OpenAI for Climate.

The next green Google is (maybe) already there.

Hi there 👋

In The Climate Tiger, I share my thoughts and analysis on business strategies, tech news, trends, and secrets to supercharge the climate transition. And I would love to get your perspective on those. 

If you are not subscribed yet, it's right here 👇

---

“Eliott, I'm against AI !”

A few weeks ago, my grand-mother, 92 years old, interrupted me with that weird sentence. I remember jumping off my chair with the feeling that she had no idea what she was talking about.

Most of my climate founder friends are like my grand-mother. They say that AI is a distraction and they prefer sticking to what they know - solar panels, electric storage, precision fermentation. It’s not for them. It’s a choice and I respect it. But is it really the right direction?

I’ve noticed that the people who care the least about AI, are also the ones who use it the most. They won’t leave their GPS, Netflix recommendation, their Google Home or Siri (yes, my grandmother is a techie). That’s the silent adoption of AI. You don’t realize it, but you are already into it. So, because you can’t refuse a better user experience, and choose to live in middle age, you can’t be “against AI”. It’s like saying you are “against” bread and butter but you like eating toasts every morning.

If AI is the future, the question becomes : “how can it benefit the climate?” And there, yes, I think we are at a turning point in history. 

Imagine two doors in front of you: the first one uses AI to grow the polluted economy (more goods, more travels, more CO2…etc). The second one takes advantage of AI to speed up the climate transition. It’s a shortcut. 

Unfortunately, I feel OpenAI and Google are pushing the wrong door at the moment, without climate consideration. So today, we are exploring the second one.

I’m convinced there is an opportunity to reshape everything we know. The opportunity to build the “next green Google”, maybe the most profitable AND sustainable of all time. That could be a decisive shortcut for the climate and we could save years or decades, just now. I would go even further. The climate direction can solve the AI’s gridlocks we experience at the moment.

I hope you are as much excited to dive in as I am. Let’s go!

---

OpenAI for Climate.

First, why do I say AI can reinvent Google? 

Because AI chatbots (or “AI assistants”) are going to replace the search business. If you think about it, from a user point of view, search engines are pretty simple. It's an input (the ask) and an output (the result). Google won the search engine battle in the early 2000, just because it provided the best output for any given inputs. That being said, you still need to browse the web to find the result you want (and it’s a pain sometimes).

Not with AI anymore. Instead of organizing the web to highlight the most probable outputs, AI will answer you directly. It’s a dramatic improvement of the search user experience.

It’s not just search. It’s all the economy which is connected with AI. Since ChatGPT launched their plugins, you don’t even need to browse the web to buy or order stuff. You can book a flight, a hotel, a restaurant or whatever you like directly from the chatbot. Google understood it, that’s why they rushed to launch Bard AI a few weeks ago.

But you might ask "If Google already does it, what can be disrupted?”

I think Google is still far from winning the AI battle. Let’s summarize the takeaways from the past few months. I promise, I’ll be quick.

The remake of David & Goliath (the AI version)

When OpenAI released ChatGPT API last November, the question was “Who is going to win between the infrastructure layer and the interface layer.” ?  Like the legendary fight between David & Goliath. The infrastructure layer is made of the large machine learning models - LLMs - developed by OpenAI or Google and able to process trillion of data (think Goliath). The interface layer is all the applications which use that technology (think David). 

Now, at first sight, if you had to bet on the winner, you might choose Goliath. That's what investors did. Everyone argued that building the infrastructure was the best way to create defensibility : the "moat" of AI. After all, if your neighbor can build the same app than you with OpenAI API, how can you differentiate? So Goliath was about to win.

Until the past few weeks.

First, mid April, Sam Altman, the founder of ChatGPT, declared that the large models were not sustainable and AI will need to find new ideas to thrive. A few weeks later, an internal document leaked from Google, with the title “We have no moat, and Neither does OpenAI”. The idea is that large models won’t create a competitive advantage and smaller quality databases would scale faster. Game over Goliath.

If Goliath doesn’t win, can David?

What if David & Goliath had a baby ? 

In a conversation with Reid Hoffman in September 2022, Sam Altman gives some elements of answers. According to him, neither the infrastructure layer or the interface layer would win. But a middle ground between those two. In other words, the future of AI is not the battle between David and Goliath but the baby they will have together. AI apps 2.0.

• These new AI apps will process and transform the data with a particular use case, but won’t build their own models.

• They will offer their own interface or launch APIs to other product interfaces.

• They’ll focus on creating smaller databases of proprietary data, but won’t need to use all web data points (like most LLMs does).

That last sentence is important. One of the big criticisms of AI today is that information is not verified.  If you only process information from a quality dataset, you solve this issue. You can be trusted. 

It's funny that the race to build the biggest, quantitative models turns out to be the one to create the smallest, qualitative ones. It reminds me when the Iphone versions became too big to fit in our pockets and Apple decided to build smaller and smaller ones.

