Follow the AI hype | #2

Friends,

Make sure you scroll all the way down. I don’t want you to miss reading David Gorgan’s view on the topic of AI hype. Moreover, there’s a link to a – I promise – quite entertaining, earworm-quality song.

Now let’s dive in.

I recently came across a “Fundraising Cheat Sheet” of an early-stage investor. It stated that if you “Level up by following the hype”, e.g. generative AI in climate tech you can not only raise significantly more but you will also have an easier time fundraising.

Fundraising is very hard. So if it’s easier to raise in 2023 as an AI-Startup should I not rethink my strategy, product roadmap, domain choice and hiring plans? Is that the missing piece for finally getting this sleep-stealing 1 million EUR pre-seed round over the line? You’ve thought anyway that this one “ML-enabled feature” was not too far off from the current plans. Why not anticipate it in the product roadmap?

It can work to raise a round on the shoulders of hype. However, you might sacrifice long-term success for short-term gains. Why?

If you’re optimising for what’s cool, your fundraising will suffer from an adverse selection of investors. You don’t want to raise from investors whose thinking is influenced by hype. Instead, optimise for angels and VCs whose judgement stands on solid grounds. Hypes are temporary, investors on your cap table aren’t. So be careful whom you take on this journey. Play long-term games with long-term people.

Furthermore, the fundraising part itself is only a means to an end and not the end itself. You should obsess about the end and not about the means. Your company exists because you make something customers actually want, you can actually build and you can actually capture a significant chunk of the created value long-term. If you’ve successfully tested your hypotheses that a fine-tuned co-pilot for lawyers makes sense, then build it. But don’t build it because a16z published an article about the proliferation of LLM-based chatbots. At best you’ve wasted months, at worst you’ve killed your company.

Hype ≠ growth. The first is a subjective noise. The second is a development you can measure because it has happened in the past or you forecast based on a foundation of solid reasoning and your (non-obvious) insights. In the context of AI: No doubt that recent breakthroughs in machine learning can play a substantial role in your company’s Why Now – one of the key questions for which you need to have a compelling answer as a founder.

Apple and Bytedance (TikTok), two of the most successful firms leveraging AI barely talk about machine learning. Cool people, don’t talk about being cool. Yes, these aren’t startups. Yet, I think the point remains the same.

Founder’s view: David from Nostos Genomics

I’m beyond excited to have David Gorgan join me for this piece. He has been deeply involved in the AI space since his days at Merantix in 2016. Nowadays he’s building Nostos Genomics together with his co-founder Rocío Acuña Hidalgo (CTO). It was probably not too surprising that he ended up founding a company after having initiated Europe’s largest entrepreneurial hackathon START HACK.

What has been your experience with hype cycles around AI?

In 2016, we went with the whole team of Merantix, a Berlin-based AI investment platform – which consisted of 5 people back then – to the 30th annual NeurIPS conference held in Barcelona that year. This is one of the most significant machine learning conferences, and many scientists actively shaping AI in industry were present. To name a few, it included Yann LeCun (Chief AI Scientist at Meta), Andrej Karpathy (previously Director of AI at Tesla, now working at OpenAI), and Ian Goodfellow (until recently Director of ML at Apple, now at Google DeepMind).

My impressions from the talks and discussions with people I met were overwhelming optimism and excitement about the real scientific progress. But I definitely also got a sense of hype. There was a particularly hilarious example of that during one of the conference evenings. In general, it was pretty common for new AI labs to be revealed by tech companies during the conference (for example, Uber AI that year), so when we were invited to a launch party for a reveal of a new lab, it sounded like this could be another very exciting one. Others clearly thought the same, because 200 people showed up at the venue. It included primarily scientists and engineers, but also journalists and investors. Five major VC funds even emailed about investing in the yet-to-be-revealed company “Rocket AI” ahead of the event. But as we arrived and the unveiling proceeded, it turned out that the whole thing was just a prank. The “Temporally Recurrent Optimal Learning” acronym should have been rather obvious in hindsight. Riva-Melissa Tez, the organizer behind it, wrote a great article about the whole thing that I can recommend reading: Link

Her article highlights that “AI is at peak hype, and everyone in the community knows it” – and this was already back in 2016. Looking back, I have the impression that the previous AI investment hype cycle only plateaued around the end of 2019, with the focus shifting towards other areas. Now, we entered yet another one since the massive achievements of foundation models / GenAI last year. For anyone investing in this new cycle, it seems highly prudent to get acquainted with the lessons from the previous one, because certain developments clearly didn’t pan out in the way they were expected (both on the positive and the negative side).

What’s the best and the worst thing about a topic/technology being hyped?

Simplifying heavily, a positive aspect of hype is that if it is tied to real technological progress and commercial applications, increased funding leads to more shots on goal, which increases the odds of at least one of the funded projects or companies ultimately succeeding. On the flip side, it can also lead to overinflated expectations about a technology’s capabilities and timeline for maturity. If those aren’t met, it can lead to significant disillusionment and funding droughts, as occurred in the past (see AI winter).

What’s one piece of advice you wish you had heard earlier?

Daniel Ek, the founder of Spotify, mentioned in a podcast around the time when I co-founded Nostos that “the value of a company is the sum of all problems solved”. It’s very easy to get bogged down by all the small and big issues in the typical rollercoaster journey of running a startup and get a sense of being stuck, especially early on. For many years, this advice has been a good reminder for me that instead of seeing problems as a negative occurrence, every one of them is an opportunity. As long as we tackle them swiftly, consistently, and successfully with a roll-up-your-sleeve attitude, we can build value every day – for our customers, patients, team, and investors.

Three recommended content pieces

• Early Days of AI by Elad Gil

• Secrets by Peter Thiel

• The current climate in AI has parallels to web3 2021 by Francois Chollet

Worried about missing out on the coolest tech & cultural events in Berlin & Munich? You should check out the FOMO newsletter of my good friend Cephas.