How To Thrive in the AI Era, Pt 2
Who wants to Be An AI Millionaire? Builders, monetizers and investors
Part Two of “How To Thrive in the AI Era”.
In part one, I presented the trifecta: Learn AI. Use AI. Leverage AI. AI is a powerful productivity-boosting technology for creative activities and knowledge work. The greatest value creation from AI will accrue to those who utilize AI to boost their own productivity and creative output.
The real winners in the AI era will go beyond using AI to be more productive to becoming providers of AI value to others. That’s why there’s an AI Gold Rush going on, many see a multi-trillion market opportunity in AI.
Just as the internet spawned new technology stacks and business opportunities, the AI technology wave presents tremendous opportunities in new applications, business models, and more. Those opportunities can be broadly put into three buckets - building AI to help others; scaling use of AI to create value and monetize it; and investing in others who win with AI:
Building AI: Building the AI models, tools, agents and applications that help support AI applications and AI use cases.
Monetizing AI: Scaling use of AI and monetizing that value, either by selling the valuable outputs of AI or by gaining an AI-based competitive advantage.
Investing in AI: Investing in AI, robotics and other companies that win with AI.
Building AI - The AI Stack
AI’s rapid advance presents numerous opportunities to create valuable new change various layers of the AI value stack. AI changes everything, and it starts with shifting entire computing technology stacks to be AI-centric. Everything from hardware, devices, infrastructure, models, application frameworks, interfaces, and AI end-user applications are being re-thought and re-built for AI.
It’s helpful to break out the types of AI building that is going on. Chip Huyen analyzed thousands of open source AI-related GitHub repos to understand the types of AI tools and applications that are being developed and gaining traction. He breaks the AI stack into 4 layers: Infrastructure, model development, application development, and applications.
There is a flurry of activity in building all four categories, in both the open source Github projects that Chip covers, as well as various efforts in many startups and established tech companies.
To grasp the opportunities for building AI, we can break out into these four and add a fifth: AI hardware. AI hardware is specialized silicon, systems and devices for AI workloads, and it includes AI-optimized processors, GPUs, as well as devices (like AI glasses) tailored for AI tasks and applications.
AI Infrastructure is the software that supports training and running AI models, such as Nvidia’s Triton, as well as cloud-based AI services, and AI model optimization and acceleration tools.
AI models and model development: Developing and improving the deep-learning architectures, training algorithms, and inference techniques for AI models is the most critical part of the AI stack. Frontier AI model training is restricted leading AI model development groups with GPU resources, but others can focus on fine-tuning AI models for niche and local use, and researching alternative architectures.
AI application development: The comprises the software platforms, frameworks, and tools for building on top of AI models to efficiently develop and deploy AI applications. This includes AI development platforms such as LangChain, llama-index, DSPy, vector databases (Pinecone, Weaviate), as well as AI agent frameworks like Autogen and CrewAI. This arena is fast evolving, and there is opportunity here to make a mark still.
Development in this category is the work of the “AI Engineer” which is defined as:
AI Engineer is a software engineer who develops AI-centric software applications.
As shown by Chip’s aforementioned Github repo study, this category has blown up in 2023, with hundreds of AI Engineering repos - AI interfaces, AI frameworks, AI agents and prompt engineering. There are many opportunities for software developers to make their mark in this area.
AI applications: Companies and developers in this space can create products and services that make it easier for organizations to leverage AI technologies and integrate them into their workflows. These are the AI-powered solutions across various industries and domains, such as tailored solutions in healthcare (an AI ‘nurse’ or elderly care doll), finance (AI financial advisors), manufacturing (warehouse robot or AI visual product inspection), transportation (self-driving delivery), entertainment, and more.
As AI capabilities advance, new and innovative AI applications will emerge, some with interfaces and use models very different from the standard chat-bot interface. Some of the most successful products may be those that take AI and use it to rethink ways of doing things.
