Is ChatGPT This Year's Bitcoin?

This is The Startup Breakdown, the newsletter where we breakDOWN startUPs (just had to make sure you appreciated the word play). By joining this rapidly growing community of dozens of curious people, you're getting firsthand access to my observations and opinions on the current state of the startups and venture scene. If you'd like to receive these newsletters directly in your inbox once a week, go ahead and subscribe to never miss an email!

Happy Monday, folks.

Pretty big news for anybody looking to incorporate AI into their hot new birdwatching app: ChatGPT is launching its API.

This update will allow developers to built ChatGPT into their products for everything from improved customer service chatbots to more complex tasks like recommended shopping lists based on a prompt. Thus far, popular apps like Snapchat, Quizlet, and Instacart have already announced plans to incorporate it, and plenty of other large and upstart services alike will be soon to follow.

It’s a relatively cheap innovation at a cost of ~$.002 for roughly every 750 words (or just one if you choose Supercalifragilisticexpialidocious) entered by end users. For developers with particularly large data needs, OpenAI has stated that it will also have a flexible plan to allow apps to utilize specific data, limit the length of customers’ conversations, and make various other UX-improving adjustments.

Though the chatbot has been available for a few months now, the OpenAI team has been slow to launch this feature, citing a prioritization of quality assurance over speed of shipping new products. Translation, they weren’t going to release it until they were confident that Snapchat’s My AI wouldn’t tell 14 year old girls how to build car bombs with their mom’s nail polish and some elbow grease.

In the same announcement, the team also updated its terms of data usage following some backlash over its prior policy of using user inputs to help further train the model. The company will no longer use your data by default but will offer users the ability to opt in. Yeah, maybe go ahead and clear that chatbot conversation history…

While on the topic of AI, would love to hear from anyone either working on cool projects in the space or interested in exploring. Have a lil something something I’m working on 😉 

Let’s talk health.

Most people invest quite a bit of money into their health every year, whether in nutritious foods and medications or in those $50 magical healing crystals which look suspiciously like the gravel outside the holistic healer’s front door.

It should come as no surprise, then, that investors are also looking at spending quite a bit of money to have skin in the game in such a lucrative market. Unfortunately, they were feeling a bit under the weather last year as the total raised by startups in the space plummeted from 2021. What’s new?

Companies in the space raised just $15.3 billion last year compared to $29.1 billion the year before. While it might sound like last year was some apocalyptic, The Last of Us-style wasteland of a fundraising environment, 2022 seems to be really just be a reversion to the mean. 2020 saw $14.1 billion in investor money flowing in, though approximately $14 billion of that was prolly spent on masks and hand sanitizer.

Where will the money go this year? Experts are predicting sectors like alternative care, drug development tech, and healthcare workload management software will dominate while DTC health companies will really struggle. Apparently, inflation DOESN’T just apply to eggs, and higher costs will make production of these goods more expensive while also impacting demand as consumers choose those $10 juevos over hair volume restoration creams.

Another interesting theme is widespread expectation for greater clinical due diligence on the part of investors looking to make deals. With money being tight (#🤙), investors are less willing to bank on moonshot ideas without proven science supporting them.

Speaking from my own experience on the VC side of things, nobody on our team was remotely qualified to make decisions on the viability of any healthcare product that sent us a pitch deck. For any biotech professionals out there, giving you the billion dollar idea of a VC consultancy service to help investors determine whether those gas station pills really can make you live forever.

Another reason for the increase in DD has been the rise of some… ~less honest~ founders across sectors. One in particular has become the well-established poster child (that’s going to look way more clever in like ten lines), and luckily, she is in the news, too!

When you hear the term Girl Boss, your mind likely goes to Beyonce, Daenerys Targaryen, or that chick on TikTok who keeps on popping up on your For You Page promoting a marketing plan which oddly resembles those pointy structures in Egypt.

However, possibly the biggest Girl Boss of them all is Hulu’s and Apple’s favorite founder, Elizabeth Holmes.

The infamous founder of fraudulent biotech company Theranos is back in the news as she has asked for more time before she must begin to serve her 11-year prison sentence as she prepares to give birth to her second child. Her initial trial was delayed as she gave birth to her first son.

To be fair, her defense team is also convinced that numerous legal infractions throughout her trial will also help to mount a defense to reduce her sentence, but all I’m really getting from this is you don’t want to go to jail? Just start popping out babies. #InfiniteLegalGlitch

Tiger Putts.

Elon Tweets.

Cade Invests.

You subscribes.

