Discover more from Adam’s Notes
The Era of AI-Native Products in Tech Consumption
Initial commit. Thoughts About The Upcoming 2020s and The AI-Native Products as the Next Tech Craze.
When OpenAI opened their GPT-3 API for free registration, it was the first shot fired in what would become known as the AI-Native Products trend. This is when The AI Ship left academia and landed on the shore of the Builders, Entrepreneurs, and Investors. Why is it so important? Because now it scales. Millions of Builders will think, innovate, build, try, and fail, rather than hundreds or thousands of academia personal and their inner circles. In today’s blog, I will outline my view of the upcoming trend in the tech consumption industry, called “AI-Native Products”, and where I think it is worth putting the focus on as a Builder and where it is NOT.
First, What I mean By “Tech Consumption”:
Tech Consumption is a name of a technology that has become mainstream and accessible for the average Joe (and widely used in B2C products). Great examples are:
1970s/1980s - Computers become mainstream with the Personal Computers (PC) Revolution.
1990s - The Internet adoption in the American household (and web 1.0 protocols).
2000s - When everyone opened a Facebook account.
2010s - The Mobile Era, when people got themselves Multitouch smartphones.
There are, of course, so many more examples, but you get the point.
So What Are These “AI-Native Products”?
Some of these “products” have already shifted to the mainstream, with DALL-E, which generates images based on a given prompt (e.g., the blog’s picture), or GPT-3, which generates text. These two “AI models” received extensive news coverage both on traditional and social media. However, this is not the trend I am referring to; this is what enables it. Let me explain:
Even though these models can generate value for certain people, they are not consumer products. They are too generalised and don't solve a specific problem for a customer. However, these models can be leveraged to create a product on top of them that does solve a problem. They are more like the "Infrastructure" of the true AI-Native Products era we are entering (let's call them “Infra-Models” from now.) They will play a crucial role in it - just like the innovative startups of the 2010s were leveraging the GPS, smartphones and App Stores to create unique valuable products that solved an actual problem for the consumer.
While the media is going crazy about the Infra-Models as they are, the era of the AI-Native product is forming in smaller groups on Twitter and Discord. Indie-Entrepreneurs and a small group of friends are trying to team up and build something new. I like to think about this stage as the equivalent of a bunch of guys in their garage building the first PC. It is still the trend's early stages, but this is precisely when new giants are forming.
Read more @ aigrant.org.
What it will NOT look like:
I’m already seeing three “sub-trends” in my circle on Twitter (check out who I am following on Twitter @adamcohenhillel), these trends might overlap, but their overall concept is wrong:
Copy the old: In the short term, companies will build products that imitate pre-AI companies with the AI twist. This might work, but it most likely won’t be the place of real innovation. Think of it in comparison to the Dotcom Era, when newspapers and magazines created websites that copied their offline presence without taking advantage of the new opportunities offered by the internet. In contrast, new companies that did take advantage of those opportunities thrived.
Twitter - Everyone becomes a journalist now.
Amazon - Why do we need to hold stock on the internet?
Uber - Let’s use the interconnectivity of GPS and mobile to fight the 2008 crisis.
Netflix - Started by copying the “Old” DVD model online but eventually shifted to the new era with the streaming service.
Straightforward 1x innovation: Straightforward innovation using Infra-Models is currently trendy and dominating the industry. However, this is likely to be a short-lived trend. Companies that use Infra-Models to create products such as “generate your profile picture” or “write-assistant” may have some success but are not genuinely creating significant enough new value to sustain. To be innovative, you should think about how you can leverage the already-possible capabilities of the Infra-Models to perform a new type of functionality rather than warping it in a nice-looking way.
AI-Focused Consumer Products: Companies that focus on “AI” in their value proposition should be careful, as this trend is missing the point. AI-Native enables a new type of great products; it is not the core product itself. No one who’s using Uber cares about the underlying technology; they care about what the product enables them to do.
So what will it look like?
This question can be rephrased as “What might be possible now that wasn’t possible before?” And to answer that, we need to understand what these Infra-Models are doing and what they’re capable of:
Understanding Human Language: One of the most amazing “byproduct” of generative models like GPT-3 is that they also “understand” human language. And by that, I only mean that it knows how to get text input and generate a great output based on it. This generative process can be seen as “understanding”. Now my point is the “generative” nature of the model doesn’t have to be the focus of a new product, but that “understanding” byproduct. (see list below).
Data manipulation: This as well, is a byproduct of the generative nature of the infra-models. Unuseful input A → to get back useful output B that can be used in the program. We are not taking the generated output as the goal itself but using it to manipulate data and form new structures and acting decisions.
Generating: Don’t get me wrong, the generative nature of these models is impressive, and there are so many things to build on top of that. Copilot and similar products are great, valuable and innovative. They will soon be used in similar products for other areas of human knowledge, like Law, Sales, to even gaming companies that can now generate worlds and quests at scale.
So, think about areas today that:
Have a communication issue. Today, we already see tools that “summarise” content, making it clearer for the reader to digest. However, content consumption is only one way of communication; this model can be replicated in other places.
Require Human Reasoning. Chatbots are an interesting example, as they require reasoning on the user’s prompt. But they are too simple for what I actually mean here. A better example might be Twitter, which can now leverage “reasoning” on its users’ tweets to create a different experience.
Depends on Human Automation. Copilot to everything.
Can benefit from new data at scale. Gaming companies, Netflix, and other content-consumption products can create content in a fast, personalised way.
This is very vague and only outlines a small percentage of what will actually be built soon. But again, to summarise the idea: these AI-Native capabilities are the backend of the new products, and consumers won’t care about them if it doesn’t truly enable a new capability for the mind. As a gamer, I don’t care if an AI generated this new map, but I would care about a game with an endless map. Do you get the point? Do you see the difference?
The Formation of a Bubble:
We should be cautious not to create another Bubble. This is a request for both entrepreneurs and investors. Find a real problem worth solving, don’t get hooked by the hype of the “cool” technology, and don’t get excited by huge funding rounds. It is not how much money you can accumulate from investors but how much value you can give your customers. Try to create a REAL value.
Conclusion, Beyond the Bubble:
A new era starts with the enthusiasm of single entrepreneurs who see ways to leverage new technology to create value. A bunch of companies spring up, some succeed, and some fail. New consumer standards will emerge, and eventually, we will see new tech giants alongside the old ones. These giants will be “AI-Native”, which means that in their core backend, their innovation is only possible because of the large AI models we call "Infra-Models."
Along this trend, OpenAI (as the pioneer) and other “Infrastructure” companies will rise to a giant dimension - as they are the ones who enable this era of innovation (Just like Apple enabled the Mobile age). Will these infrastructure companies to consumer products as well? I think so.
Thank you for reading. If you liked my writing, don’t hesitate to reach out. I’d love to talk with more folks and discuss everything: Tech, Philosophy, AI, Ideas, Lex Fridman, Startups, Software, Science, whatever!
Thanks for reading my Newsletter! Subscribe for free to receive new posts and support my work.