A picture of the Pope in a fabulous oversized white puffer jacket recently went viral. His Holiness looked like he’d stepped off a Paris catwalk. It was, of course, a fake. The image was created using Midjourney, a generative AI engine which turns text instructions into pictures.
Midjourney. Dall-E. ChatGPT. These generative engines are a phenomenon. ChatGPT won 100 million users in its first two months. And the money-making potential is staggering. OpenAI, creator of ChatGPT, raised $300m (£240m) in early May at a $29bn valuation. Google invested $300m in generative AI startup Anthropic, and $100m in AlphaSense, a provider of financial information, to add an ability to summarise information with generative AI. US tech giant Salesforce is pouring $250m into generative AI startups including Cohere, Hearth.AI, and You.com.
So with all the big players pouring money into the industry, where does this leave smaller investors? Is there an opportunity to participate in the greatest tech boom of the decade?
“Plenty,” says Professor Andy Pardoe, chair of the Deep Tech Innovation Centre at Warwick Innovation District and founder of Wisdom Works, an AI platform. “It’s a fantastic period of innovation, and smaller companies can typically move faster than the bigger ones.”
In his view, there are so many AI startups looking for capital that the chief problem for the investor is not finding a way into the market, but rather knowing where to put their cash. “It’s challenging,” says Pardoe. “But that’s why investors typically have a portfolio of businesses they invest in. They hedge their bets with multiple investments.”
The data supports this view. According to Pitchbook, which tracks tech deals, in Q1 of this year a total of $1.7bn was invested in generative AI. Some are primary providers of AI, and others harness an engine in a specific way or offer a user interface. It’s a thriving ecosystem.
The challenge of picking the winners
Some of the companies at the cutting edge are small. Midjourney, for example, is a bootstrapped business with just 11 full-time staff, but it was able to outperform the generative AI teams from Google and Adobe. Rival company Stability AI, which produces the image venture capitalists in October, is still a startup by any definition. Small enterprises can indeed succeed.
Google, for one, is under no illusions about the threat from smaller rivals. A leaked note by an anonymous but verified Google researcher published in May declared bluntly: “We have no moat. And neither does OpenAI.” The researcher went on to state: “We’ve done a lot of looking over our shoulders at OpenAI. Who will cross the next milestone? What will the next move be? But the uncomfortable truth is, we aren’t positioned to win this arms race and neither is OpenAI. While we’ve been squabbling, a third faction has been quietly eating our lunch. I’m talking, of course, about open source. Plainly put, they are lapping us.”
Open-source models are free to use. They can be run on phones or PCs and fine-tuned by the user. WizardVicunaLM, for example, is an open-source rival to ChatGPT which can be run on a laptop. The Google researcher noted: “People will not pay for a restricted model when free, unrestricted alternatives are comparable in quality.” If open-source AI triumphs, the billions invested in closed-source companies will go up in smoke.
And for the open-source alternatives, the quality is there. One leading authority on generative AI estimates it works at 97% of the performance of ChatGPT-3.5. “In the space of six weeks we have moved to the Cambrian explosion,” said the expert.
Balancing risks and rewards
The opportunities may be big, but so are the chances of backing a dud, warns Sophie Lonergan, head of startup investment at Digital Catapult, which promotes new tech adoption and investment in the UK. “With early-stage companies, there’s a risk. The returns are not guaranteed and it isn’t easy to profile the investment in the way you might with other asset classes.”
For the undeterred, there are many ways to invest. Lonergan points to the angel investing groups and online fundraisers such as Crowdcube that offer a way in for the smaller investor. The tax benefits are advantageous too. The Seed Enterprise Investment Scheme and Enterprise Investment Scheme offer generous tax relief to qualifying projects. Lonergan recommends studying the educational literature and masterclasses created by angel networks and other investment advisers.
And before rushing to invest in this booming sector, it may also pay to listen to lawyers. Generative AI is new and there are potential pitfalls. Emma Wright, partner at law firm Harbottle & Lewis, warns: “AI is seen as a high-risk sector by the UK government for the purposes of the National Security and Investment Act, which gives the UK government the ability to block or apply conditions to certain transactions, including investment in AI startups at certain levels.”
It isn’t yet clear who owns AI, or whether the models which build on existing work are permissible. Wright notes: “The position in relation to whether a third party’s intellectual property has been infringed, or personal data exploited, when building these LLMs [large language models], and then who owns the output, remains unclear. It hasn’t been properly tested yet and could impact valuations either way depending on the outcome.”
Furthermore, a sector this new is volatile. Disruptors can be disrupted. Google pioneered image generation, to be overtaken by NVIDIA’s StyleGan, which itself was overtaken by Dall-E and Midjourney, and in September Meta announced Make-A-Video, a text-to-video model. Amazon and Baidu are investing in generative AI, and have the budget to back their ambition. The leader today could be an ‘also-ran’ by next year.
For the victors, the riches are there to be won. ChatGPT charges $20 a month for its premium service, and forecasts revenue of $1bn in 2024.
The opportunity to invest is clear. Public appetite for the products is sky-high. But investors need to ask: in this industry of illusions, how can you be confident that you’re backing a winner and not a mirage?