Anyway, if David & Goliath have a baby, it's good news. I don’t know if you realize it yet, but it means that  we could reshape the cards and threaten the incumbents. So we have the opportunity to build a new Google.

I know, at this point, all of this sounds very abstract. So let’s take a concrete example with an app that (almost) nailed it. 

Yuka : the OpenAI for food and cosmetics.

Yuka user growth in the first two years (today more than 30M)

Yuka is an app that tells you whether your food or cosmetics are dangerous for your health, when you scan it in the supermarket. That’s a legit concern. We don’t want the food we consume to kill us. I’m amazed when I see my family members still using the app on a daily basis, whereas I introduced it to them 3 or 4 years ago. And Yuka user growth proves that the market needs a new source of trust. 

Yuka and OpenAI have a lot in common. They were created around the same time between 2015 and 2016. Like OpenAI, Yuka relies on open, public data source - Open Food Facts, and like OpenAI, Yuka takes advantage from user contributions to improve the algorithm and product. But Yuka goes deeper on one aspect. They transform and check the data with a mix of tech and a team of nutritionists and experts to display a qualitative dataset. It goes one step ahead from the data you might find on the web - to this extent, it’s proprietary data. And it’s a moat.

Yuka has almost all the ingredients of an AI app 2.0 (proprietary data with AI, consumer interface…). A single thing is missing : an AI integrated user experience. Imagine if you could ask Yuka for a recipe of healthy food, and get the ingredients delivered into your home by connecting your supermarket delivery system. Awesome, no?

Now, let’s imagine we could do the same, not only for food or cosmetics, but for everything (travel, shopping, leasure… ) and only with sustainable, verified products.

According to me, it could solve one of the main bottlenecks for the climate and accelerate the transition in a few years. Climate data faces the same issue as food and cosmetics : a lack of transparency. How can you know if this product is sustainable or not? What are the tons of CO2 it produces ? Why should you buy it and not another one? 

With AI, we can change the consumption habits of everyone, and build a “climate filter” while improving the user experience. That’s the “AI door” for the climate.

Introducing “the green AI Google” (without Google). 

A lot of people discovered the power of ChatGPT when they launched their plugins integration a few weeks ago. They understood that AI can, not only give you answers, but also complete actions for you.

Fewer people understood that if ChatGPT organizes the data, they also could decide which product you have to buy. If Walmart is integrated with ChatGPT and Amazon is not, you end up buying Walmart products. If both are, who knows if ChatGPT won’t still prioritize Walmart ? 

When others see a threat in that algorithm, I see an opportunity for the climate. What if we could prioritize climate alternatives instead of polluted ones in the algorithm ? Let’s take the example of the travel industry.

Screenshot Google flights

When you search a flight today on Google, you have very little incentive to buy the “greenest” one. CO2 emissions are mentioned, but it’s easy to get lost in a big list of flights with a lot of parameters (prices, schedules, itinerary…). What if we simplified the user experience by 10x with AI and book directly the “greenest” option for you ? You saved time (and the planet).

Of course, it would require important climate data checking (like Yuka does for nutrition) but we need it anyway, and it’s an opportunity to build proprietary, defensible datasets. At the same time, I expect a virtuous cycle to start. If brands want to get integrated, they would need to calculate the CO2 emissions of their products. We already saw it thanks to Yuka. A lot of industrials started to care about their positioning in the app and improved their products.

One question remains : the competition. There is no doubt both Google or OpenAI can build integrated products with multiple plugins, even if OpenAI is the first mover here. Of course, a proprietary database of climate verified results sets us apart, but why would a green alternative succeed? 

I just think we are at the right time in history for this kind of product to emerge. I don’t expect my parents to be the first adopters of AI powered search or assistants (even if they use Yuka). But my generation (Y and Z) who grew up with the internet will be. That’s the generation who care the most about climate change. 

And there is more. AI is a big reset button in the history of tech. When green search engines like Ecosia launched in 2009, the battle of user experience was already won by Google. Apart from planting trees, Ecosia didn’t significantly improve the user experience. Today, we are all at the same stage.

We need to push the AI door for the climate, now. We need to embrace it and not let tech history repeat itself and accelerate the CO2 economy. We can’t be left behind. What an exciting time to be alive and an urgent time to act!

---

Sources :

• Packy McCormick, Attention is All your need (Substack)

• Dylan Patel and Afzal Ahmad, Google “We Have No Moat, And Neither Does Open AI (Substack)

• Ron Miller, Sam Altman: Size of LLMs won’t matter as much moving forward (TechCrunch)

• Greylock OpenAI CEO Sam Altman | AI for the Next Era (Youtube).

---

That’s all for today. If you found today’s post useful, share it with a friend or two.

Thanks for reading & see you next Monday!

Eliott