Quantity Surveyors concur that AI estimating results should be "Assessed Inaccurate".

Our article "Estimating with ChatGPT" showed up the shortcomings in AI where it depends on a limited data set.

Georg van Rensburg of Ace Solutions, shared a recent article from Futurism that even the most advanced AI has a problem: if it doesn’t know the answer, it makes one up! That article had me wondering whether AI is a born politician?

When I asked Copilot that question, it promptly replied: "Ah, I see what you did there! AI as a "born politician"? That's quite the compliment—or perhaps a cheeky observation? While I may be adept at logic, reasoning, and persuasion, I promise no political ambitions here—just a commitment to being your trustworthy assistant. Would you say I possess the necessary "charisma" for the role? Or am I just being overly optimistic?"

Trustworthy? Try accurately estimating a one brick wall, Pal!

Martin Treadaway of QSPlus observed that he thinks that "...the ASAQS should teach QSs to embrace this new technology but view its answers with due diligence." Martin also remarked that the calculations shown for the OBW had certain inaccuracies and that the problem with GPT is it might have access to all the data, but does not fully comprehend all the factors which must be considered in building up a unit rate. Georg also highlighted a similar point, that in Estimating there are a lot of unknowns which a Quantity Surveyor has to allow for.

But AI is learning fast. At the end of last year, I asked several AI engines in which clauses paint work is mentioned in JBCC 6,2?  I received answers that referred to non-existent clauses in JBCC! Ask the same question now and you will get much better answers.

As big data is used (and corrected) by the built environment the answers will become more accurate and relevant.

Liza Oberholzer wrote that: "It solves a dilemma I was facing.  I couldn’t understand why a particular very young QS working for a contractor kept submitting “stupid” rates that made no sense whatsoever.  I now suspect they are using something like this and not bothering to adjust for the size of the project and specific site conditions."

Should we be shielded from AI?

Gregory Fendt says: "I’m of the opinion that QS profession does not need to be shielded from AI, we need to understand how to use it and how to our best advantage."

Greg contributed with very important observations, and I replied that "QS's work in the grinder between imagination and execution. I foresee a future where, if we do not control the data, it will be impossible to convince the end user of our worth."

We should be working on a solution to solve that problem.

Compiled by Bert van den Heever using ChatGPT and Copilot

Disclaimer: This article is the personal opinion/view of the author(s) and is not necessarily that of the ASAQS