Op-Ed: Is AI the unlikely unwitting saviour of human critical thinking? It’s unavoidable.

You don’t have to try too hard to find grim tales of the end of human critical thinking online or anywhere else.

It all started with the “post-truth era”, that folksy little well-paid global obscenity that could have been shut down from day one. The sheer volume of useless despair on the subject of critical thinking was almost as bad.

The arguments were somewhat undermined by tacit acceptance of the post-truth theory. Just because someone said it was the post-truth era, there were no other angles to explore.

That spectacularly ignorant perspective itself is the direct opposite of critical thinking. Nobody questioned “post-truth” on any level. It became a sort of ideological car park, and about as interesting.

Critical thinking and human survival

Meanwhile, critical thinking has become a survival essential. Critical thinking is where you pull information, ideas, and logic apart, and do some quality control on them. It’s also the sole method of thinking your way out of lousy information, processes, and even basic operational problems.

Much whining has gone into the issues with critical thinking. Kids aren’t trained to think critically, they say. That’s true enough. When you teach generations of kids to simply recite required information, there’s no real point of access into assessing information quality.

The information can be wrong, or more likely obsolete.  How are they supposed to know, let alone criticize? Are there academic awards for spotting garbage in the syllabus? Extra credits? The slightest acknowledgement of any kind?

While the post-truth situation and ineffectual critiques smugly festered for over a decade, AI entered the mix. Post-truth can’t manage its own failure. It can’t manage human reality in general. It has failed dismally to address AI’s glaring issues.

Critical thinking is grudgingly becoming the only way of addressing these problems. Let’s consider a few factors:

The mindless naïve acceptance of AI was universal in the corporate tech sector. Chatbots would solve everything. At this level there was no criticism, just sales pitches.

The almost total dismissal of chatbot AI in by the expert tech sector was equally universal. Nobody was, or is, even slightly impressed with a server-heavy multi-stage game of pass the parcel for every bit of real-time data. Nor are sycophantic Ais much use for anything at all.

The wheels fell off the “AI social ideology” immediately. Creating a new form of technology for the sole purpose of destroying jobs was never going to work and never will.

AI disasters snowballed, and they keep getting worse. The general and very uncharacteristic consensus among people in the sector who can’t stand each other is now that AI is going to cause unfixable problems without much more planning.

There’s a weird-looking dynamic here. All of AI’s problems were predicted to some degree. The necessary critiques came long after the predicted problems emerged. The risk levels have also snowballed when they should never have been possible.

Why? There’s no mystery here. Because the predictions contradicted the big money sales pitches, and for no other reason. Criticism, particularly correct expert critiques, threatened the gravy train. The Dumbest of the Dumb didn’t understand the issues, the critiques, or anything else.

AI to the rescue?

High schoolers and anyone using AI know that AI isn’t anything like perfect. Performance can be abysmal. It can cost far more than it’s worth just to solve the problems AI creates.

It’d be nice if institutional investors understood that, too. They have the clout and market reach to demand performance standards and enforce them. No quality standards can easily equate to no investment.

This is critical thinking applied in its more exponential form. You have to rip apart processes, AI alignment, AI dilution, AI training. and even the LLMs themselves if necessary to eliminate problems.

The most basic processes need critiques. Prompts need to be bulletproof. Productivity metrics must relate to clear success criteria. This is roughly the intellectual equivalent of deciding whether a nice breakfast cereal is more acceptable than wet cement.

It’s not that bad an analogy. AI is capable of producing unheard-of quantities of useless garbage at great expense.

Critical thinking is the definitive difference between success and failure.

There are no options.

Reinventing AI with critical thinking

AI provides a lot of raw material for critical correction. It spells out and defines its errors. At any point in any process, the failures become obvious, even if the reason for them isn’t.

There are multiple options for critical thinking in contest with AI:

Avoid unclear outcomes in prompts and testing environments. AI obviously must have unambiguous outcome requirements.

Build in critical cross-checks. This is already happening to some extent, but in this case the critiques need to relate directly to the outcomes. It’s better logic and can’t be misconstrued by the AI. Outcome X must be specified, and X must equal X.

Integration of AI outputs with outcomes must be audited. Forget “vibe coding”. Many people have. More advanced prompts are turning into novels to specifically to avoid those risks.

Don’t overrate human inputs. Any real critical assessment can’t just assume that a human input is flawless, either. The entire AI process needs clear evaluation standards.

If you’re thinking that this rather brief summarized overview looks very like the sort of character-by-character hard slog that went into digital tech, you’re quite right. It’s why digital tech works at all.

Early digital tech was very clunky. What we have now is the result of 40 years of fixes at the most basal levels. Every line of code had to be scrutinized, criticized, and workarounds found.

Critical thinking is exactly what AI needs now. Just do it.