Welcome back to The Future of Computing. This was a two part series around the potential paradigm shift in how we interact with technology. The first article, covering a talk by Andrej Karpathy on the future of computing, is here.

Today I’ll be talking about how feasible this future of computing is. Specifically the next generation of virtual assistants. Looking at them from a lens of a disruptive technology and how that plays out in the market.

Before we begin

I need to point out the ‘new kid on the block’ in terms of AI wearables, Friend. This is NOT what I’m talking about when i’m thinking next-gen wearables, but thought it’d be fun to share. I actually find this idea quite scary. I’m looking at ones focused on being a logical and productive assistant. Like the Rabbit r1 (even if they have their own issues).

What is a Disruptive Technology?

We often hear the term ‘disruptor’ or ‘disruptive technology’ when it comes to new companies. Everyone wants to be a disruptor or someone who changes the game but it’s often not as sexy as it seems. In The Innovator’s Dilemma, what is now the go-to theory for disruptive technologies, a few key traits of disruptive technologies are explained. When compared to the current generation of products they are:

• More expensive or less convenient (or both)

• Have the interest of a small market

• Have smaller profit margins

That doesn’t sound like the next big thing but rather every startup that eventually fails. So how do they ever become popular and disruptive?

Becoming Disruptive

A core reason why some technologies are so disruptive is not in spite of those issues but actually BECAUSE those issues exist. They create a little protective bubble around the innovation so large companies don’t have an interest in touching them. Let’s look at the most famous example.

Kodak

I think the most famous example of a disruptive technology sweeping away a major company is Kodak. The usual story is Kodak invented the digital camera but they were stubborn and lazy so they never really made any money off of it. Then eventually everyone else saw the obvious sign that digital cameras were better and competitors ran them out of business. Sadly, that’s far from the truth.

When digital cameras were introduced not many people had computers. And the cameras were 10x the price of a film camera. People weren’t tech-literate then so it was difficult to use. And it was so expensive to produce a digital camera that Kodak didn’t make much money off of them. Plus they weren’t making any money on the continuously selling of film anymore. So by almost every account, it was a bad business plan to try and grow the digital camera market.

Essentially the question became "should we make better products to make more money from the customers we already know or make less money making worse products for people we don’t know too well?”

I think we know the answer to this question 10 times out of 10. But these issues aren’t the same for a startup.

From Niche To Mainstream

A small startup doesn’t have any of these problems. No product, no revenue, and no existing customers. So they can take the chance. As they find the people who will spend money on the innovation, they can refine the product and their marketing speech. Then for it to become truly disruptive a few things must happen:

• Initial small market interest must grow into the mainstream

• Product must become noticeably cheaper or more convenient

First, the initial small market interest must grow into a bigger market. These are usually causes that are external to any product or brand. Although sometimes a large amount of advertising or evangelizing an idea can make this happen. But it’s usually complementary things that change and cause the product to be more accessible. For Kodak, it was that more people started having computers so more people could download the picture off of digital cameras. Plus as people became more tech-literate, the learning curve of a digital camera decreases. Making this previously scary product actually pretty accessible.

Second, the product must become noticeably cheaper or better than the current generation. Again this can be both inside and outside of the company. For digital cameras, companies began making their processes more efficient and their products better but the biggest cost savings was that computer parts were getting cheaper. So they could build their cameras for much less than before. Again making it more accessible and worthwhile to a wider audience.

Because of this, it’s actually easier for a startup to work with an early disruptive technology. They have access and are interested in smaller markets and don’t have any customers to alienate with their higher cost and less effective products. They can initially attract the innovators and early adopters in a space and then grow from there.

So how does this look for next-generation virtual assistants?

How AI Could Disrupt The Smartphone Market

So I say ‘could’ pretty notably because the truth is we don’t know. There’s a mix of product innovation and market forces at play. And understanding market forces is predicting the future, which no one can do.

I’ll start with the products and see if they land in the land of a disruptive technology. And then can see if the market could change in their favor to become mainstream.

Product

To start we’ll look at the Rabbit r1. Even though they’ve had a lot of their own issues in terms of security and product execution. I’d like to evaluate like all their promises came true and see where they would land.

First, the rabbit r1 was the first of it’s kind to make a phone without much of an interface. It has a single wheel for selections, a camera, and is controlled by your voice. It promised things like being able to order an Uber, food, and search with just your voice commands (although it’s only ever accomplished search). So while the idea is to make a phone more convenient one day, for now it’s definitely less convenient

Second, the market is pretty small to start. They sold an impressive 130,000 units of these at launch. For reference, Apple sells about 200 million iPhones a year. So definitely a smaller market.

Third, these sold for $200 while an iPhone sells for $1099 with a 120% markup. Making Apple ~$500 per phone. I find it hard to believe that rabbit is making that much off of the r1s.

So they’re making a less convenient product, addressing a smaller market, and making less profit.

Market

AI for productivity was a dream just a few years ago. And now ChatGPT and AI integrated tools are everywhere. But it seems like the hype or excitement in those tools started to fizzle out. So can it start breaking ground in more consumer applications?

I showed in a previous article that companies like Character AI are already breaking ground in moving this from a professional and productivity tool into more of a companion. While I’m still skeptical that something like an ‘AI friend’ will ever become mass market, I can see the pendulum swinging back and we’ll see a mix of ‘companion’ and productivity start emerging in AI products. As the use of AI in work and school become more useful and ubiquitous, I could see people becoming more comfortable with it and having it move more and more into their personal life. Slowly gaining a bigger market outside of productivity over time.

If there’s one continuous theme in information technology, it’s that it gets cheaper over time. With each new model release from companies like OpenAI or Anthropic, the older models also become cheaper. As they optimize their infrastructure and create smaller or ‘turbo’ models. There’s a suite of models that are relatively affordable without losing much of their intelligence. Now Meta has open sourced their Llama 3.1 models for free.

It seems like the current race in AI is to not only to make the most intelligent model but also the cheapest or most accessible ones as well. This will continuously drive down the price of using these models and allow for cheaper products to be built off of them.

So what does this mean?

It seems like this new wave of virtual assistants, which if become intelligent and usable enough can mostly replace a smartphone are an interesting technology to keep an eye on. If enough people become comfortable with and begin to trust these models, we could see a rather small and niche market grow into something that is an everyday product.

Thanks For Reading

I hope you enjoyed this edition of Subconscious. If haven’t already subscribed and would like to stay up to date with all things AI, be sure to subscribe. I’ll be moving this publication to every Friday now to keep up with demand. So I’ll see you next week!