AI versus Expert System Software.
The long explanation.

Click to enlarge

Where AI fails, and always will, when compared to systems like Mistral's, is due to a number of major issues.

Firstly, it takes a prohibitively long time to prepare and assemble a list of the numerous design parameters for any engineering project. Even if one creates a design parameter template form, the form still has to be completed, manually. With Mistral the process is speeded considerably, using a system of multiple choice clickable menus and interlinked, dynamic error trapping. What takes a matter of perhaps half an hour of work preparing for and presenting a question for ChatGPT for example, Mistral can and does so within just 10 seconds. Faultlessly. Human scientific error is also practically impossible with Mistral whereas with AI, being neither sentient nor knowledge and experience based, scientific or technically impossible errors, even simple 'typos', will frequently go unnoticed. Furthermore, a single design parameter template is practically impossible to create. No two, even consecutive projects are ever the same. Frequently not even remotely similar.

Secondly, there is no live, in real time, user interaction between user and AI. One of the insurmountable advantages of Mistral's system is that design engineers can experiment, in real time, by adjusting input and instantly, in sub seconds, witness the effect of those input changes. With AI this is and always will be impossible. Any change to a design parameter requires the whole project design sheet having to be re-sent. A gross waste of time.

Next there is the question of GUI. Graphical User Interface. AI cannot return a visual presentation of a developing project design on screen. Even a two dimensional one.

Then there is presentation. From what we have seen so far AI returns a jumbled, jargon infested mess. Heavy with totally irrelevant or unnecessary information, with no attempt made to assemble a presentation in any visually acceptable format. All but useless for a Sales Engineer thrusting this before their (usually) lay purchaser prospects. Notwithstanding the fact the returned 'document' will be riddled with spelling and grammatical errors and obvious mistakes. This latter problem may be resolved in time, as AI 'scrapers' continue to steal knowledge from more qualified sources. Providing those sources don't start pulling much of their hard earned information off the web of course.

Which brings us to this critical point. Who has corroborated AI's result? From what qualified and certified source or sources was AI's contributory data and algorithms stolen (scraped)? No one is ever told. Over estimating a multi-million Dollar Commercial Refrigeration or A/C project by ten percent then a firm will be bound to lose the contract at time of tender. One situation is far worse. A result twenty percent lower than an experienced engineer knows it should be then the engineer will spot the problem and spend yet more time honing or fine tuning their long set of design parameters, for re-presentation. However what if the error is a design which when costed is ten percent lower than it should be, and thus the contract is won, but fails upon first site commissioning? It isn't possible to make physical corrections to extremely complicated, interactive and expensive machinery, pipework and sophisticated electronic control systems. Might as well rip the whole installation out, scrap half the machinery or put it into long term storage hoping that one day some of it might get used elsewhere. In other words. Bankruptcy!

Finally, Mistral Expert System software result forms don't simply save and display the answers. They also, always save every key press and input which went into creating that result. Important? Extremely so. It means the presentation quality result file can be reloaded a minute later, or an hour later, a week later or even three decades later, and still be user interactive. Even after countless computer Operating System upgrades. Facilitating saving duplicates and editing or modifying them to recalculate late design changes or for application to new projects.

Here though is the real joke of AI. We have tried asking every AI platform an identical set of design parameters and as anticipated no two AI platforms returned the same answers. Not a case of knowing or guessing which one was correct. THEY WERE ALL WRONG! It gets better, or worse depending upon how one looks at it. Ask the same AI platform the same set of questions but two or three days apart and you will receive a different answer. Not something any Lloyds ISO9001 invigilator is going to be very impressed with!       Thought for day

Mistral's commitment:
Bringing benefits of computerisation to our RAC industry - without the commonly associated problems.

Close and return to previous page