Spinach is the best source of iron. I’ve understood this for decades now. As I ate more and more healthy foods, I’d choose spinach salads because of the “great” nutritional value, particularly iron.

With so much daily spinach, I figured I didn’t need to worry about iron supplements. Well, it turns out that spinach is not the best source of iron.

The Value of Spinach – other than iron.
Spinach is rich in iron and other nutrients like vitamin A, vitamin E and several antioxidants. It also includes a lot of beta carotene too. But its iron content is about equal to other vegetables. Spinach actually contains oxalic acid which actually inhibits the absorption of 90% of spinach’s iron content.

How did the notion of spinach as a great source of iron happen?
It turns out that the notion that spinach contained excessive amount of iron came from an 1870 study whose figures remained unchallenged until 1937. Then, it was discovered that the iron content was 1/10th of the claim. The error was due to a misplaced decimal point. This decimal point got us going in the wrong direction for decades. The problem is these errors are still happening today – we just don’t know it yet.

About five years ago, our foundation received a proposal from a well-recognized university. It included three, single spaced pages of formulas supporting their analytics. This proposal, along with the accompanying analytics, had been reviewed by the heads of relevant departments at the school before submission. We reviewed the proposal and the formulas (one of our team reviewers had a PhD in mathematics from Oxford), and it looked like something was off. We eventually figured out what it was: Their formulas had something misplaced that was imbedded. If the formulas as presented would have stood, and the analytics were conducted, the result would have had the opposite answer as that which the researchers were attempting to prove. Despite their review attempts, the error was imbedded and difficult to identify.

How does this relate to AI?
There’s a lesson here related to application of artificial intelligence, or AI. There’s the obvious question of if or how AI will edit out biases in the underlying data it absorbs, but a less-discussed question is will AI be able to flag imbedded formulistic errors like the flawed spinach study and the university proposal examples?

The instantaneous nature of the internet results in many of us blindly accepting results we find online and on social media. With AI, who is reviewing the formulas, and do they have the experiential backgrounds and then the guts to point out errors that arise? More importantly, are they ready and willing to get those errors corrected? Or will they merely be yelled down and silenced for their criticisms as is occurring elsewhere today.

It seems likely that some form of government oversight will be involved but government does not have the capacity or skill to handle the job. So, who will check the AI developers and check the AI checkers? It’s up to us. Question everything you read and see – which is what I personally do generally. This won’t solve the hard question facing AI technology applications, but it’s the safest strategy for us end users.

 

 

Featured news

Jayne Koskinas Ted Giovanis
Foundation for Health and Policy

PO Box 130
Highland, Maryland 20777

Media contact: 202.548.0133