AI Hype: Is the Tech Industry Overvaluing AI Hardware?

Introduction

Ever since Chat GPT was released in 2022, Silicon Valley and Wall Street alike have been obsessed with AI. According to Crunchbase, AI startups raised $50 billion from venture capitalists in 2023 alone. On top of that, established tech companies are spending tens of billions of dollars to develop their own AI tools.

The consensus among tech and finance professionals is that AI will transform the world economy, generating trillions of dollars of economic value. This justifies the tens of billions of dollars being invested in the industry today.

However, there are two problems with this. Firstly, most AI is not nearly as advanced as companies want you to believe. For example, a couple of years ago, Amazon released cashierless checkout at their Amazon Fresh grocery stores.

They claimed they could monitor what items a customer puts in their shopping cart with their advanced AI computer vision. As it turned out, up to 70% of transactions need to be verified by offshore workers manually reviewing the camera footage. This is economically unviable, and Amazon recently announced that it will end cashierless checkout.

Even when AI can do impressive things, it’s often hard to find real-world applications. For example, last year, Google demonstrated their Gemini AI, showing it doing various gimmicks, including keeping track of a paper ball in a cup. As it turns out, the demo was staged, and Gemini’s capabilities aren’t nearly as great as the video would lead you to believe.

But even if Gemini could do what Google showed in the demo, it’s hard to think of an economically significant use case for an AI that can keep track of moving cups and play Rock Paper Scissors.

The recent hype around AI has led to billions upon billions of dollars being invested into some frankly bizarre startup ideas. In this video, we’ll look at some of the weirdest new AI products being developed and see why the recent AI investment surge may be a massive misallocation of capital.

AI’s Current Capabilities

There are many different types of AI, but the main type of AI that’s caused the hype over the last two years is generative AI. Thus far, generative AI has two main functions. The first is chatbots, which can analyze user-generated text prompts and respond with text of their own. This can also be applied to writing computer code. The second main type of generative AI is image generation, which can create fantastic images based on user text prompts.

There are a few obvious use cases right off the bat. Tools like GitHub Copilot use generative AI to help programmers write code faster. However, for things like Chat GPT, it’s a bit harder to find use cases. They can make great party tricks, such as writing posts or creating whimsical stories on the fly, but the problem of so-called hallucination is a major barrier to using AI chatbots for real-world applications.

Even the most advanced AI chatbots today have the propensity to make up blatant falsehoods. This makes them effectively unusable for applications where accuracy is important, such as journalism or writing legal documents.

You can use AI to create fictional literature, for example. There are websites where you can buy AI-written science fiction and fantasy novels. Newsweek interviewed one of these authors who use Chat GPT and Claude to write 97 books over the course of nine months. He sells them online as eBooks for $2 to $4 per copy.

During this time, he made $22,000 of total income. We don’t know how many hours he spent wrangling the chatbots to make these 97 eBooks, but $22,000 in nine months isn’t exactly a great salary. Most critics view AI-written books as poor quality, and not that many people are willing to pay for them.

There’s a similar problem with image generation. Today, almost anyone can become an AI artist almost overnight. There are tons of people trying to sell AI-generated artwork on platforms like Etsy, but because the supply of AI art is so massive, most of the artists generate minimal, if any, sales. It’s hard to see how you can make real money with AI, and unless there are any real revenue-generating use cases, it’s hard to see how the industry can be viable.

With that being said, the industry is gradually trying to make tools with real-world applications. Many startups are developing customer service chatbots where you can manually input data that the AI will reference for frequently asked questions. This decreases the rate of hallucination but also restricts the chatbot to only answer questions within a limited domain.

The main applications where chatbots are currently being used are customer service. While this could displace many customer service representatives, it’s hardly world-changing.

Some smartphone makers are integrating AI into their cell phones. For example, Samsung licensed Google’s Gemini AI model to create Galaxy AI. This has a few features, including being able to circle an object in a picture and search for what it is on Google. It also has a live translate function where you can talk into the microphone and it will output an AI voice in a different language. It’s basically Google Translate plus Siri. While these are cool features, they hardly seem like game changers.

To make back the tens of billions of dollars invested into AI hardware and development, the industry needs to come up with far more compelling use cases. In an attempt to do this, a whole host of startups have come out with dedicated consumer hardware devices specifically designed for AI.

AI Hardware

In 2018, two former Apple engineers founded a startup called Humane Inc. For the first few years, they remained secretive, with nobody outside the company knowing what they were working on. Things started gaining steam for them in March of 2023 when they raised a $100 million funding round. The funding round included a wide range of blue-chip venture capital funds as well as OpenAI founder Sam Altman.

