Founder & CEO at bilateralstimulation.io.
Many experts now agree: AI marks the next platform shift in the world of technology.
Recent advancements in AI research, starting with the foundational “Transformer” research paper in 2017 and exploding into the public consciousness with the release of ChatGPT in November 2022, have increased AI capabilities to the point where even your mom finds use in it.
With the arrival of multimodal AI, neural nets can now read, write, see, hear, speak and maybe soon also smell.
This dramatic increase in AI capabilities has expanded the frontier of what is possible to do with computers.
What might have sounded like sci-fi only a couple of years ago can now be built. Expect to see real-life versions of Iron Man’s J.A.R.V.I.S., Her’s Samantha and even Star Trek’s Holodeck in the years to come.
The frontier of possibility has expanded—but to turn this potential into real-world impact, we need to actually build things. Entrepreneurs (and intrapreneurs) have to pick up the baton from researchers to build applications that deliver on the promise of this new technology.
The newly opened space of entrepreneurial opportunity is vast. A true platform shift means the impossible is now possible and the unprofitable is now lucrative, across many applications and use cases.
So how does the entrepreneurial mind orient oneself in this new world? How do you position yourself in this (justified) gold rush, be it as an ambitious startup or as a changemaker within an established business?
As we get closer to the one-year anniversary of ChatGPT, some patterns have started to emerge.
Let’s take a tour through some of the most exciting themes of entrepreneurial AI opportunities.
Specialized Foundational Models
At the epicenter of this platform shift are the foundational models. The current live players here are OpenAI and Microsoft, Anthropic, Meta and Google DeepMind, as well as new entrants such as Mistral AI, Inflection AI and Imbue. These are the alchemists of the gold rush.
Training a state-of-the-art foundational model is a highly technical, dynamic and capital-intensive endeavor that is tough for any mere mortal to compete in, no matter if you’re an ambitious startup or a well-endowed incumbent. Most of the time, you will be better off just building on the shoulders of the emerging giants.
But look beyond the hyper-competitive space of generalist models and you will find pockets of specialized opportunity: Companies building smaller, domain-specific models that have the potential to outperform the generalist models within their problem space, at lower cost and with better performance.
Examples of this are Replit, which is building its own small model focused on code generation, a large Indian IT company building Hindi-focused models and various enterprises building models on their proprietary data.
So far, we have all relied on ultra-generalist foundational models trained on a subset of the entire internet. What can we achieve with smaller, more-specialized models, trained on curated data, optimized for your specific problem space? What data do you have access to that allows you to build a superior reasoning engine for your domain?
Horizontal Plays
One layer below the foundational models is the emerging world of AI engineering. Somebody has to build the shovels, picks and buckets we all use to mine the artificial gold.
Similar to previous platform shifts like personal computing, the internet and mobile, this layer is currently undergoing a “Cambrian explosion” phase where we’re throwing a thousand blooming flowers at the wall. It remains to be seen what sticks.
This is an exciting layer where the state-of-the-art is advancing every month and many fortunes will be built. But if you want to compete here, you have to be a quick study, build fast and be willing to change course multiple times until you arrive at a profitable and durable equilibrium.
Vertical Plays
In the technology industry, we sometimes forget that the world is not all picks and shovels—somebody has to actually refine and put to use all the gold that we mine.
This is where vertical, industry-specific ventures come in. These ventures tend to get less public attention because they are less shiny; the problems they solve often require some domain-specific knowledge to grok and their solutions are often non-obvious to the layman.
This is a huge opportunity for you as a builder. Pretty much every industry is now ripe for having its workflows automated, its quality improved, its creativity amplified and its level of abstraction raised with the advent of the reasoning-as-a-service engines that the AI revolution has blessed us with.
Personally, I believe that this is the layer where the most real-world impact will be created for the most people.
Which industry are you in right now? What unique problem spaces have you been exposed to? How can you use these new AI capabilities to solve real problems within your vertical domain?
AI-Native Plays
The last and hardest-to-predict category is AI-native applications. What kinds of products, companies and experiences are now possible that simply weren’t before?
The advent of the internet brought us websites like Google, Wikipedia and Facebook. The advent of mobile brought us mobile apps like Uber and Google Maps. What will the AI equivalents be?
It’s too early to tell for sure, but we are starting to see some early promise in the areas of:
• Mass-personalized content (think education, entertainment, news)
• AI-powered personal assistants, tutors, coaches and copilots
• Goal-oriented AI agents that solve real problems for people
Outlook
All of the above can be built today, without any further advances in AI research.
But the research hasn’t stopped. If anything, it has accelerated as we start to grasp what might be possible with this new category of technology.
Expect more capabilities to emerge as models are scaled further, new modalities come online and new architectures are discovered.
We are in for an interesting ride.
Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?
Read the full article here