Generative AI (2/2): what will the future look like?
In my previous Article about Generative AI, I tried to set up the basics by giving a definition of this new technology trend, explaining its use cases and how the underlying algorithms where working.
Now I want to extend a little bit on which type of actors will emerge of this trends, what are the opportunities for entrepreneurs and what will be the challenges they will be facing.
Most industries could be impacted by Generative AI, but some more than others. Here I’ll discuss some of the most important ones from my point of view.
Industries that might be the most impacted
1- Marketing & copywriting
Copywriting is the most obvious and notorious usage of Generative AI. Indeed, with the growing cost of paid marketing on Google and Facebook, it has become the most powerful acquisition channel for a lot of companies out there. But it’s not as cheap as it might seem. Indeed, it takes time, a lot of time, and some good writing skills.
So, until now entrepreneurs and content creators had two choices: creating all their content by themselves or outsourcing to copywriters.
That’s where Generative AI intervenes, offering a third option: allowing anyone to use a powerful AI assistant that can help write e-mails, blog posts, tweets etc. 10x quicker than it was possible before.
Using GPT-3 and algorithms alike with a cool UI on top of it, tools like jasper.ai or copy.ai already allow thousands of persons to write blog posts, e-mails, and much more, way faster.
Because it doesn’t stop with copywriting, it impacts the overall marketing industry! These tools can help you build a slogan, a product description, and much more 🚀
We will see people using these tools more and more. And it might have a huge impact on the overall copywriting and marketing industries in the long run. I think we will see a new kind of partnership between humans and machines.
BUT …
I’m not sure a lot of companies will be really successful in this area. Or at least I think that the competition will be fierce and that companies like Jasper.ai really need to build a network effect or any kind of barrier to entry quickly if they don’t want to be overwhelmed by competitors (maybe it’s already too late for them with the arrival of ChatGPT).
Indeed, all of these companies use more or less the same algorithms (some variation of GPT-3) that they haven’t built by themselves. Ok, they might have fine-tuned it in a specific way, but anyone can access the same underlying algorithm for a very small cost and with almost no coding experience.
Knowing that Jasper already generates more than 75M$ revenues in less than 2 years, it will definitely attract a lot of competitors. And for now, it seems that the only differences between all the actors are the UI/UX and the segment of the market they target (please tell me if you think I’m wrong, I would be happy to discuss it).
Besides, these algorithms are still far from perfect on broad topics and it might take a little bit of time for a significant part of the population to use them. Regarding that, I’m very excited about GPT-4 which is currently in development and seems (from what I have heard) to be a lot more impressive than GPT-3. It might be what we need to arrive at something that can reach mass usage.
Finally, a lot of already existing tools and platforms already started to release their own assistants. So in the end, I’m not so sure we will need a single one-stop-shop for everything, but it could rather strengthen the existing actors, letting only a small portion of the market available for new startups.
2- Visual content creation
Of course, the second most obvious domain of usage for Generative AI is in the domain of art and visual content creation in general.
I have already talked about Dall.e and Stable diffusion in my last article. But what I haven’t said is that they are already used by far more than just people who want to have fun with them.
Artists use it to get inspiration and quickly test some designs and some of them even generate art 100% made by computers. We are seeing a huge growth of AI-generated art with a new kind of artist having computer science skills emerging.
But these tools are far from perfect, their results are most of the time not what we really want and they can be quite frustrating to use as there is no native way to edit the designs generated yet.
For entrepreneurs, I can see a lot of opportunities to build tools that provide artists and designers with a way to integrate Generative AI in a better way into their workflow.
I believe that we will see some tools targeting very specific areas of the design industry like “Logo generation” leveraging this technology together with tools allowing us to easily adapt and edit AI results and empowering anyone to become a better and more efficient designer.
We might also see new kinds of design agencies leveraging these tools emerge.
However, I think that there are a lot of challenges out there. Especially regarding copyrights.
Besides, as these algorithms are trained with internet data, at some point, we will need to create more images in order not to limit their creativity. These reasons might push Open AI and Stability AI to hire a lot of designers that will create content just for the sole purpose of training their algorithms.
3- Gaming and Metaverse
One of the most important challenges in gaming, VR, and “metaverse creation” is the content creation part.
Some of the most successful games of the last decade like Red Dead Redemption 2 or Assassin’s Creed Valhalla partly derive their success from the incredibly vast and rich virtual worlds they have created. It’s truly astonishing!
But they are soo expensive to create! Red Dead Redemption 2 has cost more than 500M$ for 8 years of development and Assassin’s Creed Valhalla came from a massive collaborative effort from thousands of people across 15+ studios.
