Food scanners like TellSpec.

[BrianHG]

People believe these devices are actually telling them whats in their food:

[Cyberdragon]

Ah yes, so they have invented a flouroscope/emission spectrum analyzer into a palm sized device cheap enough for the average consumer?

"I believe you..."
*slam!*
"You see these quotations I'm making with my claw hands?! It means I DON'T believe you!"

[Raj]

Tell someone,you can increase their lifespan. They'll be ready to consume mercury

[ebastler]

Ah yes, so they have invented a flouroscope/emission spectrum analyzer into a palm sized device cheap enough for the average consumer?

"I believe you..."
*slam!*
"You see these quotations I'm making with my claw hands?! It means I DON'T believe you!"

Nope, you got that wrong. No fluorescence involved here; this is just plain reflectance/absorption spectroscopy. And that means that (a) the signal is several orders of magnitude stronger than in fluorescence, and (b) it does not have to be pulled out of a much higher background of excitation light.

Also, calling this a "spectrum analyzer" may be a bit misleading for the electronically minded on this forum. We are talking about a prism or an optical grating here. (Think "stamped CD pattern"...)

Hence, a portable IR spectroscope for food analysis, built from consumer-grade components, does not seem implausible. I have not watched the video yet, but will if I have to.   What is their key concern?

EDIT: Here's a nice (although a bit dated) overview of the application space and some technical approaches: https://www.hackteria.org/wiki/images/0/03/NIR_food_analysis.pdf

[Cyberdragon]

You can't, food sucks real bad at IR analysis (as demonstrated). They are claiming you get the whole nutrition facts from any food instantly, and can even identify medicine with a terrible resolution and not to mention 0 surface penetration.

[ebastler]

You can't, food sucks real bad at IR analysis (as demonstrated). They are claiming you get the whole nutrition facts from any food instantly, and can even identify medicine with a terrible resolution and not to mention 0 surface penetration.

... says the guy who thinks this works via fluorescence detection. 

Sorry, I would need some references. I gave one above; what do you base your assessment on?

[BrianHG]

You can't, food sucks real bad at IR analysis (as demonstrated). They are claiming you get the whole nutrition facts from any food instantly, and can even identify medicine with a terrible resolution and not to mention 0 surface penetration.

... says the guy who thinks this works via fluorescence detection. 

Sorry, I would need some references. I gave one above; what do you base your assessment on?

Watch the full video at least at 2x speed before assuming anything.  The technology is specified for both and one of the cheaper brand units were already disassembled and analyzed by a third party.  They are garbage and they are not capable of doing what they are advertised to actually do.  It's full BS and if they instead sold nothing but a visible color light meter box with the same AI engine, the results and capabilities would actually improve.

[ebastler]

Watch the full video at least at 2x speed before assuming anything.  The technology is specified for both and one of the cheaper brand units were already disassembled and analyzed by a third party.  They are garbage and they are not capable of doing what they are advertised to actually do.  It's full BS and if they instead sold nothing but a visible color light meter box with the same AI engine, the results and capabilities would actually improve.

OK, I had time to watch the full video now. Clearly the two low-cost providers are massively overselling their gadgets. But -- as acknowledged by Thunderfoot towards the end of his video, and discussed in the overview paper I linked to above -- NIR spectroscopy for food analysis is an established technology to analyze e.g. grain and flour, meat, fish, milk, and fruit. Not for a full analyis of material composition, of course, but for specific key analytes in each of the applications.

Keeping that key limitation in mind, and lowering the accuracy/resolution requirements a bit, I don't see why a consumer-level device should not be feasible. Maybe SCiO's version is a bit too cheaply built, and certainly they are overselling it; but the next entry to the market might get it right.

In fact, I have been involved with a company which produces professional-grade NIR food analyzers. They evaluated the SCiO, wondering whether there might be a real threat emerging for their business -- and found that it did indeed work for one of their major food testing appliciations. Not as good as their own, high grade equipment; but certainly not "garbage".

