The Inevitability of AI in Every Facet of Your Business: 11 Deep Thoughts With Lasting Impact
The Inevitability of AI in Every Facet of Your Business: 11 Deep Thoughts With Lasting Impact
This is the Three Data Point Thursday, making your business smarter with data & AI.
Let’s dive in!
Actionable Insights
If you only have a few minutes, here’s what’s going to make your business smarter:
We’re handing over power to AI. And that’s okay; you just need to know when and what you hand over.
We should embrace AI 100%. We’re still a far shot from general artificial intelligence, so we should embrace it with all we have.
AI can be a vehicle. AI tends to reach parts of the world that previously were untouched by software; thus, it can be a vehicle to drive change.
AI is about the data. We’ve known it for some time, data is the center piece of AI, that doesn’t mean you should focus on data only, and not do AI. It means you gotta do them together by focusing mostly on the data.
Scientific breakthroughs matter. We haven’t seen all of it, in fact, I believe in serious scientific breakthroughs over the next decades that will turn our world around in terms of data and AI.
Knowledge of the data & AI space is still obscure. It helps to medidate on statements, you’ll uncover their true meaning only as time passes by.
20th Februar, it’s a warm Tuesday, and I’m somewhere in the Persian Gulf, relaxing in a hammock, reading through a book I long wanted to finish “The Daily Stoic, 365 Meditations on Wisdom…. .”
It’ a heavy book, each page has a quote from a Stoic at the top, that then is followed by a short meditation from Ryan Holiday and Stephen Hanselman on that topic. It’s kind of a short discussion, helping you to think more about the quote. To me, this kind of reading has always been hard. It requires me to read and reread the quote, then understand the thoughts of the two authors and relate them to my life, potentially coming up with a few examples of my own.
With the sound of the waves, the wind in my face, and the heavy and raw mountains of Oman rising in front of me, my mind started to drift off…across the sea, and soon it landed where it so often does: at data & AI.
Yeah, I know, I’m a geek.
I realize I’m a bit strange in that way; probably not everyone is thinking about data & AI when he’s in a hammock enjoying his vacation in the Persian Gulf, lined by beautiful nature. But then again, I do have a Ph.D. in math and have been literally buried in data & AI for over a decade every single day.
Strangeness aside, I realized two things. Whether or not I like it, I keep coming back to quotes & statements in the data space I’ve read or heard before. Again and again. I am meditating on them, changing and updating my opinion on what they mean to me.
But unlike the teachings of the Stoics, data, and AI thoughts are often obscure and hard to grasp. Data & AI are still intransparent and hard to access. Even if we in the space like to believe the opposite.
So while the idea of meditating on data & AI sounds geeky and weird, it’s kinda my thing, and I want to share it with you.
I want to share a collection of thoughts, quote from leaders in data & AI, and a few thoughts I have on them, to get you thinking. I have no clear answer to any of the questions they raise, and I keep coming back to them, revising my opinion over time.
The future tools of data
"Every medium becomes the content for the next medium. " - Marc Andreesen, in an interview with Lex Friedman
Told stories became the medium for written stories, became the content for theater, became the content for movies, became the content for DVD rental stores, and became the content for Netflix.
But what's the content of Netflix? Not just movies, it's also TV shows, games and more. Old mediums become the content for the new medium, the new tool. That’s what Marc means with this statement he has been repeating for years and years.
The world of data & AI looks turbulent, hard to forecast. Noone did predict the rapid rise of ChatGPT, and yet it is here, and it’s becoming better every day.
If you want to create a next-level business, think about mediums. If you want to know what tools you should bet on today, turn to mediums. It’s the reason I’m betting more and more on consolidated platforms like databricks and Snowflake, because I believe they will soon be the medium of mediums.
It is why I think the future of databases isn’t databases, but rather data stores that give me whatever type of database I need for the data I have at hand. And one of the reasons I’m critical of most database startups.
This mantra helps you to build something truly next level, and it helps you understand the direction markets will move into.
