Right as the new year begins, I’ll share with you a special newsletter issue about AI, which will be a primer on how the ecosystem is evolving and what might happen next.
Let me take a detour, telling you about my experience in the last six years
2022 has been an astounding year, both for me professionally and for the AI industry as a whole.
And I can tell you from commercial experience in the field that back then AI was mainly a combination of natural language processing and some intelligent features with it.
Back then, this was cool stuff.
However, nothing compared to what happened in 2019, when GPT-2 was released.
Why do I focus on GPT-2 and not the release of the first GPT model?
Well, I remember checking out the first GPT in 2018, together with the product team at the startup I worked for at the time, and from a commercial standpoint (the things that you could do on top of it as a user) I was not impressed at all.
Indeed, I almost shrugged as to completely dismiss what I saw.
Back then, I could not tell you at all I was excited about what AI could do.
I remember describing to clients what they could do one day once AI and their own data would come together to create incredible and dynamic conversational interfaces.
Yet, when you went back to reality, everything was very sluggish!
In 2019, the pace of AI started to change.
The pivotal moment was GPT-2.
That release made me think, “ah, there is something here!”
From that standpoint, I could finally see something that made sense.
Content generation was definitely the primary use case for AI.
Yet, for how good GPT-2 had become, it was still not comparable (I thought) to humans.
Things changed when in 2020, OpenAI released GPT-3.
There, I saw clearly that content generation was the perfect use case for AI.
That, indeed, led to the explosion of a plethora of AI-generation tools.
I still found them valuable for specific use cases (outline generation, content ideas, title generation, research, and so forth).
Indeed, as a writer, I didn’t find yet, the long-form content generated by the AI good enough.
Meaning, that was great to get started, but if you do write quality content, which is well-researched, factual, and grounded on experience, that was not it!
However, many have been focusing on the content side.
My argument is that this time, ChatGPT going forward, the real use case for the AI, is not content generation; it’s code generation!
Where for the previous two years, the perfect use case had become content, from 2022, going forward the perfect use case is software development.
Thus, on the one hand, competition might be much more fierce. On the other hand, the rise of AI-based tools might enable the rise of even leaner startups.
Tiny teams now have incredible leverage, as with AI-based tools, they can enhance themselves. And those tiny teams can maybe become multi-billion dollars players?
And are we looking at the rise of a new form of a lean startup?
So now that I properly set the ground, let’s get to it, but before a quick recap.
Let me break down the main stages of the commercial development of AI.
2016-2018 – The NLP Era
I’d call this period “the brute force of coding!”
2018-2020 NLG becomes viable
After years of fantasizing about NLG, with the release of a large language model like GPT-2, Natural Language Generation started to really become viable.
At this stage, AI started to be effective at content generation.
2020-2022 NLG for content generation
Here, with the release of GPT-3, it becomes clear that AI can tackle the content
2022-forward NLG for software development!
One thing is for sure, 2023 might become the pivotal year for the AI industry to consolidate around a few key foundational pillars (from both a technological and commercial standpoint), which might shape the industry in the next five-ten years.
A few key things, I believe.
Strengthening the foundational layer and multimodality
In 2023, things might get even more interesting as new updates will be released by the foundational layer’s players.
The most awaited one is GPT-4, which according to many, might actually be multimodal.
Meaning it might finally handle many prompts and interactions, be it text, image, video, and more.
This is critical.
Because if you’ve been playing with the various generative models, you know that while those a general-purpose, they are still highly verticalized.
For instance, if you take GPT-3 it’s incredible at performing a bunch of tasks, but it sucks with images.
If you take DALL-E, it can do wonders when it comes to image generation, but it doesn’t handle text.
Multimodality will be a critical component of an AI industry that can really take a further leap forward.
Whereas today, for instance, you can obviate that by building specific AI engines that combine GPT-3 and DALL-E, what if this would be done by default?
2023 might really be the year where the middle layer takes over. Here, as I explained in previous newsletter issues, we might see the rise of AI players that can automate a whole set of corporate functions.
For now, AI companies like DoNotPay are finally becoming viable.
Yet, imagine if you’ll have your own AI Lawyer, Accountant, HR, Salesperson, and Marketer.
This will also turn into a challenge for legacy players, that must need to integrate AI into their products to be able to compete.
For instance, if I were, Salesforce, I would be pretty worried about this development, as 2023 might be the year when an AI-based Salesforce comes to market and gains traction quickly.
Every software company will become an AI-based company
Every software company will become an AI company in two ways.
First, in generating code, developers will leverage AI assistants to be faster and more effective.
On the other hand, those tools will have built-in features to tackle a good chunk of the user’s journey.
From lean to leaner
I also hope that we’ll see the transition from lean organizations to leaner ones.
Tiny companies, made of a few people, can also remotely build valuable stuff at scale.
Moats are built by mixing data, algorithms, and fast iterative loops
Here, those players are able to use AI-based tools to enhance their workflows.
While also, integrating AI-based features within their software will have an incredible competitive advantage.
The Cambrian explosion of AI-based Apps!
In this context, 2023 might be the year where we see thousands of business applications built on top of foundational layers for both B2B then consumers.
The interesting part?
Among these players, the few ones that might come up as the most relevant might also be the ones who reach a billion users…
I also recorded my thoughts below!