Co-founder & CEO
Elon Musk has been banging on the proverbial ‘we need to be scared of AI’ drum, along with other prominent members of the AI community. Even Putin alluded to this recently. I largely agree with them, but not in a Hollywood way. The typical Hollywood doomsday scenario is:
1. AI acquires human-like capability
2. For some reason, it is a humanoid robot or some anthropomorphic form. Skynet is an exception
3. AI decides humans are a burden to the universe and they all must die. Depending on the movie, this might be payback for how AI have been treated by humans in the past — like the Cylons in Battlestar Galactica
4. The hero of the story saves the day, humans are safe and the world is great again or the world descends into a dystopian AI-dominated future where humans are ahem, mere mortals, who are expendable or exploited for resources a la Matrix
This narrative does not help the community’s quest for increasing AI capabilities, trending towards building Artificial General Intelligence (AGI). AGI will not emerge in humanoid robots seeking to kill humanity with crude weapons that go pew! pew! The emergence and effects of AI will be much more subtle and pervasive as we increasingly adopt aspects of AI technology in our products and daily lives.
- John McCarthy, one of the pioneers of AI
Netflix recommendations, Google ads, Facebook News Feed, iPhone battery optimization, algorithmic trading, dynamic pricing in your Uber app or when you look for flights — AI is everywhere and we don’t notice it. Even when we do notice it, we accept it as natural.
Nick Bostrom makes a wonderful case for the existential threat posed by AI in his book Superintelligence. A common refrain is — ‘AGI seems so far away, why should we start thinking about it now?’. Stuart Russell has a great response to this. Let’s say AGI is 50 years away. Now imagine that we get a message from some extraterrestrial civilization.
‘People of Earth. We will visit you in 50 years. Get ready’
Imagine the pandemonium that would follow such an announcement. Contrast that with the laissez-faire attitude we seem to have to the existential threat posed by AGI. We have no idea how long it will take for us to put checks and balances in place to deal with the existential threat posed by AGI. It could be 5 years. Or a 100 years. We should be worrying about this now.
Meanwhile, humanity needs to figure out how stay relevant. The term ’moonshot’ is business as usual when it comes to Elon Musk, and his latest venture Neuralink takes it to another level. It is an attempt to begin research on building brain interfaces so that we can augment our human capabilities with machines. This is the next logical step in a long technology discovery and adoption cycle. This started with discovering fire. Since then, humanity has been on a quest to augment human capability using external/artificial methods.
Neuralink hints at a future where humans and AI are merged in a cyborg-like scenario. While that seems scary, we are already cyborgs to some extent. The average person today has the ability to access information in an instant. Only the most powerful people in the people had this superpower a mere 30 years ago.
Gerald Ford: “I say, Dick, what is this DARPA? Can you find out for me please.”
Dick Cheney: “Yessir”
We are on the path to embracing technology and integrating it more deeply into our daily lives. The technologies we choose to develop, adopt and build upon follows the rules of cultural evolution and is dictated by the environment which provides the feedback mechanism — the capitalist ‘free market’. AI is next in line for large scale adoption and it has already begun. This has interesting consequences for product management, design and development.
To understand this better, let’s look at how software development has changed over the last 20 years. Here’s a typical web application from the mid-90s.
Take a minute to get over the barrage of information that passed for design in 1995.
To build something like this, here’s what needed to happen.
1. Write business logic
2. Store whatever is need in a database
3. Build the UI. Usually meant writing HTML by hand
4. Wire up the UI to do different things based on business logic rules
5. Find a server with a public IP
6. Deploy application
7. Deploy database
The monolithic application evolved into splitting up the problem into abstractions. The Model-View-Controller (MVC) architecture was one such approach.
1. Model — data model of the application
2. Controller — Controls the view, based on coded business logic
3. View — Render HTML based on controller’s instructions
4. Server(s) — hosts applications/databases
5. Development environment, source control and deployment
Scaling and reliability challenges forced a re-thinking of the infrastructure behind the web application. First came master-slave databases. Then the load balancer. Vertical database scaling. Horizontal database scaling. The Cloud! Auto-provisioning and scaling of servers (AWS). Service Oriented Architecture (remember that?). Micro-services. Now: Serverless applications.
