What Comes after OpenAI?
Michael Spencer
A.I. Writer, researcher and curator - full-time Newsletter publication manager.
Hello Everyone,
In my pursuit for new perspectives on A.I. at A.I. Supremacy, I’m very interested in getting the perspectives of analysts, Venture Capitalists, emerging tech thinkers, think tanks, A.I. ethics researchers, and even financial investing experts and writers. I'm lucky enough to have some very talented guest writers willing to share their unique angles as guest posts.
?? From our sponsor: ??
TED Tech: An Event for Humans Operating in the Age of Uncertainty
Join TED's first all-tech event bringing together tech, art and culture. On Sept. 18+19 in London you'll dive into the dynamic spaces shaping society with new ideas and insights to help you navigate the uncertainty around AI, Web3 and immersive tech.
Guest Post by Genuine Impact
?? Genuine Impact is a rather large Newsletter with some incredible macro analysis.
Miguel Benitez writes for Genuine Impact, a 3x weekly newsletter using data visualizations to explore and dissect the financial markets, economic climate and other complex data around us. You can subscribe to the Genuine Impact newsletter here.?
If you are passionate to learn and and enjoy explanations around investing with beautiful and simple visualizations, love charts, stock deep-dives, and investment philosophies - then this might be the right Newsletter for you.
Let’s dive right into it:
Where can AI go from here?
At first glance, that might sound like a silly question, and I get it – we’ve barely scratched the surface of AI’s capabilities. You can automate pretty much everything now, from driving to cooking to medical procedures; anything you can think of there’s probably an AI tool for it. It’s definitely an exciting time for the AI industry, and an exciting time for investors as they hope to discover the next trillion-dollar company.
Currently though, there are AI companies that look like they’re running before they can walk.
Every week, we hear of a new disruptive AI startup that will “change the world”. It has become a buzzword that almost guarantees millions of dollars in funding from the biggest names in venture capital. These startups might even get bought out by one of the Big Five as they try to stay ahead of the game.?
Endless funding, new companies appearing out of nowhere, expectations of changing the world – exciting for sure, but comparisons have already been made to something not-very-nice: the Dotcom bubble.
The candle that burns twice as bright, burns half as long???
Now of course, I don’t doubt the usefulness of these AI tools – but when your side project gets a billion-dollar investment overnight, it doesn’t automatically make you a great business owner. The main thing on the mind of an investor is how much money can this business make.
Over the past couple of years, private and public sector funding into generative AI has skyrocketed. Just like in the Dotcom bubble, we’re seeing two distinct sides: the old guard of established multibillion-dollar companies like Google and Microsoft, versus the new kids on the block – startups.?
The Big Five
Amazon, Apple, Google, Meta, Microsoft. The tech giants of today, with their considerable spending power, have vacuumed up around 100 AI startups in total since 2011. These include the high-profile acquisitions of DeepMind by Google for $650M in 2014, and Shazam Entertainment by Apple for $400M in 2017. Apple currently stands out above the rest with 30 acquisitions so far, closely followed by Google at 26. The remaining three are some way behind - Amazon with 14, Meta with 17 and Microsoft with 16 acquisitions to date.?
Microsoft’s $10B investment into OpenAI in January 2023 represented just how much ChatGPT meant to them, giving them a 49% stake in the AI startup now valued at $29B. The next few years, I’m sure, will see the rest of the Big Five and even other huge tech companies like Samsung and Oracle spend hundreds of billions on companies like OpenAI in order to integrate artificial intelligence into their products.
In terms of chatbots, Apple’s Siri was probably the most successful before ChatGPT came along. Microsoft now has GPT-4 integrated within Bing Search, while Google Bard is based on their own PaLM 2 model. Amazon’s Lex is used in their Alexa virtual assistants. Meta is the only one of the Big Five that currently doesn’t have a chatbot of their own, but they did release their LLaMA model as an open-source package to “further democratize access” to AI research.?
Startups
Hundreds of startups out there are developing their own versions of ChatGPT to be used in a multitude of different situations. Startups like Anthropic or Perplexity AI are making similar chatbots, and now we’re even seeing prompt engineering startups raise millions of dollars by just…helping you write better prompts.
Understandably, everyone wants in on it. The amount of money raised by startups in the AI industry is mind-boggling: OpenAI got $10B from Microsoft, Anthropic raised $580M, the list goes on. A study from CBInsights found that of the 13 generative AI startups that reached a $1B+ valuation, the average time to reach unicorn status was only 3.6 years - about half the average time of all unicorns (7 years).
Additionally, any large tech-focused venture capital (VC) fund you can think of has almost definitely invested in an AI startup or two within the last few years. This includes Andreessen Horowitz ($30M for People.ai), Sequoia Capital ($5M for Dust, a startup that hadn’t even been incorporated yet), and Bain Capital ($215M for HeartFlow).?
To read the original article with the best graphics go here.
