Written by Fola Yahaya
Sending out this newsletter each week is always a bittersweet moment. On the one hand, there’s an immense sense of relief from hitting my deadline. On the other, there’s the looming question of, “What on earth will I write about next week?” Yet, without fail, a few hours into my day job, ideas for the next edition start flooding in. This is because the pace of AI development is consistently dizzying.
As a result of the introduction of AI models that spend more time ‘thinking’ than just eagerly generating, we have moved from the ‘AI as a Toy’ to the ‘AI as a Tool’ phase. This week alone, thanks this mini paradigm shift, ChatGPT and Claude have enabled me to ‘build’:
All these ‘products’ work, more or less worked straight away, and were built by simply asking an AI chatbot a series of questions – starting with ChatGPT, and now using my new favourite coding sidekick, Claude – and iteratively refining the output. The speed with which AI has suddenly gifted me a new superpower is pretty mind-blowing. Excitingly (but sadly for coders), it lends weight to the argument that we’re approaching the end of coding as we know it. Why build separate apps when you can just ask an AI to piece them together on the fly?
These mini paradigm shifts in how we create and even think just keep coming, so you should ignore any ‘research’ that claims the broad spectrum of AI development is an expensive and short-lived fad. Yes, most people are wasting a lot of time still playing with AI tools, but I guarantee that smart companies are already successfully using far more powerful versions of AI models.
Take the recent comments from Jensen Huang, CEO of the trillion-dollar AI chip giant Nvidia, and C. C. Wei, CEO of TSMC (the company that manufactures Nvidia’s chips). Both these companies are the shovel sellers in the AI gold rush, but critically both have eaten their own dog food and shown how embedding AI into an organisation’s DNA can be spectacularly successful.
AI is quietly embedding itself into industries and workflows. It’s adding tangible value at a pace that’s accelerating faster than most people realise, and this is what keeps me awake at night. Much like my mixed feelings when I hit “send” on this newsletter, the excitement of every new AI superpower is always tempered by the reality of the inevitable job losses it brings.
Normally I like to finish up this newsletter with some light-hearted fare of AI Gordon Ramsay doing something inappropriate. However, I was blown away by yet another AI video feature that has just changed how every single animation will be made. The AI video generation company Runway has just teased its upcoming release. Remember all those funky motion capture dots that the English actor Andy Serkis had to use to capture Gollum in Lord of the Rings? Well, that’ll soon be unnecessary thanks to Runway’s Act-One feature.
As Runway say themselves:
“Traditional pipelines for facial animation often involve complex, multi-step workflows [i.e. lots of expensive humans]. These can include motion capture equipment, multiple footage references, manual face rigging, among other techniques. The goal is to transpose an actor’s performance into a 3D model suitable for an animation pipeline … Our approach uses a completely different pipeline, driven directly and only by a performance of an actor and requiring no extra equipment.“
Even freakier, they are planning to release a tool that can allow a single actor to be rendered as two separate actors in the same scene
Worryingly, this feature release is called Act-One, which begs the question: What’s coming in Act-Two?
One of the most frustrating things about AI is its ability to do the things we don’t need it to do, like writing love letters and creating photorealistic images, but its abject failure at doing the drudge tasks, like filling in forms, that we all hate. Despite the silicon snake oil sold by both Google and Microsoft, their extension of AI co-pilots to our computers continues to be overhyped and underwhelming. However, what a difference a few weeks make.
Anthropic’s announcement on Tuesday that its AI tool, Claude, can now now look at your screen, open your folders, click buttons and type text is a first by a major frontier AI model. This makes me think that we’re on the cusp of entering level 2.5 on my scale of AI development, the ‘Tool/Taskmaster’ phase. This is the phase where AI is not just a tool in your toolbox, but can also extend out of the confines of the chat interface to start independently making intelligent decisions about how to do things. For example, as a non-coder, you could now use Claude to create a basic but totally decent productivity app or Duolingo clone. This alone, given that it only takes an hour of focused time, is crazy.
However, you still have to ask it lots of questions, cut and paste stuff into a code editor, open a browser and then keep iterating. Level 2.5 would still require a degree of iterative prompting but would just build the app, open the code editor, run the code, test it and make changes without you having to lift a finger. So finally, we’ll have an AI powerful enough to… fill out forms ;-).
