OpenAI’s ChatGPT chatbot has reached viral status since its late-2022 launch for activities like debugging code, creating recipes, and writing essays. In the spirit of experimenting with the initial capabilities, we wondered how the bot would approach a timely ESG topic: decarbonizing the railroad industry. This involves a transition to energies that emit less greenhouse gas, which traps heat in the atmosphere and alters climate and weather patterns.
Before we get to our experiment with the chatbot, let’s look at the background of the railroad industry—as written by a human.
Railroads and Emissions
Like many other hard-to-decarbonize businesses, the railroad industry has long capital lives, which means locomotives are on the rails for a long time before economically easier to justify changing to a new design. For many years, however, the industry has been working to reduce costs through energy efficiencies—and, coincidentally, lower carbon emissions. Progress continues with low double-digit efficiency improvements expected through 2030 at least, dependent on new technologies.
Railroads generally have a cost advantage over trucks in part because of three-to-four times greater fuel efficiency per ton, which takes form in different ways. According to the American Association of Railroads (AAR), trains on average can move one ton of freight 500 miles per gallon of fuel. Plus, a train carrying truck trailers can reduce road traffic- and congestion-related emissions, according to the Texas Transportation Institute.
A train’s freight cost advantage over trucks is normally greater across these scenarios:
For many high-value shipments, however, particularly those of low volume (both in number and space required to ship), shippers and customers may still value speed over cost and carbon efficiency.
Over the past few years, North American railroads focused on service improvements to relieve labor tightness, smooth supply-chain interruptions, reduce fixed costs per ton, and work with truckers to grow intermodal and container movements. These costly improvements have helped bolster rail prices. Because those efforts must be paired with changes in logistics demand, a balance between high utilization goals and availability of excess cars (where needed), and the use of new, more efficient technologies.
Several available calculators can help shippers compare the carbon emissions of barge, rail, trucking, and air shipments over different routes. One of the sell-side’s bullish rationales for the rails is that a growing emphasis (if not a mandate) on lower carbon emissions should tilt more freight to the rails, over time and where logistics allow.
Our ChatGPT Experiment
Overall, ChatGPT seems to serve as a good overview tool, but it does not yet provide recently updated information. Deeper subject matter expertise also seems to be an issue. The answers you receive from ChatGPT vary greatly depending on the way you frame questions to it, how many different ways you ask the same questions and your willingness to sift through repetitive responses.
Here is the simple question we posed to ChatGPT: “Can you give examples of what a railroad can do to reduce carbon emissions in order of dollars per carbon ton reduced?”
Following are the results. We do not comment on ChatGPT’s cost-per-carbon-ton-reduced estimates except to say that they are unsourced and unsubstantiated and even if accurate, should probably be considered with time required and scale of absolute carbon reduction.
ChatGPT: Improving locomotive fuel efficiency: This can be done through a variety of means such as reducing locomotive weight, improving aerodynamics, and upgrading to more fuel-efficient engines. This can cost as little as $10-50 per ton of CO2 reduced.
FactSet: The engine in most diesel locomotives generates electricity that powers electric motors to the wheels. Efficiency programs, then, can focus both on engines and electric motors. Engine improvements include fuel injectors, biodiesel, or renewable diesel (waste fats and oils), and better train controls.
Biodiesel mixed as 5% of diesel fuel is becoming more common where available. Biodiesel has some physical limitations in current engines, but it’s expected the biodiesel ratio will reach 20% eventually as it demonstrates safety for engines and as biodiesel infrastructure is developed. Controls and measurement devices, including cruise control, remain critical to achieving efficiencies. Electric motor efforts are focused on moving from DC to AC motors for more power, which enables longer trains.
ChatGPT: Electrification of rail lines: This involves installing overhead power lines or a third rail to provide electrical power to trains, reducing their dependence on diesel fuel. The cost of this can range from $50-150 per ton of CO2 reduced.
FactSet: Electrification experiments with overhead wires, as sometimes seen in passenger trains, are few and so far limited to short-haul movements. Possible electrification issues include route distance, installation cost, voltage availability, switching capability, and geographic impediments (e.g., remote locations or overhead bridges).
ChatGPT: Optimizing train schedules and routes: By reducing the time trains spend idling, and using the most efficient routes, railroads can reduce fuel consumption and carbon emissions. The cost of this can be as low as $10-50 per ton of CO2 reduced.
FactSet: Reducing idle time can help locomotives improve overall efficiencies, in addition to cruise-control technologies, electric locomotives for loading and unloading in rail yards, and more efficient rail yard turnarounds. All in, most rails are suggesting potential carbon efficiency improvements from 2020 to 2030 in the 10-25% range.
ChatGPT: Implementing renewable energy sources: This can include using renewable energy sources such as wind, solar, or hydropower to generate electricity for trains. The cost of this can range from $100-300 per ton of CO2 reduced.
FactSet: We believe this answer from ChatGPT refers more to the utility sector than the rail sector and speaks to the electrification of rails in an earlier paragraph.
ChatGPT: Switching to low-carbon fuels: This can include using biofuels, hydrogen fuel cells, or other low-carbon fuels to power trains. The cost of this can range from $150-300 per ton of CO2 reduced.
FactSet: New technologies—including pure hydrogen andonboard fuel cells and batteries—will likely become more significant in 2030 and beyond. With freight typically making up a high percentage of a train’s overall weight, powerplant changeovers to batteries or the addition of a hydrogen fuel tank car and a hydrogen fuel cell as powerplants are options under exploration.
Already, some hydrogen is used as a diesel fuel additive, though it brings down fuel efficiency slightly. A potential positive for longer-term uptake of hydrogen is that rails frequently move predictably point-to-point, and rail yards are often located in industrial, oil, and gas hubs where hydrogen facilities will more likely be developed.
Pure battery applications are more challenging—except in rail yards—given route distances and that the multi-hour running times of many rail movements would require larger batteries. Hybrid set-ups with batteries charged during braking and downhill segments are being looked at, however. Advantages for hybrid powerplants are smaller batteries and alternative fuels.
Regardless of powerplant technology, the longer operating life of equipment likely means complete fleet changeovers are a multi-decade event.
ChatGPT: It's worth noting that the cost per ton of carbon emissions reduced will vary depending on the specific railway, its size and operational characteristics, and local regulations and incentives.
FactSet: True, and the bot’s response highlights two overall elements of the experiment and global efforts to decarbonize:
-
There are no blanket solutions for the “best” approach to improve fuel efficiency and reduce carbon emissions across industries, sectors, or countries
-
Time, new knowledge, and capital are all required to enable efficient emission reductions
Additional Thoughts
In time, ChatGPT and other AI programs will have this and other articles in their databases to inform future responses to the same question we asked the bot for this article. A key question in the meantime is: How quickly will the information nexus be updated for current content?
In addition, if access to written content is limited, embargoed, or placed behind a paywall, how will AI pick up the correct, most learned, or most current information? A partial response may be that ChatGPT or other AI programs will, in turn, charge users for access. Perhaps the program becomes more like a Wikipedia page or an improved Google search rather than a brain that knows everything—including complexities and nuances—and delivers it to you instantly, just as you want it.
Lastly, it will be interesting to see how AI develops around numerical data. Because gathering, organizing, collating, and cross-referencing datasets is a complex, engineering-intensive effort, many datasets require a paid subscription. And quality control and audit can be vexing problems that, if even slightly insufficient, could be quite misleading to users of bot-driven data content.
As history has often shown us, welcome to another iteration of a brave new world.
Post Comment
Insight/Author%20Bios/Headshot_Avatars_NLWF.png){kind=avatar}
