we are going to enforce the rules once they are in place.
When I thought of AI ten years ago I didn’t think of authors suing tech companies over copyright, but insane advancements in health which we are seeing thankfully around glaucoma detection (I’ve been having the puff-of-air-in-my-eye test at optometrists since I was a teenager). These models are trained on loads of photos of lesions and so can detect patterns that could lead to abnormalities better than humans.
With the good, also comes the bad… the US military is launching an AI drone initiative called “Replicator”. The AP says Pentagon officials are gung-ho that within the next few years the US will have fully autonomous lethal weapons. Advances in data-processing and machine-to-machine communications will relegate the military people to “supervisory roles”.
What’s new at Develop AI?
Develop AI’s ambition is to have a fully Pan-African set of software developers working on various projects across the continent. We are starting to pitch for and bring in work that is helping businesses of all sorts, not just media, with AI. If you think your business would benefit from a AI solution then get in touch.
See you next week. All the best,
Join our WhatsApp Community, listen to our AI generated podcast, visit our website, or contact us on X, Threads, LinkedIn, Instagram or TikTok.
Physically we are based in Cape Town, South Africa.
You can email me directly on paul@developai.co.za.
If you aren’t subscribed to this newsletter, click here.