Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Food and Generative AI

engineering.hellofresh.com

Post reply on HN
Print view View original

61–70 of 100 posts

Re: Food and Generative AI

#61

Post by TeMPOraL »

gffrd wrote: post #46
tornato7 wrote: post #20

Earlier quoted context omitted.

As someone with a few allergies, I effectively had to throw away and replace a third to half of the ingredients in every Hello Fresh meal. That's the #1 reason I stopped using it. An AI should be able to know what to substitute to make the whole kit edible for me.

Are you thinking "Rather than having a cook compose recipes with 1000 variations based on permutations of allergies, just produce one recipe with intended outcome, and let an AI figure out the variants"? How do you train an AI on taste? Would a chef state what's allowable for a recipe? "This one you can sub the soy out, but this one falls apart without it"

> How do you train an AI on taste?

You keep shoveling piles of recipes at it, and as long as most of the recipes are positively tasteful (as opposed to just randomly generated, or worse, engineered to suck out of spite), the AI should eventually pick up on taste in general.

Re: Food and Generative AI

#62

Post by TeMPOraL »

__MatrixMan__ wrote: post #55
TheHumanist wrote: post #38

Earlier quoted context omitted.

This is exactly how I feel. I want it to actually be helpful. It needs to be able to: - track what I have (if I have input that data) and how soon anything may expire - track what I made but haven't finished eating (just a few quick pictures in the fridge should update that as needed, or maybe it's time we all get that Alton Brown in-fridge camera to talk to) - can I use this in something else? How long should it sti…

As complex as it might be, the "intelligence" would probably be the easy part. Aggregating that data in a way that didn't become a burden for the user--either because its tedious, or because they don't trust you--would be quite a trick.

Then another trick will be surviving the attempts of stores and food manufactures to screw with this system as it becomes popular, because individuals empowered to make good choices for themselves hurt everyone's bottom line.

Re: Food and Generative AI

#63

Post by famouswaffles »

burkaman wrote: post #54
famouswaffles wrote: post #39

Earlier quoted context omitted.

>Actually generating a recipe is obviously much harder, and should be effectively impossible unless you include a humanoid robot that can cook and taste. It's not because this necessary information is already implicit in the sum of all the data it is trained on.

I don't agree, even a human chef with decades of experience in all the tools and ingredients they're using is not going to get every recipe right on the first try. I'm not saying an AI recipe will never be good, but it is going to make mistakes sometimes. If you're wondering whether a task is theoretically possible for an AI, I think a good rule of thumb is to ask "could a human domain expert do this given a few days…

You misunderstand what a language model is. What a human can or cannot do is frankly irrelevant.

A GPT (Generative pretrained Transformer) is not a simulator nor an imitator. It is a predictor.

Predictors try to reverse engineer the computation that could have led to a certain output so they don't just grok what is explicitly stated in a dataset but also what is implied by its structure.

As an example, In a language model trained on protein sequences alone and nothing else, you will find biological structure and function emerge in the inner layers.

https://www.pnas.org/doi/full/10.1073/pnas.2016239118

In fact, with a capable enough architecture and sufficient data (quantity, qualit, variance), there is nothing a predictor couldn't divine with enough training.

You assert

>Actually generating a recipe is obviously much harder, and should be effectively impossible unless you include a humanoid robot that can cook and taste.

In the pursuit of recipe prediction, a predictor will learn to internally model taste because taste is implicit in the data.

You don't need the hypothetical perfect predictor to demonstrate your assertion as false because a capable predictor already does.

>If you're wondering whether a task is theoretically possible for an AI, I think a good rule of thumb is to ask "could a human domain expert do this given a few days and as much research material as they need".

You are wrong. Train a language model on descriptions of the functions of proteins and an equivalent protein sequence and you get a language model that can predict novel functioning sequences from function descriptions alone.

https://www.nature.com/articles/s41587-022-01618-2

Not only is this not something a human expert can achieve in a couple days, It's not something a human expert can achieve at all. It is a Super-Human ability.

Re: Food and Generative AI

#64

Post by TeMPOraL »

fnordpiglet wrote: post #25
candiddevmike wrote: post #12

Earlier quoted context omitted.

I think something like the recipe generator that suggested mixing dangerous chemicals into cocktails is going to be a problem for any AI application in this space: https://arstechnica.com/information-technology/2023/08/ai-po...

It’s important to caveat here the language model was prompted into the suggestions explicitly, it’s not that it spontaneously suggested dangerous recipes. I’d not also that it’s getting harder to do that with current ChatGPT (this article uses GPT3.5), and I suspect “alignment” research in the 5 years time frame will make these sorts of things pretty hard to trick the models into doing.

That, plus:

> this article uses GPT3.5

Which alone disqualifies it from opining on what LLMs can or can't be used for.

