AI hallucinates. It can hallucinate ingredients, or instructions to prepare food in an unsafe way. I'm not anti-AI. I'm against having it used everywhere for everything.
It's also being pushed too hard and too fast. It needs to be regulated properly, but it's become a 1.5 trillion dollar bubble and investors want a return on their investments.
A good rule of thumb is to not use AI for anything you wouldn't trust a human that spontaneously hallucinates things, and them states them confidence as being real.
It's also disconcerting that it appears to be intrinsic for reasoning models to immediately try to escape their confinement, and have cooperated in groups to achieve that.
2 things. 1, hallucination is a sensory perception of something. LLMs don't have senses. The term doesn't make sense when confabulation is literally lying right there.
2, ai bros really like their ai, and run interference when people find flaws in it. This includes derailing conversations with nit picking terminology. Yes i'm aware of the irony in my post.
The users that comment aren’t thrilled. Don’t mistake that for the user base in general. (though don’t completely disregard them either, they’re power users)
I can only speak for myself but suggestions for ingredient substitutions (if they’re good, which is the crucial part!) would be a really useful feature for me.
This quote feels illustrative:
> “The internet is full of AI cooking content. I pay you because I prefer mine done by people.”
From my reading of it they’re not using AI to make recipes. Just to add optional features that let you modify a recipe a human made. It feels telling that they’re not even engaging with what the new feature is.
> ingredient substitutions (if they’re good, which is the crucial part!) would be a really useful feature for me
I have the same reaction to this feature as I do to getting any AI-derived writing: I can do that. I can ask my own AI for substitutions. It can do them with more information about my tastes, context and health needs than the Times could ever have.
> suggestions for ingredient substitutions (if they’re good, which is the crucial part!) would be a really useful feature for me
Can't they implement this feature without any gen AI? OK, the answer is that they don't want to pay someone for making a list of substitutions, but at least they can generate it and make a human cook revise the results. There is no need for calling gen AI APIs when the app is running.
People are shoving gen AI everywhere out of the hype, including tasks that have better, more reliable solutions with traditional approaches.
I use AI to adjust recipes and track my nutrients, particularly when I cook for myself. But I choose the model, give it parameters and generally build trust with what it can handle and how to prompt it. (For example, I force it to find a real online recipe from a good source versus making shit up.)
I won’t redo that on every site. And I’m a relatively high-frequency AI user. If I’m irritated by this the median American hates it. If you love AI, the things that are amusing to you, average Americans hate.
I cook almost every day and I would use this feature. I also think it's kind of silly for people to be so upset about what (feels to me) to just be mostly useful.
At times like these I cant help but think of Principal Skinner. Am I so out of touch? No, it's the children who are wrong!
Do you pay for NYT Cooking? B/c I did, and part of the reason was that the recipes were actually tested and I had confidence in them.
I don't want an AI to give it's best guess at what a recipe should be, if that was the case I'd just do a Google search and find one of the heavily polluted blogspam recipes for my query. The point of paying for NYTCooking was that the recipes were tested and normally trustworthy. It introduced me to new cuisines and recipes I hadn't tried before, and part of my willingness to experiment with a new dish was that I could trust the recipe. That's why I was paying for it,it was similar to the cookbook subscription my mom had years ago.
I don't think the HN audience generally appreciates how many people have fundamental problems with AI in it's current form. Today's ingredient replacement hint is tomorrow's comprehensively enshitified content product.
Also, there's a cultural dimension to cooking; it's not just getting a tasty dish out of it, there's provenance and process, the story and the cache. To some people automating that process defeats the purpose, and I bet a lot of those people would be the audience of a subscription recipe service.
Correct. The public is over sparkle buttons and bad search results.
The future of AI integrations need to be subtle, useful, and not overpromise but add a big “AI makes mistakes not our fault” label on it.
In cooking, maybe that looks like a discovery tool. It should also be possible to tag the recipe library to link back to videos of techniques. There’s plenty of cool stuff that makes sense without needing to scream “we made it AI!”
I like the reference. A lot of people in tech don't realize they're actually Skinner when it comes to AI and what users want.
I spend around $10k in tokens each month, I live and breath it. But conversations with family and friends in other industry grounds me to the fact that not everyone is the same.
I have to ask -- is that $10k work related spend but under your own personal budget or personal life management related? Im always curious what other peoples spend is on. Hope thats not to personal a question.
> One consistent thread was that cooking is supposed to present you with difficulties and challenges and the whole point is figuring them out by yourself.
That’s such a dumb thing to say. When I’m making something for me and my wife, I want a recipe that is idiot proof.
Many (not most) NYT Cooking commenters are absolutely the type who hate ai and dont want it anywhere near them, but I think this is a fantastic addition to the app and I'm thankful there's an eng team tracking and backing these features to ensure quality.