We have reached a limit of what can be built by humans while still being economically viable. And that’s one of the biggest challenges that are facing VR and Metaverse companies right now! They just don’t have enough content to attract a lot of users, so it means fewer revenues which translates into even fewer content, and on and on …
Some algorithmic approaches close to Generative AI have been used for quite a long time by gaming companies in order to build infinite worlds. One of the most popular of these approaches is Procedural Generation. That’s what Minecraft and games like No Man’s Sky use to automatically generate unique worlds and maps. The approach is quite simple, game designers provide an algorithm with a set of objects, landscape elements, and rules. And the algorithm randomly generates new worlds by following these rules. But this approach can be quite limited and are hard to apply to any kind of universe.
And that’s why I’m quite sure that thanks to Generative AI, everything will change. Imagine being able to ask a computer not only to generate a world by putting together multiple pieces of content designed by you but using these pieces of content to generate entirely new ones and build 100x richer worlds!
By fine-tuning 3D Generative algorithms with designs built by talented game designers as well as virtual replicates of real-world objects (made thanks to technologies like Instant NeRF from Nvidia) companies will be able to generate incredibly rich virtual worlds and games maybe at 1/100 of the cost and in 100x less time that they could before.
It will open a new era for gaming but also for Metaverses that could become a “less far-fetched dream”.
But it’s not all, thanks to Generative Text and Audio technology, we might be able to create really smart PNJs and game avatars like never before.
If Dall.e and Stable Diffusion can create compelling images and avatars in seconds and GPT-3 is smart enough to sustain a great discussion, how far are we from this?
I think we might see a lot of great startups emerge with tools helping video game and metaverse designers build much better virtual words in a much more efficient way. There might be a huge opportunity here for a new kind of platform that could compete with Unity or Unreal Engine!
We might also see a lot of new gaming studios emerge with deep expertise on how to use these tools to build the games of tomorrow!
Of course, it will not just impact video games but 3D in general: movies, simulation, etc. and we might see incredible companies building themselves targeting a specific subset of the 3D market.
The next step of this evolution (that might take some time to become widely used because of technical challenges), could in my opinion be something I call “Adaptative virtual world and stories creation”.
Aidungeon is one of the first examples of that. It’s a tool that automatically generates adventures and adapts them depending on how you interact with it.
In the future, imagine being able to type “I want to play an adventure game where I travel across the ancient Egypt of Ramsès II” in a prompt and a virtual world of your choice would be automatically generated!
It might change the entertainment industry forever, but not only: I believe it could also literally metamorphose the education industry, have an impact on the medical industry, and much more.
4- Coding & Engineering
In my last article, I already talked about how Generative AI allows generating code by writing what you want to build with simple words
GitHub Copilot, a tool made by GitHub for this usage, has been used by more than a million developers helping them to be more efficient!
It’s quite impressive when you see it in action, and as a part-time developer, I thought at first that it could change everything in the way people code.
But I might have been a little bit biased. Indeed everything is in the “part-time” part here. As I don’t code every day, I don’t always recall the exact formulation of basic commands and when I go from one language to another (like from Python to Javascript then to Solidity), I always struggle to get back to the perfect syntax. So for people like me, it can be a very well welcome help.
For non-developers, even if GitHub Copilot won’t be useful as it is, I believe that we will also find a lot of great use cases. A good example of that could be to use these kinds of algorithms for database querying. I’m very interested to see the development of tools like ai2sql that allows anyone to find what they need in a database from a simple prompt. I can see so many useful usages of this tech across various departments of companies: sales reps, finance guys, data analysts, and more!
But as I interviewed a dozen developers using GitHub Copilot in order to get a sense of the magnitude of the changes it could imply, it seems that they were way more dubitative.
According to them, the code structure generated can be quite bad, writing a good prompt to generate your code can be very hard, and you always have to check everything behind the algorithm.
So in the end, according to studies, they might be 1.5x more productive by using Github Copilot. Yes, it seems huge like that, but compared to the gain of 5 to 10x in productivity when you use Python compared to C, it’s not soooo impressive.
So I don’t think it will change the way people code as it is but I think it will have 4 main impacts:
- We will see beginner & part-time programmers be much more effective in producing medium-quality code
- It might be a lot easier to convert some legacy code from one language to the other
- We might continue to see other extensions & IDEs leveraging this tech better and better to allow developers to be more efficient. More specifically, some tools might let companies train these algorithms with their code and good practices to create smart assistants capable of harmonizing the code across all repositories as well as accompanying new employees.
- We will see platforms leveraging these types of algorithms together with the principles of no-code platforms like Bubble, allowing us to build even more easily very powerful web applications. Tools like debuild (which excites me a lot) seem to be on this path.
