An AI wave is about to change startups, industries, and the region.

Imagine your core value creation activity could be 2x more efficient. What could you build with the extra slack? Now imagine 20x. What about 200x? 200x is what Colombian education platform Slang predicts will be its change in course creation capacity next year.

Now picture your startup launching in different languages simultaneously. Everything from marketing, onboarding, and customer support. What about starting day 1 in every language, all at once. When Andrea launched her mental health app from Cancun, she reached 12 million users. Imagine the reach without language barriers. That future is within her grasp.

With Hi’s new fund, we seek to back Latin American miners looking for asteroids.

tl;dr

• Language processing evolved from recurrent neural networks to faster, more accurate transformer models, accelerating AI capabilities and adoption
• AI’s rapid mainstream adoption, led by chatGPT, signifies a technological revolution shifting from prediction to the generation of information and content
• Startups today should focus on optimizing processes for efficiency and/or enhancing their value proposition for user benefit with AI
• AI-native startups are primarily focusing on optimizing tech development, creating diverse media content, and augmenting human capabilities
• Non-AI-native companies have the urgency to enhance and defend their unique advantages and value propositions as we see technical founders quickly supercharging services
• The success of AI implementation hinges not on the model used, but on the problem it solves and the value it adds
• Significant potential risks of AI necessitate a focus on responsible model development, including the analysis of edge cases and biases
  

‍

All startups, everywhere, need AI.

I barely have had a chance to understand AI and LLMs. All I know is that it’s about to change startups and startup building forever.

Initially, state-of-the-art language processing was accomplished through recurrent neural networks. Then the paradigm shifted with the arrival of the "Attention is All You Need" approach, which significantly accelerated everything. Instead of relying on large language models that focused on predicting the next word based on the last one, we moved toward transformers.

These systems leverage all previous words and their respective positions to make predictions. Their ability to work in parallel has resulted in faster and more accurate processes, leading to the widespread adoption of this approach.

It took decades to establish the AI baseline, but for GPT it only took five days to reach one million users. The launches of DALL-E and ChatGPT marked the arrival of AI to the mainstream.

Generative AI demonstrated to the average person not just the power use case of everyday AI, but also the fundamental technology revolution hidden behind closed doors, which emerged when we finally had the capacity to compute and process billions of parameters.

Consequently, a technology that the mainstream only knew primarily for increasingly accurate predictions began to extract information from unstructured content and generate personalized output. While that is certainly impressive, it begs the question - what can I do with it?

There are two significant outcomes that every startup should be leveraging today; first, the optimization of their processes for increased efficiency, and second, the empowerment of their value proposition to offer superior solutions for their users. Any claims beyond these two outcomes should be met with skepticism and would recommend consulting with someone who really is an expert on the matter.

Luckily, we are on the first row and with something more than popcorn to witness these changes.

Most AI-native startups we see are focused on three topics; developing tools to optimize tech development such as creating code and minimizing errors,  generating media content such as videos, images, text, music, and brands, and innovations that augment human capabilities like providing the tools to empower a more productive human within the workplace.

For those who are not AI-native, the clock is ticking. It's time to leverage their unique competitive advantages and enhance their value propositions. We've observed that most of our technical founders have promptly decided to supercharge their existing services.

Slang is creating individualized journeys and content five times faster. Yana is extracting every bit of value from their 2 billion historical conversations, treating them like lemons to squeeze in order to train their model. Corporate expense management platform, Mendel is leveraging their close corporate relationships and integrations to make handling your travel invoices seem like magic.

New edtechs, mental health chatbots, and enterprise SaaS startups won't have these strengths, for example platforms, engaging UX, historical data, or commercial contracts available to run their technologies.

We are particularly excited to see human capital startups leverage AI. Our most recent Hi insight maps the cross-sectorial reach this technology can have: from edtech and wellness to fintech and climate tech.

One lesson Julian, founder of Senzai, an outcome-based AI startup, taught me is that it's not about the model you're using, but rather about the problem you're solving. It's about what you build on top of this new tech. E.g. in Senzai's case, it's about helping increase revenues; for Yana, it's about assisting users in navigating their emotional journey.

Historically, we have more than a few precedents on how new technologies can take some missteps, but none of them are so powerful as the promise of what AI can become. Consequently, we must build upon these revolutionary models responsibly, and understand the critical risk involved in a holistic way.

For those who are not AI-native, the clock is ticking. It's time to leverage their unique competitive advantages and enhance their value propositions.

This approach includes everything from the generation of an AI hallucination, to a privacy violation, to the ongoing problem of misinformation. Gabriel Weintraub, professor at the Stanford GSB, refers to this concept as an MVRP (Minimum Viable Responsible Product). He recommends founders to be intentional about analyzing edge cases and biases.

We, Hi, need to change or become irrelevant. Federico and I have decided to go all in AI in Latin America. Our new $100M fund will be dedicated to fund Latin American AI first and AI-enabled startups solving the hardest problems in the region. We seek to back miners looking for asteroids.

Last May, we launched LAIC, Latin American AI Club, a community of founders, investors, and corporates to collaborate to make AI a force of good for the region. Join the club :)