The future of power
"First is that AI is the first tool in history that can make decisions by itself. All previous tools in history couldn’t make decisions. This is why they empowered us. You invent a knife, you invent an atom bomb; the atom bomb cannot decide to start a war, cannot decide which city to bomb. AI can make decisions by itself. Autonomous weapon systems can decide by themselves who to kill, who to bomb.
The second thing is that AI is the first tool in history that can create new ideas by itself. The printing press could print our ideas, but could not create new ideas. AI can create new ideas entirely by itself. This is unprecedented.
Therefore, it is the first technology in history that instead of giving power to humans, it takes power away from us. And the danger is that it will increasingly take more and more power from us until we are left helpless and clueless about what is happening in the world." - Yuval Harari in an interview with Lex Friedman
Yuval Harari is in my opinion the clearest thinker on data & AI in our lifetime. I say that because I’ve been reading his stuff for almost a decade now, and he is a dedicated meditator like I am. Yuval has a unique gift for uncovering the painful and fundamental questions, the things everyone should meditate on.
We’re using AI and data to automate decisions; that’s the point of all of those tools. We are handing away power and responsibility to automated systems willingly. And that might not be a bad thing.
The question you should ask yourself is: which powers do you want to give away? In your personal lives and as a business.
And what powers do you want to keep, to multiply them? Because in the end, that’s why we’re doing all of this, to make our lives easier by getting a bigger hammer.
This hammer just happens to be pretty darn different than all the others and thus more dangerous.
It’s not yet general artificial intelligence
“is it good for everyone in the world to have access to the most frontier AI models? But right now these are still very basic tools. They're powerful in the sense that a lot of open-source software like databases or web servers can enable a lot of pretty important things. But I don't think anyone looks at the, you know, the current generation of LLaMA and thinks it's, um, you know, anywhere near a super intelligent.” - Mark Zuckerberg on the Lex Fridman Show
These times are certainly crazy, people call for a halt on all AI research, billions of funding is poured into AI, deep fakes are all around us. Events in AI seem to be going head over heels.
And yet, we also feel excited about the possibilities we see open up. There are some true everyday breakthroughs thanks to AI. Video subtitles, transcriptions, photoshop generative fills, GitHub CoPilot, and the list goes on and on. When I write a classified ad, I let ChatGPT do it, using AI voice recognition, because it’s becoming that good. It takes 30 secs and is five times better than what I can come up with in 15 minutes a year earlier. We all feel the potential, and yet we’re also scared - that’s how big changes look like.
What Mark is saying, is that right now is not the time to be scared, it is the time to distribute and democratize, to use everything we get. The stuff we have in our hands in 2024 is the webserver of the 1990s.
Don’t be afraid, use this stuff, distribute it and let everyone get a hand on it. Implement it in your business, now is the right time to jump onto the train, use it personally, just get going.
Let’s jump
Spaces like agriculture, hospitality, education, legal, construction, and manufacturing are primed for what we call “AI leapfrogging.” If executed correctly, AI has the potential to reach new demographics of users who were bypassed by the previous software revolutions.
If packaged intuitively, AI can go where no software has gone before. - Morgan Beller et. al. at NfX
Speaking of democratizing, the VCs at NfX have a consistent output of great thought and barely get any credit for it. I’ve been thinking through this statement again and again. Is AI really able to reach the non-tech-enabled industries? Is the most advanced of technologies able to go to the least-technological places?
It sounds like an odd pair, and I have no good explanation on the why, but from what we’ve seen in the past, it looks possible. AI transformed the agricultural landscape from crop development to driving tractors.
You should care about this. Software is enabling people, making it reach more parts of the world, or your organization is important. And if AI is a good vehicle to do so, that’s the way.
Maybe the best way to get a legacy oldschool business into business intelligence systems it to provide them with good (haven’t seen one yet!) conversational BI interfaces. Maybe you can convert old school sales people into CRM systems by giving them AI.
Maybe AI is a great vehicle to drive change.
