Ruminations #13: is AI a platform shift? How is Australia shaping up?
Reflections from me and the Square Peg team
Hi!
In today’s edition:
Recent shares and reflections from the Square Peg team
My AI-focused commentary for the recent State of Australian Startup Funding Report
Team reflections
Last week Square Peg Co-Founder and Partner Paul Bassat shared the following in our Team Learnings channel on Slack:
A post by a VC got me thinking over the weekend about what sorts of startups will prevail in the AI era vs previous tech eras.
If you think about the internet era, there was a mix of successful startups that did things that couldn’t be done before but there were more startups that did things better than traditional incunmbents and the change in platform meant that incumbents couldn’t compete.
Google was an example of a startup that did things that effectively couldn’t be done before (search for all of the information that existed) but Amazon was an example of a startup that did things better (buy a book or a jumper).
At SEEK, we improved the process of people finding a job (or employers finding staff) and at REA, they helped people buy a house. Expedia helped people book a trip. I would say they sat somewhere between helping people do stuff better that they had done before (booking a holiday) and creating something new (being able to book a trip without a travel agent).
The common denominator was there was a shift of platform that created new opportunities both for folks inventing stuff but even more for people who weren’t inventing stuff but were radically improving stuff. If I think about the move to SaaS, that has primarily been about new and incumbent businesses that helped people do stuff better than they were doing things previously in a traditional software world.
As I think about AI, my sense is that incumbents by and large are going to be incredibly well placed to help consumers or businesses do stuff better as long as they have the mindset and talent to execute. There is no platform shift in the way that we saw a fundamental shift of platforms with the internet.
What does this mean for startups?
I think AI (unlike the internet) is going to enable much, much more novel activity that wasnt possible to the same extent with the rise of the internet, and novel applications will be where we see a lot of value created by startups.
A framework for me in thinking about any AI startup is along the following lines:
1. Are they doing something truly new?
2. If not, why are they well placed to beat incumbents?
3. If yes, will there be a market for what they are doing within the timeframe required?
4. If there will be a market, why are they best placed to win that market?
To which my colleague Lucy Tan replied:
To build on the above, Zetta Ventures just put out the above in their email newsletter which really resonates:
"AI differs from other platform shifts of the modern era in at least one fundamental way: unlike Internet, SaaS, Cloud, or Mobile, AI is not disruptive to the technology stack itself. And that means the best opportunities for startups may look a little different than they have before.
Compare AI to SaaS: in 2002, just cloning an existing software product with SaaS resulted in a vastly superior product. You shifted capex to opex spending, improved product development velocity and shifted the operational burden from the customer to the vendor. The SaaS stack was disruptive and gave startup challengers a huge opportunity to disrupt incumbents wedded to legacy code bases.
On the other hand, if you took a successful SaaS product and copied it with an AI model, the product would instantly become much worse: more expensive to build and operate, inconsistent and unreliable, and inevitably the quality will degrade without expensive ongoing retraining. Of course you can add AI features to existing software and make it better, but ultimately that’s a feature, not a product. Most incumbents have the “right” to that roadmap. Some sleepy incumbents may miss their window, but counting on someone else to screw up is no way to build a business.
...AI is disruptive because it solves problems that software can’t. It can automate work that has resisted technology and until now could only be accomplished (if at all) by people doing tasks by hand. If AI disrupts anything, it’s not technology stacks, it’s human labor.
...So the real AI Native startup opportunity is to create new products where there are no tech companies today, where the business problem is solved with human labor (or just not solved). If incumbents exist, they are service firms. Those may turn out to be your customers (like law firms for Harvey and Casetext) or they may turn out to be the incumbents you disrupt (like taxi companies for Uber and Lyft, or translation houses for Lilt)."
The State of Australian Startup Funding - AI
I was asked to provide commentary for the 2023 State of Australian Startup Funding Report. It’s an excellent report that I recommend founders skim, as it provides some very helpful stats on the ecosystem.
Read on below!
It’s clear that 2023 was an extraordinary year for technology, in large part thanks to what we’ve seen coming from AI startups and research. Platforms like ChatGPT, DALL-E and Midjourney captured the imagination of consumers, developers and investors alike. Australia, thanks to its world-class developer talent, has a real shot at producing generational AI businesses. I personally have never felt more excited to be a technologist.
An overnight success, decades in the making
Over time AI has seen hype cycle after hype cycle accompanied by expectations that were often greater than what the technology could deliver.
In 2023 (and late 2022) ChatGPT became a global phenomenon, in part because it felt as though we were finally seeing some of AI’s true potential come to life.
This inflection point is not merely another hype cycle. It represents the culmination of decades of research, data proliferation (thanks to the open internet) and growth in computing power. These three factors have been severe limitations for the progression of AI’s capabilities.
It’s not just generative AI that’s seeing breakthroughs, either. Truly autonomous vehicles now roam the streets of San Francisco and are able to be booked as easily as one books an Uber. AI-enabled drones deliver food in Canberra and Brisbane. We are living in the future we were promised.
The Australian ecosystem
There’s much disgruntlement about Australia’s missed potential as an AI powerhouse among experts, especially since the disbanding of the early NICTA group and the overseas departure of some of Australia’s most credentialled AI researchers. Fortunately, Australian developers are well-positioned and excited to make up for lost progress.
The barriers to building with AI have been rapidly diminishing and the skillset required to build an excellent AI-enabled product have changed. Speaking to researchers at OpenAI, they find that team members who join with software engineering skills are better equipped to learn AI than researchers who try to learn software engineering skills.
Australia has world-class software engineers, many of whom are currently rapidly upskilling in AI.
Beyond startups, enterprises are taking to building AI solutions with fiendish excitement. Whilst many enterprises have started to understand the difficulty of productionising AI models, an ecosystem of support players (consultancies, developer studios and cloud hyperscalers like Google and Amazon) have been investing significantly to support them. An increased level of comfort and familiarity with AI from enterprises will be a huge revenue driver for the startup sector, as enterprises have both the deep pockets and excitement to implement AI.
What to expect in 2024
Investors this year were simultaneously extremely excited by AI and very wary about investing in the wave of AI companies being founded. Many continue to feel a degree of anxiety about evaluating the longevity of these businesses in a space that’s moving faster than any other we’ve seen.
In the year to come, I expect to see more investors find comfort in investing in these new products, but with an increase in scepticism due to:
1) the number of businesses founded to solve very similar problems; and
2) the number of fast failures or quick pivots being seen in these businesses that are struggling to find sound footing amidst the change.
There may also be some cooling of the sky-high valuations seen for AI companies as scepticism rises and investors become more conscious of the difficulty of raising new venture funds in this macroeconomic environment.
Increasingly AI-savvy investors are talking about their interest in AI businesses that combine the technology with deep domain expertise to solve a specific problem that’s difficult for larger players to fast follow. We’ll also see more cross-modality and multi modal applications and more developers moving away from building purely on OpenAI’s architecture to more tailored, open-source alternatives.
As always, thanks for reading and feedback welcome!
Thanks
Casey