10 Predictions for 2023: AI, Web3, and Beyond
[The piece was originally posted here]
What a year for 2022! As we enter the new year, we’d like to take the time to reflect on some major events that happened in 2022 and try to look ahead to what might be coming in 2023.
As an early-stage fund, Leonis focuses on companies that are powered by AI and web3 protocols. We spend a lot of time thinking about the changed and unchanged aspects of these two paradigm-defining technologies, their intersection, and the implications for society over the upcoming decades.
If you are a founder pursuing any of the areas in our AI and web3 predictions – or any other non-consensus ideas related to these supercycle technologies – we would love to hear from you! Shoot us a message at jay AT leoniscap.com or jenny At leoniscap.com.
Looking into 2023, here are 10 predictions we made about the new year:
AI: Still Looking Hot
1. More VC funding will go into AI/AIGC-related companies, possibly creating a new bubble for the next cycle.
With the launch of eye-catching models like ChatGPT, DALLE-2, and Stable Diffusion, AIGC has gone mainstream and has caught the attention of the general public, as well as many VCs who are looking for the “next hot thing” to invest in. This has led to increased investor confidence and a surge of venture capital going to AI and AIGC companies, which has in turn caused the market to become overvalued and vulnerable to a potential bubble. Investors should expect a high noise-to-signal ratio in AI in 2023. On the flip side, we will see lots of experimentation and many startups being built in 2023, setting the stage for consolidation in 2024. ****
2. AIGC will go beyond consumers’ “fun” use cases and reach enterprise services.
The 2022 consumer-facing AIGC products (like Lensa.ai) generate real revenue but are buzz-driven. The buzz might fade in 2023 and it remains to be seen whether a once-hyped product can transform into a venture-scaled company. At the same time, AIGC will be increasingly used for enterprise functions, such as internal project coordination and workflow optimization. It is already used in external-facing activities like sales and marketing (Jasper.ai and Copy.ai are great examples). But it would be interesting to see how the technology can be used to reduce internal inefficiencies, particularly in industries like healthcare and manufacturing.
3. AIGC models will get significantly better but data will be a key bottleneck.
GPT-4 will probably be released in early 2023 and it will be a big game-changer. The technology will enable a whole new generation of AIGC startups. As we argued in our generative AI blog post, the underlying models determine what applications can be built. As DeepMind’s Chinchilla paper demonstrates, the best way to train powerful models is to feed them more data rather than increase their parameters. Models like GPT-4 are unlikely to be much bigger than the largest current-gen models like Microsoft and Nvidia’s Megatron-Turing NLG (570 billion parameters), but they will require training datasets orders of magnitude larger than those of GPT-3. Acquiring training data will be a challenge for AI labs building state-of-the-art models.
4. 2023 might mark the year for “AI-first companies,” setting the new benchmark for software value propositions and sales.
As we mentioned in a 2020 post on AI and the new industrial revolution, AI-first companies behave differently than their predecessors. They are more value/ROI-driven. As AI moves from predictive AI (goal is to label data points) to generative AI (goal is to output data points), the ROI for generative AI companies is even greater than previous-generation predictive AI companies!
Because AI-first companies are a “different species,” they will take over incumbents’ market share in surprising ways. For example, what replaces Google might not be another search engine, but an AI chatbot and content producer like ChatGPT that offers more targeted, concise, time-saving, and possibly ad-free search results. What replaces Salesforce is not likely a static CRM database but could be AI-first software that can adapt and provide actionable information as the sales process unfolds. In 2023, we will see some early experimentation in using enterprise SaaS sales approaches and pricing models to evaluate AI-first companies (perhaps the focus is less on revenue per seat and more on value creation).
Web3: Rebuilding Better
5. For the web3 world, 2023 will continue to be a bearish year for tourist investors but a fantastic year for agile builders.