I don’t think that Generative AI on its own will change how we develop web applications but put together with the progress we have seen in no-code and tools like “Figma to React” (allowing us to convert automatically Figma Design into React components) I’m pretty sure it will!
5- And much more…
These use cases might only be the tip of the iceberg and I see many other transformative use cases already tested or in coming.
- Using Generative AI to create new molecules and medicines: Thanks to their underlying computational power, Generative AI models can be used to generate new molecules and medicine that humans couldn’t think about or wouldn’t take the time to test. Microsoft developed a model that seems very promising in this area. It is one of the most promising domains for Generative AI, so I could have created an entire part dedicated to it, but since at Red River West we don’t invest in Biotech and I’m far from being an expert on the subject, I won’t risk it :)
- Cybersecurity: Black hat hackers already use Generative AI to generate new ways of attacking and corrupting systems. On the side of the white hat, some companies are leveraging it to prevent attacks and defend corporates. It might become a huge market as cybersecurity is one of the most important matters of the XXI century. The threats it will create are so huge that the opportunities for “white hats” should follow.
- Pictures and video edition: We have already seen some great actors emerge on this side leveraging AI for specific pictures & video edition use cases. The French PhotoRoom is a very good example of that, offering a Photo Studio at your fingertips.
- Well… I could continue on and on, but I really don’t want this article to last an hour, so I’ll wrap up here ;)
Who will be the actors of Generative AI?
I think we will see 3 different types of actors emerging:
1- Global actors focused on huge markets & tackling multiple issues in one product:
In the next 10 years, I believe that we will see a few Decacorns (companies valued at more than 10B$) with Generative AI at their heart emerge.
But I think there are not so many markets in which Generative AI can produce 10B$ + exits and I don’t believe there will be a lot of companies in this bracket. Gaming might be the sector with the most potential for a 100B$ + company to emerge but only if existing actors don’t monopolize the market before.
As I already explained, knowing that for now, most companies using generative AI use more or less the same algorithms. Entrepreneurs will need to build a huge barrier to entry if they want to have a chance to create this type of business. It could be creating a huge network effect, collecting proprietary data that would make their tools 10x better than the others, combining Generative AI with other proprietary technology, etc.
2- Companies targeting very specific niches or use cases using Generative AI:
For me, that’s where generative AI might create the most value. Generative AI will have a strong impact on most of the industries we know about (Real Estate, Investment, Media, Video Games, cybersecurity, and much more) and tools focused on professionals of these industries with Generative AI powered features might be the real game changer.
If you want to create a company, I strongly advise you to look at industries you like and reflect on the main issues they have that could be solved by AI.
Some incredible startups will emerge in this area tackling a lot of issues that we couldn’t before. A company I love in this bracket is runwayml which helps anyone create and modify videos like never before. Some of them will reach “Unicorn” status and others will be bought by corporates trying to add new features to their products.
I also think that in some cases, the “niches” on which these companies will apply generative AI will be too small to make them VC fundable (or at least to become unicorns one day), but will rather create some very profitable small businesses with high EBITDA margins.
A very good example of that could be tweethunter, a company that seems to employ less than 4 people part-time while generating more than 1.2M$ ARR! Applying Generative AI to a very specific niche might be one of the best opportunities right now to create profitable side businesses that don’t require a lot of technics and time to operate.
3- Incumbents using generative AI to improve their software:
I think that in the end, even if most innovation in the Generative AI space might come from startups, most of its everyday usage will come from already established tech companies. Big techs and scale-ups already started entering the race and a lot of them are working on integrating GPT-3 or Stability powered features into their existing products.
Adobe has been developing Generative AI features for quite some time now and I believe that the likes of Adobe and Canva might be some of the biggest actors leveraging this technology.
Notion is on the verge of releasing its Generative AI assistant, and it has been years since Google complete your sentences in Gmail (it will become better and better). Ubisoft has already used Generative AI in their games and most established Gaming companies will follow.
These innovations can happen from inside, with companies building their own tools, but also from outside by acquisition. Thus I believe that we will see a lot of acquisition of AI companies from established actors in the next 5 to 10 years, giving great exit opportunities for entrepreneurs and VCs.
Conclusion
As you saw, I’m quite excited about generative AI and the possibilities it will bring. I really believe it’s one of the biggest revolutions that we have seen in the last years and the concretization of a lot of hope we had for AI. It will create incredible possibilities for founders but also a lot of challenges! But it also comes with a lot of challenges and I hope it won’t create a tech bubble in the next years.
As a data scientist by training and a VC at Red River West, I will follow the development of European startups leveraging this type of tech with a lot of attention.
Please contact me at abel@redriverwest.com if you want to discuss it or if you are building in this area!