When the bubble pops
In recent years, the media has portrayed tech startups as a more exciting place to invest your money compared to the slow growth of companies like Coca Cola. And rightly so - the late 2010s and even the pandemic years brought a feeling of “We’re living in the future now” as startup after startup obtained unicorn status. However, the bubble eventually popped - pandemic darlings like Peloton absolutely tanked during 2022, and mass layoffs are still, unfortunately, occurring regularly.
The AI industry of today has two different types of company. On one side is Big Tech - ChatGPT backed by Microsoft, Google Bard, Apple’s Siri. On the other side are the startups like Jasper, Perplexity and Poe. Plucky underdogs versus the big boys, with seemingly enough space for both sides to thrive. Who doesn’t love that storyline?
The reality is, most startups fail. According to the US Bureau of Labor Statistics, 20% of startups fail within the first year. After five years this goes up to 48%, and after 10 years, 65% of startups will have failed. Digging deeper, the information industry - which most AI startups belong to - has the highest failure rates. 26.4% chance of failure in one year, 51.8% chance within five years, and 72.1% chance within 10 years.?
Venture capital funds and angel investors aren’t just looking for cool new tech that might change the world. They’re interested in the business as a whole - is it revenue-generating, or is it just an idea that needs money to build? Is the team made up of experts in their field with no business knowledge, or corporate executives who would need to hire a whole lot of people? So many important questions need to be asked before even thinking of transferring money, and so many startups might not have the right answers to those questions.?
Here’s another problem for you to ponder: a LOT of startup chatbots use OpenAI’s GPT-3, 3.5 or 4 models. And yes, they’ve all put their own spins on it, and everyone’s wowed by all of their capabilities, but is that a sustainable model to base a whole business on? A single tool that uses the proprietary model from another company will get you a cool $20M in funding today, but what happens in 3 years when you’ve spent all the money on hiring 100 people and not developing anything new? There’s a reason why the Big Five are still growing so consistently even decades after their creation.?
Innovation.
Innovation breeds success
ChatGPT was, and still is, a resounding success for the AI world. It only took the chatbot 5 days to reach 1 million users after its public release in November 2022, and OpenAI got a sweet $10B check from Microsoft for their efforts – job done, right? Of course not. In the months following ChatGPT’s explosive start, OpenAI added a premium subscription with plugin support, API keys, and improved language models with every new release. It seemed like the world was theirs for the taking.
Fast forward to June 2023, and monthly website traffic and unique visitors declined for the first time ever. 7 months. That’s all it took. I’m not saying OpenAI is going to fail tomorrow, or even at all – they have plenty of funding and support from one of the biggest companies in the world, and one that knows how to innovate.
Innovation is the key word here, and the Dotcom bubble gave us plenty of examples. Companies that don’t innovate will die out, whether startup or established Fortune 500 – remember Pets.com? Or Webvan? Market leaders in their prime, now bankrupt or absorbed by rivals. They didn’t innovate, couldn’t become profitable, and didn’t have sustainable business models.
Regulations are another important note that should also be addressed. Is it too early to put shackles on an emerging technology that is changing every day? Some people say it is, and to see what sort of capabilities we can achieve first. But we’ve already had plenty of examples of people ‘hacking’ chatbots, with ChatGPT able to provide comprehensive instructions on how to build a dirty bomb and hotwire a car. The general mindset in the AI community seems to be supportive of regulations, but there are worries as to whether the ones currently being discussed are much too stringent - remember when Italy outright banned ChatGPT in April over privacy concerns?
Okay, enough doom and gloom. The bubble might pop tomorrow, or not pop at all. And the likelihood of whether Google, Meta, or any AI startup fails is not something we can predict reliably. Just think - generative AI is what everyone is talking about today - imagine what we’ll have in another 5, 10, 20 years. Will OpenAI stay at the top of the AI charge for the next 20+ years and grow into a trillion dollar company? Possibly. Will some of the unicorn startups today ultimately fail? Statistically speaking, yes.
??? Read Past Issues | ?? Start a Substack |???Sponsor Us |?? About
?? Enjoying this content? Subscribe to their Newsletter. ??
?? Also by the Genuine Impact team ??:
How to bet on AI…safely
Now obviously you can invest in the Big Five and any other big tech company that is using AI in their business. You can also bet on riskier startups, which could skyrocket in value and give you those sweet, sweet dividends. A third option, perhaps less risky but with a similarly high upside, is to look one level deeper. What sort of infrastructure do all these AI and other tech companies need today and in the future? Whether it’s Apple or some guy who made a chatbot as a college side project, there are several key things that all AI companies need to survive, and indeed thrive. Let’s take a quick look at what they are:
Data storage and cloud services
Artificial intelligence hasn’t quite yet progressed to the point of being able to ‘think’ creatively and come up with new ideas. As such, companies need to train their models using millions of data points. OpenAI reportedly used 100 trillion parameters to train ChatGPT-4, and you can imagine that a 1TB hard drive probably won’t do the trick to store all that data. Startups and smaller companies generally can’t afford to build the huge data centers needed to store all their data, and cloud services companies do all this for them.