Despite a series of crypto project crashes in 2022, the fundamentals of web3 have not changed. At Leonis Capital, we focus more on the fundamentals of decentralized ledger technology and its profound implications on how data can be stored and verified in a trustless, decentralized way. While Celsius, 3AC, and FTX dominate headlines, we have also seen a major merge by Ethereum this year, paving the road for scaling in 2023 and beyond. Although there will be less hype and more pessimism in the web3 world, under the hood, the web3 ecosystem will continue to be buzzing with builders’ activities. There are real use cases for the decentralized storage and trustless verification of data, such as ID management, credit score monitoring, banking, and even supply chain management. We expect some early non-gaming dApps take traction towards the end of 2023.
6. 2023 will be the year of reckoning to switch from “pseudo-web3” companies to truly decentralized projects and thoughtfully designed protocols.
2023 will mark the fifteenth year since Bitcoin’s white paper was first published, a white paper that describes a fully decentralized payment processing and verification network. Since then, much wealth has been created in blockchain-related projects, and much has been lost in pseudo-decentralized ones.
Over the past four years, especially during the crypto bull run of 2020-2021, people have been blinded by the incentive and temptation to “print money out of thin air.” Collectively, the market has chosen to ignore the difference between truly decentralized protocols that is the real web3, and centralized platforms that facilitate token transactions. We all have learned a costly lesson in 2022.
7. Regulations are coming for web3 and it could be a good thing.
For years, the U.S. has been reluctant to provide guidance for DeFi companies, resulting in major exchanges fleeing the U.S. market (while still taking U.S. customer funds). Regulators ultimately had to “relearn” the lessons of traditional finance because so-called DeFi companies are still centralized entities. But such relearning often comes too little, too late. In the aftermath of the FTX scandal, regulations are coming into place. With regulation guardrails, we will see less over-promise and hype in web3 projects in 2023 and beyond. It’s a healthy and natural progression for web3 companies to truly enter the mainstream.
8. In 2023, we will start to see more intersection use cases of AI and web3.
AI and web3 are technologies that can complement each other in powerful ways. AI could analyze large amounts of data and enable new types of applications, while web3 technology can be used to securely store and manage the data that AI systems rely on. For example, AI can optimize supply chain activities and smart contracts can self-execute transaction agreements. AI can also conduct biometric recognition and then store the data securely and transparently on a blockchain. Yet so far, we have rarely seen the two technologies intersect in meaningful ways. At Leonis Capital, we firmly believe that AI and web3 have the potential to revolutionize the way we live and work. We are particularly excited to see how the two technologies can intersect and create new use cases.
The Macro: Mega-Trends
9. There will be two main types of VC investments in 2023: mega deals on “consensus-supercycle” companies and value deals on later-stage companies.
In 2023, VCs that have raised mega funds previously will face pressure to deploy. One type of investment is doing mega deals in companies we put in the “consensus-supercycle” quadrant. These are large rounds with high valuations in hot sectors. The companies tend to be founded by high-profile founders who were often executives from big tech companies or headline-grabber superstars (think Magic Leap, Quibi, and even Theranos). When the macro turns conservative, bigger VC funds’ investment committees tend to seek false comfort in herds.
Another type of investment is “value deals”. Large VCs, regretting overpaying valuation for the past two years, will be circling like sharks for down or flat rounds in later-stage companies. For early-stage investments, VCs will be taking more time and trying to seek lower valuations whenever they can, so founders should expect the fundraising will take longer time to negotiate and to close.
10. 2023 will be a tougher year to exit but a great year to build and invest.
In a gloomy economic environment, capital will have the upper hand and will be expensive. The tighter capital supply forces startups to remain heads-down, focus on building, and remain frugal. This is why some of the most iconic companies are built in times of economic downturn. However, it’s possible the lackluster financial markets mean that it will be hard to make meaningful exits, either through high-valuation IPOs or through large acquisitions. If the market condition stays on its current trajectory, we expect 2023 is going to be the year of “heads-down building”, that calls for patience, focus, and resilience.
Wish you health, prosperity, and happiness in 2023 and beyond.
Stay curious, keep on venturing. 🌌