Hawaii’s Kilauea is actively erupting, and I’m really questioning my life choice to not chase down grad school at the University of Hawai'i. The temperature at my house this morning was a frigid -11 F. Because I am both deranged and committed to consistency, I still got my workout in at 5 am. By the time I got back, it had dropped another 3 degrees. 🥶
To warm us up from our head down, I start this week’s edition off with a hot take on AI trends: For me, the “it’s coming” feeling around the AI/ML revolution really seems to be changing into an “it’s heeeerre!” arrival, as more companies are demonstrating use, value, and continued expansion of AI. Let’s take a look.
Sarah Compton
Editor, Enspired
Tech Trends From an AI and ML Expert
Anggalih Presetya/Shutterstock.com
I recently connected with Patrick Ng, co-chair of AAPG’s Deep Learning Technical Interest Group and founder and CEO of data science, automation, and analytics company Shaleforce, to talk trends and the next big thing in the tech space. The technical interest group is meant to help members better understand, and more actively participate in, the AI-revolution.
Ng’s background:
After completing his undergrad at University of Oregon and his master’s in geophysics from Yale, Ng nabbed his first R&D job at Western Geophysical, where he would build a foundation 22 years in the making.
That foundation led to his passion building models that combine physics with various numerical methodologies, including neural network simulations and MPT (modern portfolio theory).
MPT basically quantifies risk and allows the user to adjust their portfolio to meet their desired level of risk based on the combination of assets in that portfolio. I think this is what we have to thank for those age-based retirement options in our 401(k)s.
The catch-all phrase “AI and ML” is starting to age poorly, as people (including us geoscientists) want to understand, with specificity, what exactly we need to be learning and how to apply it.
Here are a few key takeaways from Ng and I's conversation:
On his desk: When it comes to hardware used to deploy AI, he favors Nvidia Project DIGITS, a small desktop powered by one of Nvidia’s superchips. Ng called it a “tiny powerhouse with the strength of a data center.”
Software selection: On the software side, his voice goes slightly against the grain. Rather than LLMs (large language models such as Chat GPT), he sees small language models as the way to go, and this fits with his overall message of specificity in AI and ML applications. Small language models use more compact architecture and fewer parameters than LLMs.
Up next: Ng points to smaller, faster, and smarter AI that is fit-for-purpose and will perform specific tasks. I am a big fan of this, and I think most geoscientists will be, as it goes along with the drum many of us have been beating that says, “AI is not a solve-all-your-problems solution!”
A message from AAPG Academy and Eliis
Register now to join AAPG Academy and Eliis on 28 January at 9am CST for a free webinar to learn more about downscaling seismic and geological interpretation to construct geological and reservoir models.
Expert speakers Nicolas Daynac and Jean-Claude Dulac from Eliis will demonstrate how to:
Use synthetic seismic data from a digitized analogue model alongside real-world basin seismic datasets to show how the Relative Geological Time method enables characterization of depositional sequences at well log resolution
Despite the reputation of geoscientists as being worse in math and physics than our engineering counterparts, we have all taken at least one college-level physics course, so we understand that matter (and therefore energy) cannot be created nor destroyed.
It can only be transformed.
To that end, hermetically sealing a hydrovoltaic cell to increase its efficiency by preventing the escape of water should surprise no one, but thought experiments justifiably do not draw the same excitement and enthusiasm as successful, real-life experiments.
Let me explain:
When water interacts with other surfaces, energy can be created.
Hydrovoltaic cells capture this energy, though such interactions typically rely on sunlight as the instigating power source and require a constant stream of water to keep the reactions going.
Researchers at the Chinese Academy of Sciences have combined a few surprisingly affordable and common materials to create a tiny powerhouse of consistent and sustainable energy using this information.
How the tech works:
The team created a hermetic hydrovoltaic cell (HHC) with two main parts: a sealed container and an electricity generation unit (EGU) inside the container.
The EGU is placed in about 5 centimeters of water and is composed of a polyethyleneimine-modified carbon black (CB) layer and a layer of tissue paper. Together, these form a heterogeneous wicking bilayer on a quartz plate under the effect of capillary force.
The capillary force pulls water up (this is where the electricity is generated!), the wicking layer removes the water from the EGU and causes diffusion. The ambient temperature in the HHC then evaporates that water, which later condenses back into the small layer of water at the bottom of the unit.
By hermetically sealing the cell, water is never lost and can be constantly recirculated through the system.
The catch? It’s extremely tiny. The output was roughly 150 millivolts (about a tenth of a volt, and your average outlet is 120 volts), but it persisted across several days. Increasing the number of cells in parallel did increase the energy output.
Increasing ambient temperature, including through adding sunlight, increased the efficiency of the reaction, indicating great promise for this tech in water-poor and high-temperature areas.
👍 If you enjoyed this edition of Enspired, consider supporting AAPG's brand of newsletters by forwarding to a friend or colleague and signing up for our other newsletters here.
➡️ Was this email forwarded to you? Sign up for Enspired here.
✉️ To get in touch with Sarah or suggest a topic for Enspired, send an email to editorial@aapg.org.
AAPG thanks our advertisers for their support. Sponsorship has no influence on editorial content. If you're interested in supporting AAPG digital products, reach out to Melissa Roberts.
.png?upscale=true&width=220&upscale=true&name=Sarah-Compton-Headshot-Signature%20(1).png)
