Written by Fola Yahaya
I have a confession to make. I write my content from scratch. I know, I know, in the ChatGPT era it’s crazy to spend precious time thinking, coming up with ideas and writing and then re-writing endless drafts. However, as a writer, I find that the seductively fluent ChatGPT first draft just gets in the way of good writing.
Yet, I appreciate that most people don’t feel the same way. We can all remember the tedium of being forced to learn grammatical rules and spelling dogma so that we could write essays and school reports. Many of us will have also been frustrated by our laughable attempts at making an apple look like an apple in art class.
AI is being driven by people who ultimately don’t enjoy the birth pains of creativity and seek to shortcut it. But while it ‘democratises’ creativity, we need to be mindful of what we lose by avoiding the ‘pain’ of the creative process.
Below I discuss the implications of Sora, the latest and greatest in a long line of ‘instant creativity’ tools. I also discuss the implications of the potentially ground-breaking new chip architecture that makes creating any content – from art to real-time conversations – truly instant. My feeling is that tools such as Sora herald a new era in which all content starts with an AI draft and that the creativity lies in tweaking it.
Just as I was putting the finishing touches to last week’s newsletter, OpenAI dropped Sora, a text-to-video product that blew everyone away with its jaw-dropping brilliance.
The video below of the Tom Hardy lookalike astronaut was generated by inputting a simple prompt: A movie trailer featuring the adventures of the 30 year old space man wearing a red wool knitted motorcycle helmet, blue sky, salt desert, cinematic style, shot on 35mm film, vivid colors.
Released for testing to a handful of key developers and artists, Sora is to text-to-video what ChatGPT was to text-to-text. What’s freaking people out is not just the ease with which these up to 60-second videos can be created, but their incredible realism. And it’s another example of a hugely funded company creating a solution that could put whole professions out of business. I spoke last week about the death of stock image libraries and the army of photographers that eke out a meagre living from their royalties. We can add to that videographers who survive on advertising dollars. Sora ushers in a new era of instant (eventually personalised) and dirt-cheap video for the masses. More examples here.
The exponential rate of progress in AI video means that potentially in a few years, a streaming giant like Netflix which sits on a universe of data will be able to produce fully-fledged films tailored to our tastes or fixing terrible casting mistakes e.g. replacing Jesse Eisenberg as Lex Luthor in Batman v Superman: Dawn of Justice.
Also I would hate to be one of the investors who’ve poured over half a billion dollars into GenAI start-ups such as Runway that are about to be blown out of the water by OpenAI.
Groq (nothing to do with Elon Musk’s ChatGPT competitor Grok) is an AI chip company that is making waves in the tech world due to its Language Processing Units (LPUs). These LPUs enable 10 to 100 times faster processing of large language models, far exceeding the current gold-standard Graphics Processing Units (GPUs) from NVIDIA.
In plain English, Groq’s chips allow AI bots to respond in real time to questions or in conversation.
I’m an avid language learner who is frequently frustrated by the tools on the market. Memrise, Duolingo, you name it and I’ve signed up, used and soon discarded it. Despite their efforts to shame me into daily practice, fundamentally I learnt very little and came away poorer and linguistically none the wiser.
So for me, large language models (LLMs) represent a huge leap in personalised, mobile and cheap language learning. I spoke about this in a previous newsletter but to recap, to get this working you just have to:
Et voilà, you’ve now created a language teacher that’s available 24/7 to encourage you on your learning journey. So far so awesome.
However, what I really wanted was to learn Yoruba, my native tongue that embarrassingly I don’t speak. I’ve tried courses, YouTube videos and books but despite the fact that I already speak three other languages, Yoruba doesn’t seem to stick. So creating a Yoruba tutor seemed like the perfect solution to allow me to learn how to have a basic conversation with my parents in their own language before it’s too late.
However, ChatGPT and the majority of GenAI bots offer a limited number of mainly European and Asian languages. Apart from Swahili, there is limited voice data available for the 1,500 or so African languages. This means we have AI-colonialism: we’re all drinking from the same fountain that has been flavoured with largely US data.
As Jensen Huang, NVIDIA’s CEO, recently said at the World Governments Summit, every country needs its own AI and voice data sets. If countries want to preserve their digital tongues or train their populations in using and leveraging the benefit of AI, they will need to focus on preparing language packs.
It codifies your culture, your society’s intelligence, your common sense, your history – you own your own data. Jensen Huang
A great article in the FT last week discussed the profound impact of AI on the advertising industry. It begins with how Publicis, one of the world’s biggest advertising agencies, used AI to send a deepfake video message of its CEO engaging in the hobbies of each of the company’s 100,000 employees and in their native languages. The stunt was designed to illustrate to staff how technology allows ads to be created quickly and cost-effectively, at a scale and with a level of personalisation previously unattainable.
However, we’re back in double-edged sword territory with this one. If anyone can use AI to create reasonable campaigns for peanuts, what’s the point of advertising agencies?
The article references a UK marketing chief – who (of course) declined to be named – who launched a new vegan brand for a client last year. Rather than paying tens of thousands of pounds for a team of people to design a new name and logo, they simply asked an AI chatbot for six ideas and selected the best.
This disintermediation, or removal of the middle man, represents the biggest upheaval of the traditional advertising agency since the arrival of the internet. If brands can turn directly to AI for inspiration, what use are the advertising executives who ply their trade by selling creative concepts? Financial Times
With Sora, it’s clear that OpenAI are capable of doing amazing things quickly, so we have to take it seriously when Sam Altman likens the current ChatGPT model to a “barely useful cell phone”.
He anticipates a very near future where AI becomes a personal tutor, provides personalised medical advice and helps solve global challenges.
Thanks for reading this week’s newsletter. Strategic Agenda helps clients create content to get the right people to do the right thing. This newsletter is our take on the most important aspects of AI and how it impacts the content and communications landscape.
Network Hub, 300 Kensal Road, London, W10 5BE, UK
We’ll send you quarterly updates so you can see what we’re working on, what the future holds and how we’re shaping it.