As it turned out, Humane was partnering with OpenAI to create what they call a wearable pin. This was a consumer hardware device specifically built for AI. In November of 2023, they officially unveiled the device. It’s a small computer with no screen that can clip onto your shirt.

It has a camera, a speaker, a microphone, and a projector that projects a display onto your hand. The camera will track your hand motions to navigate the menu. You can think of it as conceptually similar to Apple Siri or Amazon’s Alexa. You can use voice commands to ask it questions, search the internet, read your emails, dictate text messages, etc.

The device is currently only available in the US and costs $700. You also need to pay a subscription of $24 per month. This gets you unlimited talk, text, and data through T-Mobile. The Humane AI pin cannot connect to a smartphone via Bluetooth and has its own unique phone number. Thus, it would only be viable if it replaces your smartphone. Otherwise, you’d have two different phone numbers and need to pay for two different cellular plans.

As soon as the device began selling in April of 2024, the reviews were brutal. Critics complained of its slow processing speed, overheating, poor battery life, and frequent mistakes. According to The Verge, the only thing the AI pin can do reliably is tell the time.

But even if it was extremely fast, had great battery life, and could reliably do all the things as advertised, what’s the point? For $700, you can buy a capable smartphone that can do everything the Humane AI can and much more. For example, you’ll never be able to play mobile games, watch videos, or record high-resolution photos and videos.

While sales numbers have not been released, I would be shocked if they sold more than a few hundred units given how excoriating the reviews have been. As of May of 2024, just one month after launching the product, the company is reportedly exploring strategic alternatives.

The idea behind the Humane AI pin is laughably dumb. A 12-year-old could probably have told you this product would flop. Yet, they were able to raise $240 million from the likes of Microsoft, LG, Qualcomm, Tiger Global, and Sam Altman. How could so many people invest so much money into such a dumb idea?

Solution Looking For A Problem

To understand the mindset, let’s look at what Humane’s co-founders told CNBC shortly before the product launched:

“At the heart of it, we believe that AI is really going to drive the next generation of compute, but that in order for you to fully get the power of it, we needed to rethink everything. So, we don’t do apps, right? There’s a burden that comes with having to know that an app exists, searching for it, downloading it, managing it, logging into your account.

“AI allows all of that to disappear. So, the way that our platform works, it’s both a device and a platform called Cosmos, is that we do all of the heavy lifting. All you have to do is ask for something, and we do it for you. That really is about streaming things on demand for you. It would find stuff from apps that exist, or this is like no apps in the world—no apps. It’s all about streaming AI experiences.”

The entire selling point of the Humane pin was its focus on AI. They licensed OpenAI technology to provide what they call AI experiences to their customers. The AI experiences are similar to Alexa and Siri, which have already been available for many years. The co-founder’s explanation that apps are a great burden for smartphone users is almost comical. The idea that consumers will give up their smartphones in favor of a screenless pin is absurd.

Startups aren’t the only ones trying to create new AI hardware. Samsung recently released an $800 vacuum cleaner powered by AI. It uses AI to sense when you’re vacuuming carpet versus a hard floor and adjust the suction power accordingly. This supposedly increases battery life.

At the 2024 Consumer Electronics Show, they unveiled a new AI companion called Ballie. The product is not available for sale yet, and its pricing has not yet been announced. You can give it instructions with voice commands, and it can connect to smart devices within your home. It also has a projector whose main use case appears to be playing workout videos.

The product looks like it’ll suffer the same fate as the Humane AI pin. It’s a solution looking for a problem and doesn’t do anything useful you couldn’t already do with a smartphone. When you have a hammer, everything looks like a nail. The tech bros have bought into their own AI hype.

They think the technology is so great that anything with AI in it will automatically be a great success. That’s why they fund things like Humane AI, which anyone outside of the Silicon Valley thought bubble could have told you was guaranteed to fail.

Conclusion

This is highly reminiscent of the dot-com bubble from the late 1990s. Internet-based startups were able to raise venture capital funding regardless of how bad their business plans were. The vast majority of these companies failed, with investors suffering billions of dollars in losses. History doesn’t repeat itself, but it does rhyme. With tens of billions of dollars being pumped into questionable ventures, we can probably expect similar results.

Alright, guys, that wraps it up for this video. What do you think about the new AI hardware devices? Let us know in the comments section below.