You’re all going to rush to hit the unsubscribe button or leave me a bunch of mean comments on Twitter. Probably both.

Just hear me out, though.

Generative AI is this year’s crypto bubble.

To be clear, I’m a firm believer in the long-term potential of the technology. I believe that even within 10 years, artificial intelligence has the potential to completely reshape our lives through its ability to empower the individual to accomplish and build things that they don’t currently have the capacity to do, and it will make people and companies far more productive, theoretically enabling us to take more power over how we spend our time.

I have the same high optimism for web3. However, it’s too early for both.

If 2021 was NFTs and 2022 was the year of the metaverse, 2023 is the year of AI. Check social media, your inbox, or any other internet source of indisputably correct news straight from the mouths of experts with 52 followers, and chances are that you’ll see people talking about the many applications of the technology.

There have been plenty of cool new tools to emerge in the space over the past few months. Beginning with DALL-E this past fall when everyone with a phone could finally recreate their worst fever dreams with a few seemingly descriptive phrases, thousands of tools have since been released for things like writing content, producing transcripts of audio, and generating code. New products are released ever day, becoming even more niche by the minute.

No tool has gotten more attention than the tremendous achievement that is OpenAI’s ChatGPT, the AI chatbot that can answer the universe’s most burning questions like “Is the Loch Ness Monster real?“ or “Can I microwave a red solo cup?” in seconds.

Now, everyone is suddenly an ML engineer pumping out “groundbreaking” services. I have personally bookmarked tabs for generating Excel formulas, plannings weddings (including writing the vows with GPT), and of course, the now-banned endless AI-generated Seinfeld episode on Twitch.

There’s been a rush of content creators looking to capitalize off of the salience of the space with hundreds of newsletters, podcasts, and TikTok accounts dedicated to covering all of the latest developments and new tools. Each has amassed thousands of followers.

Most big tech companies have entered the arena, notably Microsoft through backing OpenAI and competitors like Google and China’s Baidu producing their own versions of robo-genies. There are even reports that Elon is founding a competitor to OpenAI (the nonprofit which he cofounded lol) to produce a less-censored, possibly even more unhinged version of ChatGPT.

Consumer-facing brands with over-capitalized marketing teams have sought ways to make headlines by rolling out their own applications incorporating the tech, including the mentioned Instacart’s customer service chatbots and Snap’s AI assistant tool. Not even your 13-year old cousin is safe from having it shoved down his or her ~changing~ throat.

Business leaders like McKinsey are racing to publish reports and outlooks on the technology’s potential and strategies for implementing generative AI into workflows. Financial institutions have invested into building out their infrastructure and support teams for the tools.

Where there’s hype, there are starving investors looking to cash in. Last year, investments crossed $2.6 billion, a more than 425% increase from 2020. Most of this money went to large, established startups that are building the massive, data-intensive base models, but any startup with the word “AI” in its pitch deck is also raising millions from salivating investors.

Does this whole story not sound a bit familiar?

In 2020, crypto prices skyrocketed alongside the stock market as bored investors had nothing else to do while in lockdown and were armed with stimmies. These factors also largely fueled the craze that was NFTs in 2021. Startups like Yuga and Dapper became blue bloods, raising millions for their Bored Apes and Punks.

Twitter timelines were nothing but “wen moon?” and “gm” coming from animated jungle animal profile pictures. Every project flaunted its innovative roadmap with plans for movies and video games and merch and ~community~

There was a blizzard of content created to cover the space, from Youtube creators to the infamous TikTok financial advisors telling kids how to get rich fast by just buying the jpeg and selling it when the price goes up. Truly brilliant stuff.

Every brand wanted to be hip and got in on the action. None could compete with the cultural icon that was 2019 Wendy’s Twitter, but everyone from Coca Cola to Porsche to even the NBA was getting involved with the crypto craze to varied success. In reality, few were actually providing their customers with any utility of substance.

Eventually, even the most change-averse of financial institutions got involved as banks and funds opened crypto operations. Thought leaders like Bain published papers on the web3’s power. “Blockchain” and “web3” were the biggest words at the center of the word cloud, and VCs ate it up. 2021 saw the industry raise $26 billion, and even as recently as Q2 of 2022, the industry brought in more than $5 billion in institutional funding.

We’d seemingly reached the tipping point, and full-scale decentralized society was here.

The similarities in each’s rise are uncanny. They followed the cycle of a big development ➡️ social media hype ➡️ oversaturation of startups and coverage ➡️ consumer brands riding the wave ➡️ slow-moving institutions adopting ➡️ VCs pouring way too much money in, creating unjustified valuations ➡️ crash (?) as we realized that there was not enough utility yet

In the peaks of each’s hype cycle, the technology had (has) some major deficiencies that many are more than happy to sweep under the rug.

