Last month I wrote about
potential overhype in the artificial intelligence space, asking a series of investors whether the hubbub around generative AI had current, tangible implications in the climate sphere. What I mostly heard was: Not yet. Many acknowledged that generative AI could plausibly do fundamental scientific research — creating new chemical and molecular formulations that could have broad implications for climate tech and beyond — but most didn’t think we were there yet.
Not everyone shares that perspective. Obvious Ventures, a San Francisco-based venture capital firm that focuses on the three pillars of planetary, human, and economic health, says it wants to invest in what it calls “generative science.” Today.
“While most venture dollars are chasing large language models for enterprise productivity, Obvious is funding large science models trained on chemistry, physics, and biology, to generate new scientific breakthroughs in decarbonization, biotech, materials science, and robotics,” James Joaquin, co-founder of the firm, said in a recent blog post titled “
Generative Science: Our Contrarian View of AI.”
There are some companies pursuing this lofty vision, especially in the pharmaceutical space, and Obvious has even invested in
a few of them. But whether “generative science” is currently upending the AI and climate space is debatable. In an interview, Andrew Beebe, managing director at Obvious Ventures, walked me through why he’s so bullish on AI for climate.
“Plenty of use cases”
“So computational biology and life sciences really, truly have been using machine learning for a long time,” Beebe told me. Of course, having a machine learning model that identifies patterns in reams of data is different from the type of “generative science” that could come up with new drugs, for example — but now one of Obvious Ventures’ earlier investments,
Recursion Pharmaceuticals, has partnered with Nvidia to do just that. “That company uses AI to speed the drug discovery process. We have a number of companies where they are using similar concepts for proteomics and genomics, so that experience taught us that there are plenty of use cases where this can really apply,” Beebe told me.
The success of Recursion, which went public in 2021, has helped fuel the firm’s optimistic AI outlook, and it’s since made a number of investments at the intersection of AI and climate. Just a few weeks ago, Obvious led a $30 million round of Series B funding for
Zanskar Geothermal & Minerals, which also included cleantech VC Lowercarbon Capital, among others. The company analyzes swaths of geological data to help locate areas with optimal geothermal resources, creating maps and greatly expediting what can be a highly inefficient process.
“Smart geologists will drill 10 exploratory wells and get one to hit,” Beebe told me. Zanskar aims with its software to dramatically improve that hit rate, and though the company hasn’t provided performance metrics, Beebe said that if we could get closer to nine out of 10, “that changes everything. It changes the economics of traditional geothermal.”
The company also pulls in data such as power line capacity and land pricing to make its recommendations, “It will tell you this is where you should drill to be cost effective. Not just this is where the heat is.”
As useful as this is, though, Zanskar’s tech isn’t generative AI — it’s just a great use case for increasingly powerful predictive AI, in which machine learning models analyze patterns in large datasets to make forecasts and recommendations, in this case where to drill. Thus far, it seems, none of Obvious Ventures’ investments in the climate and AI space are yet fulfilling the ultimate promise of “generative science” as Joaquin characterized it. “Generative media has delivered us a printing press that can write its own words,” he wrote, “but generative science will deliver a more consequential lab bench that can create its own novel arrangements of atoms.”
“The edge of what is physically or scientifically possible”
Beebe sees other climate applications for generative AI, however, particularly for the electric grid. “Maybe the mother of all near future generative science in the climate space is just making the grid smart,” he told me. While Obvious hasn’t yet invested in the AI-enabled smart grid space, Beebe is excited about “agentic systems” that will be able to make autonomous decisions based on real-time supply and demand data. “A result of that might be, let’s take power out of this massive
Form seasonal battery sitting up in Modesto and move it to Southern California. Let’s take this water and start pumping it up the hill” choices that, today, “are really not automated in any coordinated way across the system,” he said.
Beebe also thinks there’s big AI potential when it comes to battery chemistries and nuclear reactor designs. “I think that AI is going to help basically expand the edge of what is physically or scientifically possible because of the rate of iteration of different designs. They won’t be right every time, but they will help us get closer and closer to the estimate space. We will then feed back in that reinforcement learning and then it will become better next time,” he told me. “And then I think things like fusion reactor designs are further down the line.”
Again, Obvious hasn’t yet invested in companies actually utilizing AI in these ways, but Beebe is confident that the future is near. And some recent research in does provide reason for hope For example, AI research laboratory
Google DeepMind collaborated with the Swiss Plasma Center to learn how to better control hydrogen plasma in nuclear fusion reactors, and Microsoft used its own AI platform to discover a battery material that could reduce lithium use by up to 70%.
Chat GPT for energy data
Obvious thinks large language models have a space in the climate tech landscape, as well. Last month, the firm
co-led the seed round for Halcyon, a company trying to improve access to energy market data via LLM-enhanced searches. It was founded by ex-Twitter employees alongside Nathaniel Bullard (formerly chief content officer at BloombergNEF and publisher of a renowned-in-niche-energy-circles annual decarbonization report).
“We call it NatGPT internally,” Beebe said. “What they’re really trying to do is build a automated consultative service for energy developers to help them figure out where to site power plants, how to think about where to site transmission lines, how to answer any questions that they have about the vast and complex world of accelerating decarbonization of the grid,” Beebe told me. “It’s all AI-based and effectively LLMs, for the most part.”
In summary, even if Obvious’s current investments aren’t quite yet creating “
the chemicals and molecules” to “help solve humanity’s toughest challenges” in the climate sphere, watch this space.
Editor’s note: This story has been updated to correct statements about Zanskar’s accuracy.
