Top 8 takeaways from the Cerebral Valley AI conference

Here are my top 8 takeaways from the Cerebral Valley Summit last week in SF. Panelists included a few dozen CEOs from the top Generative AI startups.

1. There's a big debate among investors about whether startups can build defensible companies by leveraging OpenAI's APIs. Absolutely! Companies that are closest to the customer, like Google, Apple, and Amazon, are the most valuable. It is much easier to build down the stack vs. build up. For example, Google Maps originally licensed services and data from companies like TomTom, but built all of the technology in house over time. 

2. OpenAI is set to capture 70%+ market share for API usage. Companies like Opentable, Instacart, and Stripe have already integrated their technology, and the diversity of data from these integrations will improve future models, add intelligence, and drive down costs.

3. The AI community is working to develop open source, smaller, and cheaper models. Databricks already has 1,000 customers using various LLMs, and Mosaic trained a GPT-3 quality model for <500k. Keep an eye on the adoption of these models over the next 12 months.

4. Solving the last 5% of customer problems takes a lot of work. Fine-tuning and prompt engineering are essential to adopting OpenAI's models to a specific task. Users are discerning, and startups need to delight them to avoid churn and retain customers.

5. The most important thing for startups is to get their product to market. In the age of AI, adoption is growing at an unprecedented pace, and the coming 12 months will see a dozen startups hit exponential growth with low churn and high retention.

6. Entrepreneurs need to spend more time thinking about monetization models. Many startups had very little to say when asked about how they would monetize. Develop a business model that captures the value you're adding to your customers. At some point, just like self driving, the venture capital dollars flowing in will stop. 

7. To develop successful AI business models, companies need to focus on the value they're adding to their customers and find ways to capture that value. For instance, how can you measure the value of increasing a paralegal's or an engineer's productivity by 5x? This requires deep thinking about monetization models and developing a pricing strategy that captures the added value. Companies that can do this well will be well-positioned for success in the competitive AI market.

8. The most exciting companies to me at the conference were Adept, Code Complete, and Tome. Adept is building AI that can increase productivity by 10x, Code Complete is building Github Copilot for enterprise, and Tome is building AI-native multimodal storytelling.

I am still learning, and drinking from a firehose, so take these takeaways with a grain of salt :) Welcome any feedback or comments, feel free to email me at justin@ovtr.ai

Overall, I loved the conference and it was incredibly energizing to see the optimism, engagement, and innovative spirits from the panelists and audience. 

Special thanks to

Eric Newcomer

for organizing!