[dferyance]

This feels to me like just another entry in the AI / machine learning / cloud fad. This is a difficult problem...oh well lets throw the cloud and AI at it, that'll solve it! There is a lot that these systems can solve, but that doesn't mean they solve *all* problems.

From their site:

The combination of machine learning, bioinformatics techniques and traditional spectroscopy gives Tellspec the ability to extract nutritional information from a spectrum, the unique fingerprint of the food.

I think the idea is, throw a crappy sensor out there and hope that machine learning can make up for it. It might even work in some cases, but no way it will be reliable to any degree.

[BrianHG]

This feels to me like just another entry in the AI / machine learning / cloud fad. This is a difficult problem...oh well lets throw the cloud and AI at it, that'll solve it! There is a lot that these systems can solve, but that doesn't mean they solve *all* problems.

From their site:

The combination of machine learning, bioinformatics techniques and traditional spectroscopy gives Tellspec the ability to extract nutritional information from a spectrum, the unique fingerprint of the food.

I think the idea is, throw a crappy sensor out there and hope that machine learning can make up for it. It might even work in some cases, but no way it will be reliable to any degree.

REMEMBER: The manufacturer 'knows' that not a single human being will scan their food, then send that entire item to a lab to be tested to verify the scan's accuracy.  So, whatever the scanner results are, especially when you have to tell it what type of food item you are scanning, it make no difference at all.  Which is why I say the results a pure BS.

[Mr. Scram]

Although the introduction veiled it a little bit it turns out to be another Thunderf00t rant video. Maybe he should change his name to The King of Ranting.

[BrianHG]

BUSTED $3 MILLION Food scanner RESPONDS!

[Mr. Scram]

10 REASONS to AVOID videos with TITLES LIKE THESE!

[mdszy]

10 REASONS to AVOID videos with TITLES LIKE THESE!

YOU WONT BELIEVE what I have to DO TO MY TITLES to get YOUTUBE VIEWS (100% REAL) {NOT CLICKBAIT} [GONE ELECTRICAL]

[electronics man]

as a chemist this makes me want to cry

[Cyberdragon]

as a chemist this makes me want to cry

Don't worry, no chemists were harmed (or used at all) during the making of this bullshit.

[ebastler]

Don't worry, no chemists were harmed (or used at all) during the making of this bullshit.

Right, the CTO behind Scio is an electronics engineer.   
https://www.consumerphysics.com/business/about-consumer-physics/

[GeorgeOfTheJungle]

Imagine if a tester instead of having a -say- 5% error in the readings 100% of the time, gave good readings 99% of the time but misbehaved badly the remaining 1%. It would be totally useless, wouldn't it? That's the problem with things like this: you can never know if the reading is about right or completely wrong.

[Cyberdragon]

Imagine if a tester instead of having a -say- 5% error in the readings 100% of the time, gave good readings 99% of the time but misbehaved badly the remaining 1%. It would be totally useless, wouldn't it? That's the problem with things like this: you can never know if the reading is about right or completely wrong.

Consumer food is not an exact science anyway, so the whole concept is stupid right outside the box...

[Mr. Scram]

Biology is never an exact science. Caloric values can be and are fairly useful, as long as you don't make them into religion.

[GeorgeOfTheJungle]

Consumer food is not an exact science anyway, so the whole concept is stupid right outside the box...

I was watching this the other day:

The Royal Institution: The Physics of Hot Air - with Shaun Fitzgerald
https://youtu.be/zXPYPkgd-JQ?t=11m40s

It says a person is a 100W heater, and in the video you've posted it says one needs 2000 calories a day. Funnily enough, the figures match, look:

2000 calories is 2000 [Calories] * 4184 [Joules/Calorie] = 8.36 MJ
100 [Watt = Joules/second] * (3600 * 24 [seconds]) = 8.64 MJ