Here too, there is opportunity for developers, but there may be even more for those with a mix of domain knowledge and development skill. Innovators with a combination of domain knowledge and development skill can create intelligent AI systems that automate processes, enhance decision-making, personalize experiences, or solve complex problems within their specific niche.
Lastly, even if you can’t code, you can be an AI builder in a niche where AI meets a specific domain: You can build custom GPTs, develop Character AIs, or develop the prompt templates to solve specific tasks. Again, knowing a domain niche to actually solve real problems will be more important than development skill.
As AI continues to permeate various aspects of our lives, the potential for innovation and impact in this field is vast, making it an exciting and rewarding domain for individuals and organizations alike. To summarize the different “AI Builder” opportunities, you can:
Be an AI model builder or AI model fine-tuner.
Be an AI Engineer or AI application developer: Develop AI interfaces, AI agents, AI frameworks, and/or AI applications.
Be a custom or niche AI product builder: Leverage AI platforms to provide prompt engineering solutions, custom GPTs, or custom AI agents to solve specific workflow issues to make AI effective.
Monetizing AI - the AI-based Business
There is another layer of value, not in building AI itself but in using AI to create valuable things, and then scale and monetize that value creation.
Since AI automates and mechanizes expression and content creation of ideas, creativity, and knowledge, the cost of creation goes down to near zero. However, nothing of value is push-button, and even if it was, you need to know what buttons to push.
That’s where the AI-based business model comes in. It is conceptually simple: Running AI well to solve niche problems is a skill, so use AI to solve a business niche, and scale that AI solution. Add unique value and you have a defensible AI Era business.
Let’s take the law. The AI startup Harvey is building custom-trained legal AI models for lawyers on top of OpenAI models. Additional training on 10 billion tokens of US case law data yielded a model that “97% of lawyers preferred over GPT-4.”
The Harvey startup is pursuing a “build” opportunity in making custom AI models for Law firms. A customer law firm that creates a superior AI-based legal workflow using such an AI application can serve up lower-cost legal solutions. Combine with a full-service offer and you can market and scale that AI-based legal solution, pass on lower costs, and win business and scale.
Scaling AI-based solutions for business productivity will be a source of competitive advantage for many businesses not directly in the AI business. Just as successful e-commerce sellers use Amazon, Shopify, Stripe and their internet, Saas and cloud infrastructure, so too, successful service provider companies will use AI under-the-hood to be more successful in their industry.
Monetizing AI in Creative Arts
AI can write tweets, blurbs, essays, and even novels. AI can compose songs and melodies, even if Suno is amazing because it was trained on all the music, including copyrighted music. AI can conjure up images and videos both super-realistic and impossible and surreal.
While this AI recapitulation of human creativity is both a challenge and an opportunity for creators, it is certain that AI will play a big role in the creative arts. Many who embrace it will be able to use AI to make better creations faster and cheaper.
AI film-making will become a big part of Hollywood. Pixar was built on computer animation, and you can expect a next-generation studio built on the latest in AI. In that vein, I have become a fan of the “Curious Refuge” YouTube channel, which dives deeper into AI film-making tools, technology, and techniques.
That trinity - tools, technologies, and techniques - is what creatives in any field will need to be on the look out for.
There are now tools that will take you from prompt to completed TikTok or YouTube explainer video. You can prompt for B-roll, prompt for a script, call up a narrator, and more. In other words, each step of the process can be automated, or even all of it.
The challenge in a world where bad content will be near zero cost, the trick will be to stand out and make something better. The future of AI is not an all-powerful oracle or auto-generated content. It’s savvy users using generative AI models and tools to accelerate production to create things better, faster, and cheaper.
AI-first creators will build successful businesses around AI as a creative co-pilot, doing as much as possible with AI. However, the human will still need to be in the loop. The best future creatives will learn that balance and scale the right formula, making AI plus human equal more than either could do alone.