But seriously, features like this mean we are now entering the awkward-sounding “Agentic era” of AI. The first step towards this was OpenAI’s new Strawberry line of AI models, which are different because they spend a bit of time thinking (known as inference) before giving you a response. The next step is letting AI agents loose on our systems. Bill Gates, who has been banging the agent drum for the past two years, interestingly predicted that by 2028:
“You won’t have to use different apps for different tasks. You’ll simply tell your device, in everyday language, what you want to do. And depending on how much information you choose to share with it, the software will be able to respond personally because it will have a rich understanding of your life.“
This looks increasingly conservative. For example, Marc Benioff, CEO of Salesforce (the customer relationship system used by most large corporations), envisions a world where there are one billion agents (AIs not people) operational within the next year. These agents are AI-driven microservices – modular pieces of code with brains, performing tasks, making decisions and delivering outcomes. So rather than paying for Adobe Creative Suite or HubSpot user licences, customers will commission work from a network of AI agents. These will be smart, autonomous services that can operate independently or as part of a larger network. Critically, Benioff argues that companies won’t be paying for the capability to run the software, but rather for the value the agents create. This is the end of traditional software licensing as we know it – and it could change the entire structure of the industry.
Having been empowered by AI to code usable apps, this upending of the software development status quo makes total sense, and will come sooner rather than later. Within a year, AI agents will be able to crawl the web, cobble together bits of code and create our own personal apps without a user licence in site. Tricky times ahead for another major industry and those it employs.
HeyGen creates video avatars. It can essentially clone you and your voice, and allow your digital twin to appear in a any video and lip-sync practically any language better than Britney Spears at a concert. In short, it is the acceptable face of deepfakes. It’s now gone one step further by creating the first Interactive Avatar, which lets digital you step into Zoom meetings at any time, across multiple meetings simultaneously. Now here’s where it gets all Black Mirror. According HeyGen’s marketing blurb:
“Your avatar won’t just look and sound like you, it’ll think, talk, and make decisions [based on the knowledge and persona you give it], just like you.“
Woah. So you can now send a deepfake of yourself to a Zoom meeting and trust that its ChatGPT brain will not mess up or hallucinate or, worse still, fail to dodge the bullet of the annoying task no one else wanted to do. I think not. Yet again, this uncanny valley stuff is a problem looking for a solution. There’s a reason why people have meetings. To C.O.N.N.E.C.T. with other humans. Sending your digital twin just shows how little you care about those who attend or, more importantly, that that meeting shouldn’t be happening in the first place.
As an experiment, I created a digital twin here. (Disclaimer: I don’t have issues with my jawline, and I don’t sound like ‘Ryan’ – I was just too cheap to pay for full voice cloning and an advanced avatar).
In a week that saw over 10,500 creatives, including ABBA and Thom Yorke (Radiohead) sign a statement demanding AI platforms like Suno stop ripping off their music, multi-Grammy Award winning artist and producer Timbaland has gone all contrarian. In fact, you won’t get a more enthusiastic promoter of AI, as he says:
“… [Suno] is the new everything. It’s the way you’re going to create music, it’s like: ‘I’ve got this idea, but I don’t have to run back to the studio – I just run to Suno.’“ (Music Radar)
He goes on to say:
“It’s gonna be about which prompt you’re using. It’s not gonna be about ‘Give me that sample,’ it’s gonna be, ‘Yo, give me that prompt.’ So you selling prompts now.“
Whilst American AI developers might be dissuaded from rampant creative theft, there are thousands of models out in the wild that have already done just that and will soon only need synthetic data to create output. Sadly, for creatives, the horse has already bolted and it’s time to lean in.
P.S. Thanks for helping me reach 2k subscribers!
It’s been about a year since I launched this newsletter – a project that started as a simple guide for friends and family, trying to break down both the promise and the pitfalls of AI.
So, a huge thank you to everyone who thought it was worth passing along.
Please do share!!!! Because sharing is caring 😉
That’s all for this week. Subscribe for the latest innovations and developments with AI.
So you don’t miss a thing, follow our Instagram and X pages for more creative content and insights into our work and what we do.
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.