You cannot exist without AI
“Facebook today cannot exist without AI. Every time you use Facebook or Instagram or Messenger, you may not realize it, but your experiences are being powered by AI.” - Joaquin Candela in wired
This thought might scare you, or it might excite you. Or maybe you’ve already digested it, and you think it is obvious.
I can tell you, I haven’t. I stumbled over it first when I was reading a book on the dangers of AI. It was meant to tell us how vulnerable all our experiences are to the exploits of AI systems.
And yet, for me, a different lesson stuck. I’m a data & AI expert, I know what Joaquin says is true for Facebook, and almost every other company. And still, every time I contemplate this statement, I realize, I’ve underestimated the extend. Again and again, I’m underestimating the reach of AI.
And you’re underestimating it, too. Your company would not exist without it, you wouldn’t have your internal systems, you couldn’t write code on GitHub, you couldn’t use your fancy CRM system, and nothing would work without AI; I doubt there’s a single piece of software today, that runs without a heavy dependency on AI, in one way or the other.
Fifteen million images make a difference
“The “it” in AI AI models is the dataset.” - jbetker, an OpenAI employee (not verified)
While the quote above might not be from an OpenAI employee, and while it may be relatively recent, the idea has been one I’ve been struggling with my whole professional life.
I’ve been struggling with it since I founded a startup to collect data for vacation rental owners and then provide recommendations based on it. I’ve struggled when building out an internal recommendation system for a large procurement platform, and I’m still in general struggling with it.
The evidence is overwhelming that it’s the data that matters, not the model. A weird consequence is also, that it’s the low paid people annotating data, not the 100x as expensive data scientists and ML researchers that make the difference.
Where I’ve failed in the past is to then conclude the most valuable thing is the data, but it is not. The data is nothing without the AI system that uses it, even if that might be pretty trivial from the perspective of the overall picture.
It’s been like this since AlexNet and 15 million images that made the difference, and launched the machine learning revolution, it is why Scale AI is a $7b company, and why the father of machine learning Andrew Ng is still all in on data-centric ML.
AlexNet wouldn’t be here without Alex; Scale AI wouldn’t be valued at more than a dollar without image recognition ML systems able to learn based on the annotated data.
It’s always been complex
“”Tell m what do do!” What advice can I possibly give? No, a far better request is, “Train my mind to adapt to any circumstance.”” - Epictetus, Discourses, also in The Daily Stoic.
We data people, me included, keep on highlighting the rising importance of data in decision making because of the growing complexity of the world, because of the quick changes, of the rapid increase of the rate of change.
Epictetus lived 2,000 years ago and was apparently pretty much giving the same advice based on the same idea. Things change too fast; there’s no use in giving advice on what to do, but rather help people adapt.
The OpenAI engineer quoted above noticed something AlexNet has been proving since the dawn of machine learning, maybe the complexity of the world is increasing, but maybe we’re overestimating the importance of it?
Maybe we’ve always lived in complex times and old ways and principles to still work. We’re keen on throwing the baby out with the water, but maybe old stuff is still gold.
The data space is closed and locked
“Blockchain networks are the only known technology that can reestablish an open, democratic internet.” - Chris Dixon in Read Write Own
Chris Dixon isn’t a data leader, and yet his insights permeate what I’ve been thinking about the data space.
I’ve invested a ton of time into open source as a business tool, the primary way the data space thinks it will advance towards an open and democratic internet. Mark Zuckerberg loves to promote open-source AI as the ideal.
But Chris got me thinking a lot about this lately. The fact is, open source in the AI space isn’t open or democratic at all. Facebook and Google sponsor 99% of the machine learning market with TensorFlow and PyTorch. AWS and the other big clouds are developing more and more lock-in pieces of machine learning tech to get people into their realms.
Almost everyone in data engineering thinks open source is the holy grail, and yet, 99% of the profits of the “modern data stack” are captured by companies like Fivetran and Snowflake, and almost nothing is left on the table.
Right now, the data world isn’t creating an open world, quite the opposite. The single piece of technology that seems truly democratic and open is PostgreSQL, invented 27 years ago. That’s 27 years of closing up a space that we think of as open and democratic.