Some of the Big Five companies already have their own cloud infrastructure like Amazon’s AWS or Microsoft’s Azure, and they could stand to profit heavily - but there are other opportunities if big tech isn’t your type.
Snowflake ($SNOW), a pandemic darling that bombed pretty spectacularly at the start of 2022, is now trading at a more reasonable price. The company has beaten EPS in the last four quarters, and has almost doubled revenue every single year. With the spotlight no longer on pandemic tech, the onus is now on them to ‘grow up’ from a young startup, making Snowflake an interesting investment for the long term.
DigitalOcean ($DOCN) is another pandemic stock that provides cloud services for small businesses and startups. Yep - the next OpenAI might be using DigitalOcean. The company is (slowly) on its way to profitability, and the stock is up 86% YTD. This is a small-to-mid cap stock whose growth could stay under the radar for a while yet.
If individual stocks aren’t your style, two cloud ETFs are worth looking at. The First Trust Cloud Computing ETF ($SKYY) has 64 cloud stocks with about $2.9B assets under management, and is the biggest cloud ETF currently out there. Otherwise, the Global X Cloud Computing ETF ($CLOU) with 34 stocks has slightly outperformed First Trust since its inception in 2019 if you want something more concentrated.
Cybersecurity
Data and algorithms wouldn’t be as valuable if everyone could get their hands on it. According to Morgan Stanley analyst Hamza Fodderwala, "Ransomware attacks are up 30% year-over-year and email phishing campaigns have increased significantly since the launch of ChatGPT." Cybersecurity companies could therefore have huge growth opportunities come their way in the near future as companies and individuals look to secure their data.
Palo Alto Networks ($PANW) stock has been steadily on the rise, up 81% YTD. The company recently broke into the S&P 500 in the last quarterly rebalancing, and beat their last 4 quarterly earnings.?
Crowdstrike ($CRWD) uses AI and machine learning to level up their protection services, linking all connected endpoints to a single threat graph that gets updated every time a new threat is detected. Basically, it trains itself all the time - and new connections (i.e. new devices) means more data to train on, which means it gets better every time a new device is connected. Crowdstrike also beat their last 4 quarterly earnings, is up 45% YTD and is trading at a more reasonable price after a steady decline in 2022 brought the stock down -47% from its all-time high.?
GPUs
As datasets and processing complexity grows, the need for GPUs (Graphic Processing Unit) also grows. When developing a large language model like ChatGPT, the most resource-intensive task is generally the training phase. Imagine manually sorting through a million different parameters - it would take years. Even your average laptop without a dedicated GPU could take years with a neural network of over 10B parameters. GPUs essentially allow computers to process a lot more operations, a lot faster.?
NVIDIA ($NVDA) is the GPU company on everyone’s minds right now. It seems like the stock?hits a new all-time high every day, bringing the chip maker’s market cap to over $1T. NVIDIA’s GPUs are used not only by AI companies to train their neural networks, but also by cloud service companies to help process data faster. The company is also looking to invest in chip maker Arm upon their IPO, further solidifying their grasp on the chip market. Two birds with one stone?
Semiconductors
Exploring GPU production even deeper, TSMC ($TWD) is currently the world’s largest semiconductor manufacturer. This company quietly supplies chips to hundreds of companies in every industry: iPhones, F-35 fighter jets, even NVIDIA’s own GPUs all use TSMC chips. In a world that is still dealing with a semiconductor shortage, TSMC is also arguably one of the world’s most important companies at the moment: no TSMC = no iPhone 15. The stock is up 28% YTD after recovering from a shaky 2022.
So where does the future of AI lie?
Perhaps you’ll finish reading this and come out with more questions than when you started. That’s not a bad thing, especially when investing money. If you’re trying to make a profit from this new and exciting industry, asking questions is important.?
It’s also an industry that will have huge ramifications in plenty of other industries. AI-assisted agriculture could improve crop yields and make farming more sustainable, while personalised learning with the help of AI could transform education and help people learn new languages and skills in a much shorter timeframe than simply following a course.?
Just remember, nothing that you’ve read should be taken as financial advice. I’m not a certified financial advisor, nor do I have a PhD in machine learning.?
What do you think about the future of AI? Do you invest in these companies already? Are there any I missed?
Thanks to Miguel Benitez for the insightful post. Their charts are amazing. They also share their charts on LinkedIn here. I hope you have found some value from these perspectives.
??Also by the Genuine Impact team??:
Thanks for reading!?
See you next time!
What do you think is after OpenAI?
More features, AIs integrated into the OS (from Linux to MacOs), AIs integrated and free in software factories, word processors. from cheap to $10 per year we will begin to review a price battle for the profits of the powers users (the general public is out of the race)
"What will GPT-2030 look like?": https://www.lesswrong.com/posts/WZXqNYbJhtidjRXSi/what-will-gpt-2030-look-like :-)
Back on the road again
1 年Open AI Marriage
AVP FinTech Product Manager DEI Driver Cannabis advocate
1 年Dude there are so many startups that learning it seems more overwhelming But Shirlfriend is here for it Let’s go!!