How blind faith can put your health and wellness at risk
to you
How great does it feel to ask an important question and instantly get an answer you can trust?
But how do you know your trust is justified?
To jokingly (but half seriously) challenge people for believing things they see on social media without even the slightest bit of critical thinking, I often sarcastically say, “I saw it on Facebook, so it must be true.”
Yes, perhaps a bit annoying on my part, but as someone formally trained in and fully committed to seeking truth (a.k.a., science), I can’t help myself.
Now, as AI continues rising in popularity, I may have to adjust my sarcasm to, “AI told me, so it must be true.”
AI is an incredibly innovative resource for finding information and formulating ideas. But in many cases, it’s just a gateway and is far from the final word.
Remarkably, it can understand nuanced questions and organize and interpret information from a variety of sources into a concise answer, something that would otherwise require multiple web searches and potentially hours of reading and sorting through the results.
And the responses from AI are impressively articulate. Impressive enough that it’s very easy to accept what AI says as fact.
Despite that, you’ve likely heard some of the stories suggesting otherwise, like this case of an official government health report including citations to research studies that don’t exist. (Science.org, the home of this article, is the website of one of the most respected and prestigious research journals in the world.)
If you’ve been burned by AI, don’t worry, our government apparently has been, too.
I’ve seen a lot of AI mistakes firsthand.
But that’s OK, at least for me, because the great thing about AI is that I can interrogate it ruthlessly without offending it, which I fully take advantage of.
Staying true to my scientist roots, I ask for citations and read through them to verify and better understand the claims that AI generates.
And I keep asking questions about the reasoning it uses to justify those claims.
No doubt, it’s astonishing that AI can string together a series of related ideas and form a conclusion based on them. This, after all, is what puts the “I” in “AI” (not to be confused with A1 steak sauce).
But the more I challenge this process, the more I find myself backing AI into a corner, getting it to admit that it’s contradicting itself. (Although, it’s definitely a good thing that it’s willing to fess up!)
Here’s a recent example: I asked AI for the full citation of a study it cited to support a claim. The citation provided was not only for a different study, but also contradicted the original claim!
AI's response to me pointing this out:
Are we really sure it’s the truth this time???
Here’s the context: I was looking to evaluate the safety of the hip thrust, which is a new-ish exercise often used as a back-friendlier alternative to squats and deadlifts for training hip extension.
I wanted to get to the bottom of how confidently I could conclude that the hip thrust is indeed more back-friendly, and by how much.
I was specifically hoping to find data on the amount of shear force the hip thrust imposes on the lumbar spine.
Unfortunately, neither I nor AI found any such data. Despite this, AI made a pretty big leap in telling me that the shear load is “minimal”.
And it wasn’t until I pressed further with questions that it became obvious this claim was based on indirect evidence (making it a weak claim despite being stated firmly).
Then I asked for the citation to support that indirect evidence, and when I did, AI completely changed the claim, with the explanation for the “inconsistency” shown above.
AI said: “In other words, the evidence is mixed, and I should not have stated unequivocally that erector spinae activation is lower. That was an overstatement.” (Bold and italics as displayed by AI)
An overstatement, indeed! And it significantly weakens the overarching claim that I fortunately didn’t assume was valid. But how many people are going to believe AI without doing the interrogation that I did?
Adding insult to injury, this particular AI model went out of its way to tell me that it was intentionally erroring on the side of safety.
I like that – I like it a lot – but only if it’s what’s really happening!
There’s a good amount of anecdotal indication that the hip thrust is indeed friendlier to the low back than squats and deadlifts. But this particular AI model, in it’s own words, “overstated” this indication, making it seem far closer to fact than it really is.
This inappropriately makes the exercise in question seem safer than it might really be.
That’s the opposite of erroring on the side of safety.
The principles we teach in our one-on-one coaching are grounded in science, and refined with practical experience to fill in the gaps and keep our recommendations realistic and sustainable – all with an emphasis on what matters most for you.
Add to that Coach Kayla’s compassionate dedication to keeping you focused on what matters most and helping you overcome obstacles.
The only remaining ingredient needed for a better future is your desire for long-term success!
Schedule a free call with Coach Kayla to learn how we can help you build a rock-solid foundation for getting more out of life!
But this isn’t about nitpicking AI’s shortcomings – AI is an amazing tool, when used appropriately.
When I say appropriately, I mean applying critical thinking. Critical thinking isn’t only important for using AI, it’s important for evaluating all of the information you take in.
Sure, it’s a lot harder to evaluate information without a science background, but that doesn’t make it OK to accept everything you hear as fact.
Be skeptical and ask questions – the health and wellness of your future self may depend on it!
To finding truth,
Dr. Vin