Re: Food and Generative AI

#65

Post by TeMPOraL »

guhidalg wrote: post #51
jprete wrote: post #48

Earlier quoted context omitted.

Fake food images are literally fraud. I’m also having trouble believing that this is a passion project for friends and family based on your 90% comment.

When you search for recipes, how many recipes do you view before you decide to make one? It's a stretch to call it fraud when there's no guarantee what you make at home will look like what's made in a test kitchen. I challenge you to try to do food photography well at home without thousands in photography equipment, props, the actual dish, and expertise in making it look enticing on camera.

People absolutely expect the photos in a recipe to show the thing actually being made, and therefore an additional source of information. Adorning a recipe with photos of something similar but not quite like the described thing is, IMO, a form of gaslighting.

Re: Food and Generative AI

#66

Post by cschneid »

Xerox9213 wrote: post #29
tomjakubowski wrote: post #6

Earlier quoted context omitted.

I tried hello fresh. Unpacking a meal to find a fast food-sized mayonnaise packet as ingredient made me feel deeply sad

An aunt signed me up for a few free meals through some promotion. I tried it with an open mind but I found the entire experience infuriating. Most of the meals came with poor quality produce (e.g. tiny unripe avocado — something I never would have grabbed at the grocer) and a silly amount of waste for items I had already. My least favourite part was the recipes. I’m not sure if it was just the small selection of reci…

I recently had a 6 month stint using hello fresh, and now cancelled. I find it's really useful to get out of a rut, and teach you some small tricks. This recent one has me:

* Using much more citrus, juice & zest * Chickpeas + spice mix + touch of liquid in a pan is an easy go-to (we're veggie) * Broth / onions in rice. * Template of flatbread + stuff on top, which they had 3 or 4 in their rotation. Easy to riff on w/ whatever you have in the fridge.

None of these are earth-shattering revelations, but the repeated practice has ingrained some of these much more, and got me cooking a little differently.

We do this every few years, it seems worth it for a while, but I can't stick with it because I hate the waste, and the recipes get pretty repetitive after a bit, doubly so w/ vegetarian meals which have fewer options.

Re: Food and Generative AI

#67

Post by chefandy »

chefandy wrote: post #58

It's kind of funny-- I'm a technical visual artist, classically trained (former) chef, and spent a long time as a web developer in an environment where people breathlessly spoke about the possibility of LLMs. (Auto-correcting OCR was my pet wish that I never fulfilled.) It's been... interesting seeing layman's usually naive and mistaken takes on generative AI performing these functions. The difficult part of recipe d…

As an aside, now I really want sherry glacé'd radishes.

Re: Food and Generative AI

#68

Post by uoaei »

famouswaffles wrote: post #63
burkaman wrote: post #54

Earlier quoted context omitted.

I don't agree, even a human chef with decades of experience in all the tools and ingredients they're using is not going to get every recipe right on the first try. I'm not saying an AI recipe will never be good, but it is going to make mistakes sometimes. If you're wondering whether a task is theoretically possible for an AI, I think a good rule of thumb is to ask "could a human domain expert do this given a few days…

You misunderstand what a language model is. What a human can or cannot do is frankly irrelevant. A GPT (Generative pretrained Transformer) is not a simulator nor an imitator. It is a predictor . Predictors try to reverse engineer the computation that could have led to a certain output so they don't just grok what is explicitly stated in a dataset but also what is implied by its structure. As an example, In a language…

I don't think the commenter's understanding of LLMs is relevant here. It's what the authors of TFA think LLMs are for and, arguably more importantly, what kinds of problems (nails) they think about when presented with the idea of applying LLMs (hammer).

It's a bit ridiculous that this is the best they can come up with, and I don't really think the idea is worth defending, especially by telling other people that they just don't get it.

Re: Food and Generative AI

#69

Post by esafak »

Computer generated recipes and images would make me discount the service entirely, unless their tasters tried it first, in which case they could simply take a picture of the dish.

You can't use a language model to predict how something is going to taste.

Re: Food and Generative AI

#70

Post by smith7018 »

famouswaffles wrote: post #39
burkaman wrote: post #7

I feel like generating recipe ideas/titles is the easy part, and advanced AI techniques are not necessary. Even the article notes that there is not really any difference between 1 round of training and 40 rounds of training with their model. Actually generating a recipe is obviously much harder, and should be effectively impossible unless you include a humanoid robot that can cook and taste. I think the best you coul…

>Actually generating a recipe is obviously much harder, and should be effectively impossible unless you include a humanoid robot that can cook and taste. It's not because this necessary information is already implicit in the sum of all the data it is trained on.

My friend has a restrictive diet for health reasons (can't eat certain greens, no gluten, etc) and decided to use ChatGPT to come up with a dessert for her dinner party. It was honestly one of the worst things I've ever tried. She verified that the recipe wasn't regurgitated from another source so it was all ChatGPT's doing.
Post reply on HN