We now know that deep learning techniques can generate convincing synthetic content. We’ve seen for some years how deep-fakes are able to magically switch out faces, and how OpenAI’s ‘too dangerous to release’ GPT-2 (now GPT-3) algorithm is able to generate reams of human-like text with basic inputs.
Uses though have been limited to gimmickry and the spread of misinformation. But what about more commercial uses for this technology in the creative industry? As the tech gets fine-tuned, I’m convinced that the next decade will help power a new proliferation of creativity and storytelling, giving a new voice to the creative industry (and yes, internet trolls will benefit too). Here’s a few thoughts for what could happen…
The real gamechanger will be in the convergence and simplification of the underlying technology.
When text generators, configured and monitored to produce realistic text on a subject, are merged with auto-generated video, we’ll all gain superpowers.
- We’re already seeing this start in newscasting. AI can help journalists write short factual articles, and deep learning can compile video news bulletins from text scripts; all presented by synthetic news anchors and instantly translated across all languages.
- Stepping this up to generate autonomous avatars who can read out, or even act out content written by a text generator is the next natural step — making it easier to produce (or even entirely automate) the production of video clips.
- Content creators could use these techniques to configure bots that spit out high-quality content on a subject or nuance of their choice. A prototype out from the team at AI2 is able to produce images from captions for example. Imagine being able to create your own version of the Kardashian’s. Or the Kardashian’s own 24 hour TV channel about their dogs. We already sign up to follow people in our millions, subscribing to their own versions of reality isn’t a big leap.
- What it means to interact with a character may change fundamentally. Film, TV, and gaming franchises could encode a character’s personality and behaviours into a bot, allowing people to create their own destinies and plots within a predefined universe. Think of the Marvel franchise for example. A gamer could generate their own plot in a VR game, followed by thousands on Twitch, spun off into a video series watched on YouTube. All the while following a carefully curated set of rules and behaviours defined by the copyright holder.
- The commercial opportunities in licensing and subscriptions are significant, both for the incumbent content houses like Disney or Sony, and for individual content creators alike. ‘Deepfakes’ could gain their own legitimate marketplaces, where characters, likenesses, and even behaviours are licensed out to content creators. Although expect the illicit trade to far outpace anything legal in the short term at least.
Tools that orchestrate AI-content will be big business
“With these tools, I can create an almost infinite number of parallel universes” — François Brousseau (aka EZRyderX47)
OK, admittedly fully immersive AI experiences in VR is a long way off, but the path to get there is filled with opportunity. There’s an entire industry of AI-powered tools for the content industry that will help creators generate high quality text and video content faster, and for way cheaper. From post-production touchups, translators, scene generators, idea generators, content-scrapers, background-creators — the list goes on.
Content production is still highly labour and cost intensive industry, particularly video, and in comparison to technological progress in other sectors may as well still be stuck in the stone age. The collapse in cost of software development enabled by open source and cloud computing is the backbone our contemporary tech industry is built on.
With an expected growth in digital content at a CAGR of 16.5% until 2027 and only achievable through the production of net-new tools, content production will prove to be a zoo of unicorns. Expect incumbent social networks like Youtube, TikTok, Snapchat, and Twitch to bake in these tools into their services in a new arms race to attract and retain creators.
A final word… what does all this mean for what and how we use the internet?
There can be only so many shows about Kim K’s dogs we can handle before we switch off:
“The mere idea of AI-synthesized media is already making people stop believing that real things are real” — Technology Review
The oncoming tide of AI content will change the already limited trust people have in the content being recommended in the first place. If we can no longer trust the open internet to give us genuine, factual information, then what will we trust it for? Spam filters simply won’t be able to keep up with the tide, lending further credence to the political argument that the internet must be broken up.
People will flee in increasing numbers to trusted, verified sources of information behind expensive paywalls. Great news for the whole content and technology industries, terrible news for the free and open exchange of ideas across borders and the already eroded trust we have in the information we receive.