In web3, a major challenge to adoption has always been bad actors. There were just as many rug pulls as there were projects with meaningful communities. An estimated $2.8 billion was lost through these scams in 2021, and the numbers have gotten larger as more recent incidents of frauds have come from massive institutions like FTX. (side note, I call dibs on SBF for Halloween).

Difficult onboarding experiences and technical concepts like wallets and Layers further complicated the space, leading to rampant hacks and breaches, and high gas fees from massive demand for block space contributed to Blockchain having a large net negative on the environment as the power necessary to power these validation processes from computer systems rivaled that of many countries.

We (myself absolutely included) were all too quick to brush these concerns off as Monkey Dollar shot up 40% on a daily basis, but once the luster of quick cash wore off, the downsides were all too apparent, though the remaining web3 community has made it priority number one to address them.

With AI, the story is the same. As was demonstrated in Google’s public demo of its Bard chatbot, these systems have the tendency to pull info out of their ass. They’re also easily maliciously coercible and can be made to give literally societally threatening information like how to build bombs or how to make your own Arby’s roast beef sandwich.

Is that the kind of stuff we want in the hands of society?

The information being used to train these models largely comes from social media and internet culture, not exactly a hotbed for commendable civic behavior. This results in racist, sexist, and otherwise biased outputs that further disseminate values that we are actively trying to weed out of public discourse.

Finally, AI suffers from the same negative environmental impact as does crypto. Training a single system requires emitting 250K pounds of carbon dioxide, and the industry has as negative an effect on the environment as does the notoriously harmful aviation industry. Even worse, this impact is doubling every few months.

As long as the technology is lauded as this final stage of societal development that will ensure we’re all sitting on golden toilets, nobody is going to bat an eye. Once the space hits a rough patch, though, there will be a race to address these barriers to the technology ever reaching its full capacity.

web3 popularized the concept of “utility.” In essence, it refers to “wait what am I actually getting from this?”

Products without any tangible benefits for their users skyrocketed amidst the broader ecosystem of FUD, and this had the effect of pumping up the market like Violet Beauregarde in Willy Wonka until the fears of inflation impacted the entire market, hitting the speculative spaces such as crypto the hardest.

Very few of these NFTs or DAOs offered any sort of real life benefit (apart from being part of the group that nearly bought the constitution and all).

Unfortunately, we’re in this same stage with generative AI. Yes, it’s cool to see the 10th tool that lets me see what I’d look like with a buzzcut (pretty darn cool if you ask me), but in reality, these are offering about as much real world utility as CryptoZombieKittens did.

As for the brands looking to implement the technology, how much value is actually being added for end customers?

Most big brand NFT drops came with no real utility. Everyone from Gucci to Budweiser played the game because everyone else was doing it, the culmination of a 15 minute marketing meeting where the C-Suite asked the token Gen Z intern what was hot right now and said intern remembered the wave of PFP NFTs on his or her Twitter feed.

Similarly, what is an AI assistant adding to the Snapchat app? The product announcement was something along the lines of “Our tool can help you come up with birthday present ideas for your best friend.” I guess you get to save the two buttons it takes to open up Safari?

Most of these tools will burn out. There is too little differentiation and not enough benefit, and many will fail. Already, we’re seeing some begin to shut down as they’re plagued with bugs, and maintenance efforts are no longer worth it.

We need real utility. There’s too much froth helping to support these “developments” that is unwarranted when they’re nothing but a fun fad more comparable to Flappy Bird than to the dawn of the internet.

We need real use cases that are going to warrant the onboarding process for the >95% of people who aren’t actually using the tech to deeply change their lives for the better, both in AI and in web3.

Again, I must reiterate my absolute conviction in the long-term potential of both technologies. They truly have the capacity to change society for the better, and I am perhaps overly optimistic in believing that this change could occur in even the next decade.

However, we’re not there yet. That doesn’t mean stop developing in the space, but unfortunately, that does spell some upcoming short-term pain where the last suckers on the train get left steering an empty vehicle with no brakes as it hurtles towards a mass reset of the AI space.

TLDR: A bubble has arisen around Generative AI, one which shares eery parallels to what we saw in 2021 and 2022 with web3. While each technology has the potential to completely change society for the better in the long-run, neither offers enough of a value-add to the everyday person today to warrant overcoming the technological learning curve and usher in widespread adoption today.

Cheers to another day,

Trey