We might think we’re working in the open, but we’re not; the data space is sadly one of the most controlled and closed spaces of our time. Don’t get caught up in the illusion it’s not.
Reproducing
Reproducibility and replicability are foundational to the scientific method. The greater the claim made using analytics, the greater the scrutiny on the process should be. In order to be able to trust the data, the process by which results were obtained should be transparent and reproducible. - Maxime Beauchemin
I’ve first gotten in contact with the importance of reproducibility in the DevOps context, where words like “immutablity” are used frequently. But I think the key idea to all of this isn’t the reproducibility, but rather the scientific method, the idea that there is a good process to discover knowledge.
Reproducibility is a word that gets thrown around a lot these days. People discuss how AI can reproduce previous results and how it can be reliable and not prone to hallucinations.
But I think Max always knew it’s not really about reproducibility but rather about the scientific method and, more generally, about discovering knowledge.
The most important thing in your work is that you get it done. The second most important one is the knowledge generated in the process.
Do you really need the Netflix algorithm to be able to reproduce what it showed you yesterday? Do you really need to be able to reproduce the analysis an analyst generated for you, and you made a sound decision based on it? Data might be stochastical. Data might only be 90% good. But the knowledge you generate is real.
The true question then becomes: Do you know how to generate knowledge from data that you might not be able to reproduce in detail?
Human-digital beings
“I believe that the merge of digital and human beings in the physical world is the future [...], in a way that will seem very natural” - Rana el Kaliouby in “Girl Decoded”
When Rana wrote down the quote above, she was talking only about the retail space. But I don’t think she thought just about that, after all, she’s building systems into our cars to cage our emotions, she’s basically creating driving feeling robots.
I keep on coming back to this idea because I’ve heard Yuval Harari say something similar that I still think is true: Machines are better equipped to be emotional than humans. They simply have more sensors.
So, given that machine will very soon be better with the emotional stuff, what does that mean for us? I don’t have a good answer. I just know, emotional AI is a super interesting field to start a company in. I know we can use emotional AI in most of our products already today. I know I have to get accustomed to machines that work with my emotional state - a power I might be willing to hand over.
What about you? What will you do when faced with a human-digital being?
Technology is 50% of it
“[The Nobel prized inventions of the transistor and the integrated circuit] were crucial but science alone wasn’t enough to build the chip industry. The spread of semiconductors was enabled as much by clever manufacturing techniques as academic physics.” - Chris Miller in Chip War.
What is it that drives good businesses forward? As a product manager, I’ve been thinking about that question a lot. One thing is for certain: Giants like databricks, Google or Microsoft are propelled forward by scientific breakthroughs just as much as by effective execution of product strategies.
Just like Bob Noyce coinvented the integrated circuit (and would’ve gotten the Nobel price, if he would’ve not died before) and then build a company around it to mass manufacture them, great data businesses are built on both, breakthrough research and amazing product strategy execution.
As a keen observer of the data space however, it strikes me as odd, that so many companies try to go by without one or the other. Dbt Labs tries to productize the most mundane thing ever: templated dbt. Snowflake essentially just does smart indexing (with amazing product execution). More academically inclined people think they can get by without good execution of a product strategy - and thus (in theory) good projects fail.
It looks to me, that history has a point, we need both. You should be wary of products that have just one or the other. If you’re a founder or product manager, keep in mind that in the data space, science and technology breakthroughs matter.
Stop!
Don’t leave yet. You’re an amazingly smart crowd of readers, so now it’s your turn to share a thought you’ve had for some time.
Let me know! (Either reply to the mail to write to me in person or leave a comment)
Here are some special goodies for my readers
👉 The Data-Heavy Product Idea Checklist - Got a new product idea for a data-heavy product? Then, use this 26-point checklist to see whether it’s good!
The Practical Alternative Data Brief - The exec. Summary on (our perspective on) alternative data - filled with exercises to play around with.
The Alt Data Inspiration List - Dozens of ideas on alternative data sources to inspire you!