The AI Apprentice and the AI Influencers
A swarm of AI agents can become your AI Apprentices, helping offload repetitive tasks, grunt work and more. The use cases will expand as AI gets better. If it can do your grunt-work, why not hire it out to deal with the work of others?
Any early peek of this was when Devin AI Agent Successfully Posted a Thread on Reddit Asking for a Job and Demanded Payment.
The step from leveraging AI for yourself to monetizing AI for others is a matter of scale. Consider AI agents for marketing or PR: First, figure out how to do marketing or PR better with AI; perhaps you automate communications or response funnels; then build out the AI Agent(s) into a robust automated workflow. Then scale - pitch your solution and take on more business.
Did you know you can create an AI influencer in minutes? Some of the ‘best’ AI influencers on Instagram are now making … millions?!? Example: Lil Miquelas, projected to make over $10 million this year!”
The AI influencer business is an extreme example of automated content creation being scalable and profitable. The low barrier to entry and the lucrative possibilities lead to a plagued of bots on social media. So while I won’t encourage the idea, the concept of how to make either AI apprentices or AI influencers into lucrative business is similar and simple:
Find a niche - a topic, industry, hobby, like-minded people, etc.
Automate with AI - Use AI agents to automate work flows and content creation.
Scale - Use AI and marketing to grow your audience and business.
Investing in AI
The venture capitalists are sinking a lot into AI; the rush is on to invest in AI. The recent Y-Combinator batch had most of their companies focused on AI. This landscape of AI startups shows just how crowded it is.
Will these investments pay off? They are all chasing pieces of what the VCs and others see as AI’s multi-trillion-dollar opportunity.
Cathie Wood’s ARK predicts $220 trillion in market value from 5 new technologies (the five they hype are AI, Energy Storage, Robotics, Multiomics, and Public Blockchains). Throwing around massive numbers is a sign of a hype-bubble, but we are in the early days and there will indeed be a lot of value in the AI and robotics business opportunities.
Here’s a new twist on an old joke: How do you make $3 billion in AI? Start with $50 billion.
The promise of the upside has led to tens of billions in AI startup investments, AI chip investments and more. The downside is the AI revenue hasn’t shown up yet. What if these bets miss the mark What if AGI ends up as elusive in the next 10 years as self-driving cars have been for the past 10 years?
The Sequoia observation is a warning: In the end, any business needs to have profitable revenues and make a return on investment, and the $50 billion invested (and more to come) is because the market opportunity VCs anticipate is huge - in the trillions.
If you are not a VC and just a stock-market investor, what to bet on? Just like Google and Facebook weren’t businesses at the dawn of the internet, many future great AI companies are either not yet public or may not even be born yet. While you wait for the next generation of IPOs, the Big Tech companies will present opportunities.
There are stock market winners in AI already. The best stock-picking bet would have been investing in Nvidia in 2020 when GPT-3 was trained on Nvidia A100s. If you missed that ten-fold increase in NVidia, more opportunities are to come. Nvidia itself is becoming more than a chip supplier, with AI Supercomputers, data centers, and even inference service revenue streams ramping up. Others (AMD, Intel, Broadcom) seek a slice of the AI chip business.
Microsoft and Meta have gained in market cap on AI news, while Google and Apple seem out in the cold. But the laggards could yet become leaders; imagine if Apple hooks Gemini 1.5 to Siri and has a hit on its hands.
As with the internet era, we will go through a season of IPOs, hope-and-despair cycles, with an AI bubble and perhaps bursting of it if too many businesses fail to earn their keep.
Either as an investor, venture capitalist or entrepreneur, there will be significant alpha in correct bets on the AI systems, applications and business models that succeed. Stay the course and succeed.
Summary
There are multiple ways to be successful with AI: As an AI builder, AI monetizer or investor in AI. Use your own mindset and skillset to figure out where and how you personally can best take advantage of AI.
AI presents many opportunities. My best advice - take them. As Warren Buffett said, "Opportunities come infrequently. When it rains gold, put out the bucket, not the thimble."