Those insane AI videos... we need to talk
PLUS We are tackling disinformation in Africa head on...
If you didn’t see the splashy AI videos this week courtesy of OpenAI then you need to. Called Sora, the yet to be released initiative promises you will be able to create short “realistic and imaginative” videos from simple text commands. This is the new world: sending a WhatsApp to a stranger and getting a fully-formed creative project as a response.
It has been cleverly noted that the AI generated Will Smith where he eats spaghetti like a Men In Black monster made us all scoff in disgust… only a year ago. Back then in old 2023 we thought it would be YEARS before we got to this point, instead it has been months.
“A woman walking down a Tokyo street.” “A white SUV speeding down a dirt road.” “An adorable Dalmatian looks through a window.” OpenAI is promising perfect videos from these prompts… or anything else you can think of.
And, at this point, we need to absorb the fact that the outrageous amount of money and effort that is going into creating this tech means in another year we will all be able to create movies instead of just write stories… without actors or cameras or talent… but will any of us have the time to engage with what is being created? Are we headed for full-blown content burn-out as a species where only live, in-person entertainment will suffice? Or is that just secretly what I want?
For “journalism” there is clear implications for disinformation. I mean, as soon as this tool hits the market the luxury of believing anything on one of your WhatsApp groups at face value is over. Forever. And perhaps this will push people back to “trusted news sources”. For the last decade or so we have lived in a strange era where we have essentially put the digital equivalent of “a man blabbing at you in a pub” on the same level as the BBC. Perhaps we are going back to believing the authorities? And if you fancy a different news brand that’s fine. Even watching Fox News is better than believing a random Internet person’s AI generated video fantasy.
So, this could be good for established news brands. But it surely isn’t great for those beautiful, creative people who produce fantastical video content. They are going to need to be nimble with how they embrace and exploit these tools. OpenAI wants their offering to be as easy to use as Google search and it isn’t… yet. And that is the skinny space where creatives can thrive for the time being.
What I do foresee is journalists using this text-to-video tool for recreations in their documentaries. And that is worrying. Here’s why: I tried to make a static podcast cover this week with ChatGPT and though the image I created was interesting, it couldn’t redo it and make a small adjustment to the size of the picture. It needed to regenerate the whole cover again from scratch. It even suggested, almost apologetically, that I use an image editing program to get the size I wanted (I used a good friend instead). But I can see journalists creating realistic videos and settling for them to be “accurate enough”.
And maybe that is actually the world we are going to move into: a place where there is a fluidity with detail and facts that we can’t quite grasp yet. We will ultimately settle for a yet to be determined percentage of how much is “real” in our daily lives… and live in a permanent state of unresolved tension as a result.
In the news…
As a follow up to our previous letter on AI and elections, a long dead dictator was brought back to life to juice a political party. As a side note, they outlawed robocalls in the US because of what it could mean for their elections this year. And it is all pointing to my prediction that there is going to be a global trend of the wealthy portion of the population becoming completely “unplugged”. Those who can afford it will hand over their digital lives to managers and agents… like the old days.
This week’s AI tool for people to use…
I’ve been trying alternatives to ChatGPT. Anthropic's chatbot Claude 2 is free and friendly. Perplexity is still great, being essentially a chatty search engine. But you need to dip into Pi which has gone all in on being “supportive” and “smart”.
What AI was used in creating this newsletter?
None, except the lady in Tokyo I pulled from the Sora video on the OpenAI website.
Coding Corner (the gradual process of a journalist learning how to code)
If you remember last year in October I went to Namibia to coach a workshop on AI and data journalism with Peter Deselaers for DW Akademie. We are busy helping 7 teams from across Africa transform their AI and data focused concepts into working prototypes. The goals is for the teams to use these new technologies to improve their newsrooms. I was on a catch-up call with the teams yesterday and they are in fighting shape. They have learnt a great deal and I look forward to sharing their journeys individually in future newsletters.
What’s new at Develop Media?
The trailer for Challenging The Truth is OUT. The show trots across Africa and shines a light on the experts tackling the problem of information disorder plaguing the continent. We are in a different country (with a different host) each week. It is produced by Develop Audio in partnership with DW Akademie and the German Federal Ministry for Economic Cooperation and Development.
Please help us improve our Develop AI community by answering a few quick questions about what we are planning to do next. We are thinking about AI courses and new editorial products, so help us by giving some valuable input. Thank you.
See you next week. All the best,
Join our WhatsApp Community and visit our website.
Contact us on X, Threads, LinkedIn, Instagram, YouTube and TikTok.
Try our completely AI generated podcast.
Listen to Develop Audio’s podcasts and sign up for their free newsletter courses.
Physically we are based in Cape Town, South Africa.
Email me directly on paul@developai.co.za.
If you aren’t subscribed to this newsletter, click here.