I said AI can’t taste cheese. And I stand by that (Mostly)!
I say, mostly because there are people out there building electronic tongues and electronic noses. Which sounds less like food science and more like something Wallace and Gromit would accidentally invent while trying to improve breakfast. But they do exist.
Electronic tongues and noses are being used in food research to measure things, compare samples, spot patterns and support quality assessment. One review describes electronic noses and tongues as sensor-based tools that simulate aspects of the human nose or tongue, and combines them with pattern-recognition methods to assess food quality.
So yes, machines can measure some things linked to taste and smell, but that’s not the same as tasting cheese. And it’s definitely not the same as fancying cheese.
What an electronic tongue actually does
An electronic tongue does not sit there thoughtfully with a cracker and a glass of something nice. It uses sensors to measure responses from a sample. Then those responses can be analysed to look for patterns, differences or similarities.
That can be useful, I’m not dismissing it. It might help with research, quality control, product development, consistency or early warning signs that something has changed. But before anyone starts picturing a robot at the counter saying, “I’ll take 200g of the funky blue one, and don’t wrap it too tightly,” we should probably calm down.
This is not AI tasting cheese in the way you or I taste cheese. It’s measuring, it’s comparing, it’s classifying and it’s looking for patterns.
The cheese experiments
This is where it gets interesting (and a little bit odd). There are already cheese-related studies using electronic tongues and electronic noses. Not because the machines are sitting there having a lovely time, but because they can sometimes detect differences that are useful in research or production. For example:
- Electronic Tongue and Consumer Sensory Evaluation of Spicy Paneer Cheese looked at spicy paneer cheese with different levels of capsaicin, the compound that gives chilli its heat. The electronic tongue was used to discriminate between paneer samples with different capsaicin levels and compare that with consumer sensory evaluation. So, in a very specific sense, a machine was being asked to tell different spicy cheeses apart.
- Cheeses Made from Raw and Pasteurized Cow’s Milk Analysed by an Electronic Nose and an Electronic Tongue used an electronic nose and electronic tongue to analyse cheeses made from raw and pasteurised cow’s milk. The electronic tongue distinguished cheeses made from raw and pasteurised milk through whey and curd analysis, while the electronic nose was more sensitive to ripening differences.
- Influence of Storage Time and Elevated Ripening Temperature on the Chemical and Sensory Properties of White Cheddar Cheese looked at Cheddar during maturation. The study reported that the electronic tongue could classify samples by ageing month with a validity value of 92.59%.
- Discriminating Aging and Protein-to-Fat Ratio in Cheddar Cheese Using Sensory Analysis and a Potentiometric Electronic Tongue looked at ageing and protein-to-fat ratio, using descriptive sensory analysis, consumer panels and electronic tongue measurements. The electronic tongue distinguished changes linked to formulation and ageing time, and the authors suggested it may be useful for monitoring taste quality during ageing.
- Comparison of Cheese Aroma Intensity Measured Using an Electronic Nose Non-Destructively with the Aroma Intensity Scores of a Sensory Evaluation: A Pilot Study compared electronic nose readings with sensory evaluation scores for cheese aroma intensity. And because cheese is never just about the tongue, this feels like a useful reminder that smell is doing a lot of the work too.
So, fair enough, the robots have tongues now (sort of). They also have noses (sort of).
Interesting? Yes. Tasting? No.
This is where I think the distinction matters.
- A machine might be able to tell that something has changed.
- It might be able to compare one sample with another.
- It might be able to support research, quality control or product development.
- It might even detect something earlier or more consistently than a human panel in a very specific setting.
But it still doesn’t taste cheese.
- It doesn’t fancy cheese.
- It doesn’t have memories attached to cheese.
- It doesn’t know whether a customer is ready for the really funky one.
- It doesn’t understand the difference between “technically mature” and “absolutely singing today”.
- It doesn’t know that someone who says “nothing too strong” might still be gently persuadable.
- It doesn’t know when to explain and when to shut up and let the cheese do the talking.
And for me, that’s the real point.
Where I land on this
I should probably say here, I’m not anti-AI, quite the opposite. I’m quite geeky about AI, I’m geeky about technology, and, as you may have noticed, I’m also quite geeky about cheese.
So yes, I’m interested in what electronic tongues can pick up. I’m interested in what electronic noses can detect. I’m interested in where AI, sensors and food science might help with research, consistency, safety, quality control or product development.
For larger manufacturers, I can absolutely imagine use cases. If you’re producing at serious scale, and you need consistency across huge volumes, then tools that help spot changes, compare samples or flag issues could be valuable.
But when I think about artisan cheese, I find myself hoping the future doesn’t become too tidy. I can’t imagine a time when I’d want artisan makers to rely on AI to mature or develop cheese.
Maybe that sounds a bit romantic. But I’d be disappointed if the direction of travel was less judgement, less instinct, less patience and less human feel – those are the stories that make good marketing (and great cheese)
The thing I love about artisan cheese is not that it’s perfectly predictable. It’s that it carries the hand, care, judgement and personality of the people who make it. It changes, it responds, it has better days and stranger days and it asks for experience, not just measurement.
Camembert in the luggage compartment
This is where AI runs out of road for me.
AI will never have gone to France on a school trip and tasted soft cheese properly for the first time.
It will never have had that moment of realising that cheese could be something completely different from the mass produced Red Leicester in your packed lunch.
It will never have ignored the sensible souvenir options, walked past the Eiffel Tower snow globes everyone else was buying, gone to a cheese market instead, and proudly brought home a fresh Camembert.
It will also never have learnt, the hard way, that 24 hours in a school coach luggage compartment is not necessarily the ideal maturation environment for a soft cheese. Especially if you were hoping it would be received as a thoughtful gift rather than a small biological incident.
That’s cheese. It gets attached to memories, first tastes, bad decisions, good surprises, smells you remember for decades, markets you stumbled into, the makers you met and the meals you still think about together with the people who introduced you to something new – A machine can’t have that.
And those lived feelings and experiences are exactly why people buy cheese, love cheese, talk about cheese, gift cheese, seek it out, argue about it, build careers around it, and occasionally spend more than they planned because someone gave them a taste of something they didn’t know they wanted.
So what does this mean for the AI Lab?
For the Cheese Hub AI Lab, this is not a route into “let’s get AI to taste cheese”. The point is to explore where AI can help around the cheese.
- Can it help makers explain what they do more clearly?
- Can it help mongers turn everyday customer questions into useful content?
- Can it help suppliers organise knowledge and support training?
- Can it help busy cheese businesses write, plan, sort, promote and reuse their ideas without making everything sound like it came from the same bland machine?
That’s where I think AI gets interesting, not when it tries to replace taste but when it helps real cheese knowledge travel further.
So yes, apparently the robots have tongues now. Sort of. But they still don’t taste cheese, they don’t fancy cheese, and they definitely don’t have a slightly traumatic school trip Camembert story – which, for now, feels like a win for the humans.