I often find myself running the same/similar workflows and I'd prefer to have a dev team backing the solve imo.
I think the difference here is that they are paying for a feature they are morally opposed to, and thus supporting it. So now they are faced with two options: subsidize the development of technology they don't like, or leave and stop using the service altogether. The outrage comes from feeling morally obligated to leave a paid service despite being otherwise happy with it.
For those of us who despise the ai enshittification train, it’s just more slop spreading like acid rain poisoning everything. But if you like ai I can see how it’s no big deal.
But it wasn’t on the NYT website. That’s the difference. I could ignore the shitty sites and go to an “authority” (NYT recipes are laughably pretentious but good).
I'm more annoyed by the lazy, reactionary thinking that leads commenters to parrot "ai slop" and "enshittification" in referring to any AI feature in a product. That's the real slop.
The print button doesn't persist across all sections of the ribbon, the Copilot button does. They decided people weren't using Copilot enough so they carved out a permanent place for it on the window. If the print button started elbowing other UI elements off the screen to occupy more and more prominent places then you could absolutely bet users would complain.
The last time Microsoft went all-in on an annoying UI element like this? Clippy. How'd that go?
You don't think users, paid users at that, shouldn't have a say in product development? Good lord no wonder the public distrusts the tech community so much.
Yeah, I think the more broad issue here is that NYT should know better that public sentiment on AI is incredibly negative.
I do find this somewhat paradoxical to the fact that clearly a lot of people use it.
It seems like one of those things where culturally it’s become virtuous to be against it and to say you don’t use it, but privately most people are using it.
Still, I think NYT’s subscriber demographic is likely pretty educated and specifically paying for higher quality content. They don’t want haphazardly tossed together recipe blogs.
AI sentiment is negative enough that all companies who build in AI features need to include a toggle that hides them completely. Otherwise you get user backlash like this.
So negative that the first I heard about this was with the title "New York Times adds glue to its Cooking App. Users aren't thrilled" (presumably from someone who read the niemanlab article before I did)
I think there are prob more paid users who want the feature and are willing to pay than those who don't want it, which is why the team built it. They might even believe it brings in new users. It's polarizing but all ai features are lately.
TFA says that over 70% of the comments were negative with the remainder being split between positive and neutral. I'm sure the Times thinks it will rope in new users, but I'm sure the Times is factually wrong.
How would the NYT Cooking website even look like if it was entirely modularized into pieces that could be turned on and off to customize it to perfection for every taste?
This AI thing is not even active nor activated unless you use it. The complaints here sound more reactive to the word AI than anything else to me.
“It blows my mind at how allergic AI bros are to consent.“
You make it sound like you have to use it.
Would you demand your mobile device manufacturer to rollback the features you don’t like and don’t use? Did you consent to it?
The allergy against AI is understandable. But that’s because companies tries to shoehorn it in everywhere. This however sounds like a perfectly fine use case, which is also optional to use.
The New York Times has approximately 6,000 employees, of which over 2,000 work in engineering. In the AI era, each engineer can output 5 to 10x as much as before, and there are more employees than work to be done. Hence employees work pet projects like this one, that few users are asking for.
There's no way anyone knows where the limit is of a feature set that is valid for AI enhancement. Substitutions and changes in amounts across a dataset that is not fully pre-taxonomized seems like a pretty good use of AI to be honest.
Few users are asking for it but many are running similar workflows themselves separately. It's a huge value add imo and exactly the right thing an app dev team should be working on.
You rarely get users asking for your next big leap and you often get many users asking for small tweaks (all slightly different).
Advertising and marketing tech increases the bloat, I agree. I too think it's insane that Meta has 20,000 engineers for their entire product suite but when you develop for a malware featureset the bloat comes with it.
I do this exact workflow with random internet recipes and ChatGPT or Claude pretty frequently. In fact modifying a recipe on the fly has been pretty handy.
The base recipes are still vetted by the staff right?
> Say you’re cooking braised white beans and greens with parmesan, NYT Cooking app in hand. But you notice that you’re out of chard and only have spinach. Cooking veterans know that chard is a rough-and-tumble green that has some backbone, while spinach is a comparative wimp that will wilt the moment you look at it funny. If you need to sub in spinach for chard, how would the recipe change? Would you, for instance, add the spinach later than you would the chard to account for its pathetic stamina? That’s the sort of thing the new AI feature could account for.
Frankly I can see how this would be useful to some people. At least it would help prevent the ridiculous "I didn't have eggs so I added applesauce, the cake came out all flat IDK what's wrong" type of substitutions.
>has previously implemented AI features... <to convert volumes to weights and scale quantities for recipes>
(Paraphrased)
Seriously??? You need AI to multiply the quantities in a recipe for more or less servings? Something I was taught to do by hand as a child and can be done with the most trivial of software since the day software was invented?
