By Mike Anderson, P.E.
In February’s “National Geographic”, I saw a headline that read “AI just deciphered part of an ‘unreadable’ ancient scroll. Here’s what it says.” (See the end of the article for the quiz.)
I shared the headline with my family and made up a few responses that I thought could be punchlines. I shared my responses at the end of this article along with another response that came when I asked ChatGPT to provide a punchline as well. Take the quiz and see if you can pick which one is AI-generated.
My hypothesis, similar to an argument advanced by Detective Spooner (Will Smith) in the movie “I, Robot,” is that a machine can’t write humor in the same way a human can. Humor is personal; there’s a certain illogic in what people find funny that I doubt can be captured by an artificial intelligence. Stephen King even used this concept as a plot resolution point in his Dark Tower saga. Although I don’t have the technophobia that Will Smith’s character has, I still have some skepticism about relying on AI for everything.
Having said that…AI is great at sorting through lots of data and looking for patterns. It is that skill that allows it to predict the response of a sample based on a limited set of actual observed data. In other words, in the asphalt world a technician could potentially run just a few tests on a sample and AI could fill in the blanks to estimate what the rest of the sample properties are.
AI can work well within the confines of the data set we have. But what if something changes and it is presented with a situation that is new or different? That is where the challenge comes in.
We see this in the asphalt binders used today in the PG specification system. The PG Asphalt Binder Specification is over 30 years old. When it was developed, it was based on asphalt binders that were commonly available in the late 1980s and early 1990s. Those binders were often produced from single source crudes using conventional refining techniques. Today, the crude slates, refining and blending techniques and additives/modifiers available to use are different enough that the asphalt binders produced now can be said to belong to a different subset than those asphalt binders of yesteryear. If true, then maybe it would be wise to take another look at the system to see if any revisions are needed to address the products in use today?
The premise behind the NCHRP 09-60 research project is exactly that…the formulation and manufacture of asphalt binders used in pavements have changed significantly since SHRP (Strategic Highway Research Program) ended in 1993 and a reassessment of the specification is needed. That project started in 2016 with two planned phases and is currently scheduled for completion in 2024 after adding a third phase.
In Phase 2 of the project, the research team, led by Western Research Institute, recommended that the PG Binder Specifications (AASHTO M320 and M332) use Delta Tc (ΔTc) as a rheological parameter related to binder relaxation that is indicative of “premature thermally-induced surface damage” – in other words, early loss of asphalt pavement durability due to surface cracking. This type of cracking can be related to the compositional balance of the asphalt binder…meaning it can pick up changes due to the natural effects of oxidative aging as well as the addition of additives/ modifiers that cause an imbalance in the structure.
The issue is that while relaxation properties such as ΔTc are determined in the linear viscoelastic (small stress and strain) region, cracking is a failure/damage phenomenon that happens in larger strain regions. To address this and fully capture the benefits of enhanced strain tolerance imparted by polymer-modified asphalts, the research team added a qualifying parameter, Delta Tf (ΔTf), which is determined using Tc,S from the BBR (the critical temperature where the low-temperature Stiffness is 300 MPa) and the critical cracking temperature determined from the Asphalt Binder Cracking Device (ABCD) test procedure in AASHTO T387. A higher value of ΔTf is indicative of enhanced strain tolerance that might allow an asphalt binder to overcome issues with relaxation properties.
The NCHRP 09-60 research team has completed two phases of work and is currently further validating their recommendations and examining the variability of the proposed procedures – a critical need for use in a purchase specification – in Phase 3. In doing so, they are adding to the data set and improving the capability of AI in the asphalt world…at least until the next stage of evolution comes along.
For more information on the research, testing, and training services provided by the Asphalt Institute Laboratory, please contact Wes Cooper, Gary Irvine or Mike Anderson. Our avatars will be available to assist you; their logic is undeniable.
Mike Anderson is the Director of Research and Laboratory Services at the Asphalt Institute.
In February, National Geographic reported that AI had deciphered part of an “unreadable” ancient scroll. Here’s what it says:
1. If you can read this, thank the Phoenicians.
2. Be sure to drink your Ovaltine.
3. It turns out the ancient Egyptians were just really into cat memes.
4. So, I said to Nebuchadnezzar, “We better get started. Those gardens aren’t going to hang themselves.”
Answer:
3. Was the ChatGPT answer. Honestly, it was a better answer than I thought I would get.