There’s an interesting thing happening in the generative AI world.
More and more, established companies are leaning into generative AI by either adding generative AI features to their existing suite of tools or partnering with AI companies (mostly OpenAI or the open source HuggingFace) to launch entirely new products. Their deep pockets, brand recognition and existing infrastructure offer several advantages.
Examples includes companies like Zoom with ZoomIQ, Bloomberg with BloombergGPT, and of course Google and Microsoft with their slew of AI-powered enterprise and consumer products. The list goes on. On the other hand, check out ProductHunt or a few AI newsletter and you’ll quickly hear about more than 1,000 generative AI startups, with new ones popping up daily.
Don’t get me wrong, AI is no silver bullet. Chegg, which lost over 40% of its stock value after the CEO remarked that the company’s tutoring product is being challenged by ChatGPT, has been working hard to incorporate generative AI features into their own product with GPT-4 in the form of CheggMate, but so far with little results.
There’s a big question being asked by investors in generative AI right now: who will reap the most benefit from this innovation. Is it startups or incumbents?
According to CBInsights, there are already 13 generative AI unicorns, but despite the peak hype and constant media attention, it’s hard for new startups to stand out in this space and VCs are finding it difficult to place their bets. As I mentioned in my post on the LLM Benchmarking, venture capital investment in AI startups was down 43% in Q1 2023.
My hypothesis on why AI investments are down significantly in Q1 2023:
First signs of trouble in the horizon
We’re starting to see the first signs of generative AI upstarts folding or pivoting. For example, Neeva, a startup that tried to push the boundaries of web search and challenge Google by being the first ‘AI-powered’ search engine, found the hard way that’s one thing building a product and quite another to get users to change their habits and switch search engines. The company is shutting down its consumer product in early June.
Neeva is a reminder that for a startup to be successful in the generative AI space, they need not only to nail the product, but also the distribution (a key advantage of the incumbents) and their unit economics – a tougher job in a constrained capital environment where rounds have gotten smaller and there’s a high cost and shortage/cost of GPUs.
The big opportunities are still out there for the taking
If you listen to what the top CEOs and thought leaders are saying about AI, the opportunity is huge. For example, take the latest remarks by Sundar Pichai, Alphabet’s CEO:
“Well, definitely I see it as an extraordinary platform shift. Pretty much, it’ll touch everything: every sector, every industry, every aspect of our lives. So one way to think about it is no different from how we have thought about maybe the personal computing shift, the internet shift, the mobile shift. So along that dimension, I think it’s a big shift.”
Sam Altman and others say that the foundational model race (and funding opportunity) is largely over and that the competition now is on the applicational layer. I tend to agree, with the exception of open source LLM APIs which are taking off massively and not following much behind in quality from Google’s Palm-2 or OpenAI’s GPT-4. This Open LLM Leaderboard on Hugging Face is a good example of the developer excitement and engagement in this space. In the future, there will be less platform dependency (as not all startups will be built on 1-2 APIs only in the application layer).
I’m particularly excited about the following opportunities at this point in time:
- Vertical use cases for generative AI that is embedded in the team’s normal workflow (like what Jasper.ai did for marketers)
- Developer tools and enterprise tools to safely and creatively incorporate generative AI in the organisation
- Consumer applications that offer reliable automation and increased productivity
- “Painkillers” vs. Vitamins solutions tailor made with the user in mind
- Use cases in gaming, virtual worlds, education, health, climate… the market is big
It’s hard to predict whether the big winners from generative AI will be the startups or the incumbents, and despite the perception of a crowded market, I think we’re still early in this space. If Generative AI is truly a platform shift like the early days of mobile and cloud, think about the first generation of apps or cloud services. They were much less impressive than what you have now, and also there weren’t that many new apps in the very early days. But the number of apps kept multiplying (there are 500-600 new games introduced to the app store every day vs. a few dozen in 2008 or so when the iPhone came out). And many of those games that didn’t exist in the early days of the iPhone are billion dollar companies.
What else do we know? Models will get smaller over time, the computation will move from the cloud to on-device, and costs for incorporating generative AI in both new and existing products will keep dropping.
I’d love to connect with more generative AI founders that are tackling the big opportunities in this space.