The large-scale adoption of the Agile development methodology by the software industry drove the need to shorten iteration time, development cycles and enable experimentation. This led to the entire industry figuring out how to deploy, version and experiment on applications really fast. Automated testing. Automated deployments. Continuous integration. Experimentation frameworks. An outsourcing of non-core activities via abstractions like SDKs and APIs for things like the development framework, analytics, crash reporting, customer outreach like support chat, push notifications, sending emails, A/B testing, staged rollouts.
The creators and enablers of these abstractions capture the most value.
These technology developments have allowed the industry to work at increasing levels of abstraction. Doing so has resulted in being able to build upon the progress so far, which leads to even faster progress. This ends up becoming a power law where progress is exponential. This concept is elegantly captured by Ray Kurzweil’s law of accelerating returns.
‘Building a platform’ is back in vogue. Problem is — no platform company was successful because they decided to build a platform from day 1. They all built a great product first. Building this great product allowed them to pull off building a great platform and capturing the market. Unity solved rendering 3D scenes for games before becoming a platform for game developer services. Amazon built a massive scalable e-commerce site which needed an internal version of AWS to succeed. AWS is the culmination of years of internal struggles and pains that Amazon went through when scaling amazon.com. Facebook is a great ad platform because they build a compelling product that users want to use. Same for Google.
This has resulted in the entire industry moving up the abstraction chain. Someone who knew how to manipulate HTML and build CSS was a prized developer in the 90s. That is now easy to do. In general, the industry values the ability for engineering teams to build products that solve real problems for users and create value for the company. And do this in a sustainable, long lasting way.
This is where product designers and product managers come in.
We are at the beginning of the abstraction building process for product designers and product managers. In the same way that parts of software engineering/web development got abstracted out, automated, and platform-ized, the same thing is happening with the product designer and product manager.
Now. Back to AI. We now have — product design + product management driven by increasing abstraction + increasing AI capabilities. The next logical step on the abstraction ladder is the incorporation of ML/AI in product design and management. With AI and machine learning techniques becoming increasingly commoditized, the capabilities available to product managers will increase. How we choose to blend these techniques with human creativity will define the future of product design.
Human consumption + creativity + AI are interlinked.
1. Human consumption drives the technology marketplace. This follows the framework of culture evolution (feedback mechanisms, iterations, the best survive and are built upon)
2. Products are built for human consumption. Increasing levels of abstraction don’t happen in isolation. They serve the needs of the marketplace
3. The future is to blend human creativity and AI capability to make better products for human consumption
A fundamental assumption here is that human consumption dictates market forces, because human consumption is the economy in which we operate in.
When it makes sense to have AI control aspects of the product, it will happen because market forces will move in that direction. Human involvement and value will continue to shift. We need to adapt and continuously put ourselves in places where humans are able to climb the abstraction ladder to stay relevant. Luckily, the product management and design community embraces continuous learning and iterations as a way of life. I have no doubt that we will continue to stay relevant.
I was explaining to someone who just turned 21 that the AI/ML space right now feels like the internet space in the early 90s. There are lots of companies — big ones and startups, trying things out and we have no idea who’s going to win. We don’t even know what winning looks like. An Amazon or Google equivalent will emerge from somewhere, probably where we least expect it. I was feeling quite smug about my pseudo intellectual assessment until I was met with a blank stare. This person was born in 1996 and had no idea what I was saying!
The future is going to be awesome. We are only seeing the beginning of the impact of AI on product design. Content recommendations, dynamic pricing, targeted promotions. The world is moving towards providing randomness to users, with the hope of giving a personal, serendipitous experience. Who knows where this will go.
This is why Prashast, Indradeep and I started Product ML. We want to shape the future of human creativity in product design, married with AI and ML.
More articles on Machine Learning and Product Experience:
At Product ML, we believe that all products in the future will be dynamic. We’re building a platform that redefines product management and user experience, starting with dynamic difficulty in games using machine learning.