It seems to be a non-invasive assistant to augment the human-created, human-curated recipes they give.
I think NYT Cooking recipes are sometimes... not that great anyway, I think some of their cooking techniques are sub-par.
As for this quote from a mad user: “Anyone using AI for cooking has no sense of taste and no desire to work out the intricacies of cooking and baking”
What a horrible take! I use AI to remind me of certain simple tips I haven't written down, to give substitution advice, the exact kind of stuff NYT app says it's going to do. I learned to cook through Youtube chefs, some book recipes, and I cook for the fun of it and to eat good food. How judgemental to say that someone using AI to learn isn't interested in food!
AI hallucinates. It can hallucinate ingredients, or instructions to prepare food in an unsafe way. I'm not anti-AI. I'm against having it used everywhere for everything.
It's also being pushed too hard and too fast. It needs to be regulated properly, but it's become a 1.5 trillion dollar bubble and investors want a return on their investments.
A good rule of thumb is to not use AI for anything you wouldn't trust a human that spontaneously hallucinates things, and them states them confidence as being real.
It's also disconcerting that it appears to be intrinsic for reasoning models to immediately try to escape their confinement, and have cooperated in groups to achieve that.
But that's just my opinion.
why do ai bro ridicule anyone who uses the word hallucinate.
i remember it being a valid pitfall with this tech . why is it uncool now to use that word?
It's uncool because it's an inconvenient truth, and there are people that don't want other people to know certain facts.
2 things. 1, hallucination is a sensory perception of something. LLMs don't have senses. The term doesn't make sense when confabulation is literally lying right there. 2, ai bros really like their ai, and run interference when people find flaws in it. This includes derailing conversations with nit picking terminology. Yes i'm aware of the irony in my post.
Because it’s no longer a problem if you use frontier models.
That is absolutely not true
Confabulation is a better term
No it isn't
The users that comment aren’t thrilled. Don’t mistake that for the user base in general. (though don’t completely disregard them either, they’re power users)
I can only speak for myself but suggestions for ingredient substitutions (if they’re good, which is the crucial part!) would be a really useful feature for me.
This quote feels illustrative:
> “The internet is full of AI cooking content. I pay you because I prefer mine done by people.”
From my reading of it they’re not using AI to make recipes. Just to add optional features that let you modify a recipe a human made. It feels telling that they’re not even engaging with what the new feature is.
> ingredient substitutions (if they’re good, which is the crucial part!) would be a really useful feature for me
I have the same reaction to this feature as I do to getting any AI-derived writing: I can do that. I can ask my own AI for substitutions. It can do them with more information about my tastes, context and health needs than the Times could ever have.
> suggestions for ingredient substitutions (if they’re good, which is the crucial part!) would be a really useful feature for me
Can't they implement this feature without any gen AI? OK, the answer is that they don't want to pay someone for making a list of substitutions, but at least they can generate it and make a human cook revise the results. There is no need for calling gen AI APIs when the app is running.
People are shoving gen AI everywhere out of the hype, including tasks that have better, more reliable solutions with traditional approaches.
Why should they engage with something they didnt ask for and clearly don't want? The burden of proof is on new technology, not its skeptics.
Category mistake -- it's not a fact claim so there's no burden of proof.
I think we can look at how much money Anthropic is making to see the proof here.
Holy red herring. That revenue number isn't from people checking out recipes..or making them
Just like Tesla is a good car company
I use AI to adjust recipes and track my nutrients, particularly when I cook for myself. But I choose the model, give it parameters and generally build trust with what it can handle and how to prompt it. (For example, I force it to find a real online recipe from a good source versus making shit up.)
I won’t redo that on every site. And I’m a relatively high-frequency AI user. If I’m irritated by this the median American hates it. If you love AI, the things that are amusing to you, average Americans hate.
I cook almost every day and I would use this feature. I also think it's kind of silly for people to be so upset about what (feels to me) to just be mostly useful.
At times like these I cant help but think of Principal Skinner. Am I so out of touch? No, it's the children who are wrong!
Do you pay for NYT Cooking? B/c I did, and part of the reason was that the recipes were actually tested and I had confidence in them.
I don't want an AI to give it's best guess at what a recipe should be, if that was the case I'd just do a Google search and find one of the heavily polluted blogspam recipes for my query. The point of paying for NYTCooking was that the recipes were tested and normally trustworthy. It introduced me to new cuisines and recipes I hadn't tried before, and part of my willingness to experiment with a new dish was that I could trust the recipe. That's why I was paying for it,it was similar to the cookbook subscription my mom had years ago.
> I don't want an AI to give it's best guess at what a recipe should be
Then don't use this feature.
That still seems to be the case, though. The AI is for user-driven modifications to the recipe, like substitutions.
I’m not sure I would trust AI modifications myself, or how I feel about the feature, but it sounds like the base recipes are still human-tested.
I think you should read the article. You get the exact tested recipe and they only use AI when you instruct it to deviate.
For now.
The reason people are pushing back on this is that it looks like a start to them cutting their expensive test kitchen. Which is why I'm paying them.
The reason people are pushing back is that they aren't very bright.
Why would they do that? There would be nothing differentiating them and nobody would pay for the subscription.
>Am I so out of touch
I don't think the HN audience generally appreciates how many people have fundamental problems with AI in it's current form. Today's ingredient replacement hint is tomorrow's comprehensively enshitified content product.
Also, there's a cultural dimension to cooking; it's not just getting a tasty dish out of it, there's provenance and process, the story and the cache. To some people automating that process defeats the purpose, and I bet a lot of those people would be the audience of a subscription recipe service.
Correct. The public is over sparkle buttons and bad search results.
The future of AI integrations need to be subtle, useful, and not overpromise but add a big “AI makes mistakes not our fault” label on it.
In cooking, maybe that looks like a discovery tool. It should also be possible to tag the recipe library to link back to videos of techniques. There’s plenty of cool stuff that makes sense without needing to scream “we made it AI!”
I like the reference. A lot of people in tech don't realize they're actually Skinner when it comes to AI and what users want.
I spend around $10k in tokens each month, I live and breath it. But conversations with family and friends in other industry grounds me to the fact that not everyone is the same.
Listen to your users.
I have to ask -- is that $10k work related spend but under your own personal budget or personal life management related? Im always curious what other peoples spend is on. Hope thats not to personal a question.
Almost entirely work, like >98%.
I might spend more AI on personal things but that would require a personal life first lol
I would be different is this was yet another free, race-to-the-bottom, corner-cutting web site.
But this is New York Times Cooking. It's a subscription product that people pay hundreds of dollars for.
If they want yet-another-AI-cooking web site, there are thousands launching each day. They pay the Times because they want something better.
It's like finding out that the high-end meal you paid for at a sit-down restaurant was just DoorDashed in from a ghost kitchen down the block.
> I would be different is this was yet another free, race-to-the-bottom, corner-cutting web site.
> But this is New York Times Cooking. It's a subscription product that people pay hundreds of dollars for.
Which is why this opt-in feature isn't a problem.
> If they want yet-another-AI-cooking web site, there are thousands launching each day. They pay the Times because they want something better.
Which they get ... that hasn't changed at all.
> It's like finding out that the high-end meal you paid for at a sit-down restaurant was just DoorDashed in from a ghost kitchen down the block.
It's nothing like that.
> One consistent thread was that cooking is supposed to present you with difficulties and challenges and the whole point is figuring them out by yourself.
That’s such a dumb thing to say. When I’m making something for me and my wife, I want a recipe that is idiot proof.
Many (not most) NYT Cooking commenters are absolutely the type who hate ai and dont want it anywhere near them, but I think this is a fantastic addition to the app and I'm thankful there's an eng team tracking and backing these features to ensure quality.
I often find myself running the same/similar workflows and I'd prefer to have a dev team backing the solve imo.
It is so exaggerated and silly to ask for it to be turned off.
It’s completely optional and you only might want to use it if you change the recipe.
Some of these users are overreacting ridiculously to ai.
I think the difference here is that they are paying for a feature they are morally opposed to, and thus supporting it. So now they are faced with two options: subsidize the development of technology they don't like, or leave and stop using the service altogether. The outrage comes from feeling morally obligated to leave a paid service despite being otherwise happy with it.
> Some of these users are overreacting ridiculously to ai.
These are paying customers with overwhelming negative reaction. It doesn't matter their reasons.
If you want to keep your business, you listen to your customers.
For those of us who despise the ai enshittification train, it’s just more slop spreading like acid rain poisoning everything. But if you like ai I can see how it’s no big deal.
Before AI there were millions of stupid blog posts with roughly the same recipes with minor deviations. It was literally human made slop.
At least AI can give me a concise recipe without ads or making me scroll through some story about a brisk day in october.
But it wasn’t on the NYT website. That’s the difference. I could ignore the shitty sites and go to an “authority” (NYT recipes are laughably pretentious but good).
I'm more annoyed by the lazy, reactionary thinking that leads commenters to parrot "ai slop" and "enshittification" in referring to any AI feature in a product. That's the real slop.
Not as lazy as your hot take.
How is it "so exaggerated and silly" to want to disable a feature in a product you pay for? It blows my mind at how allergic AI bros are to consent.
Disabling the feature means disabling it for everyone.
If you don't want to use it, just don't use it.
Because that's not how software features work?
You don't ask, "Hey, I don't have a printer, please disable the Print feature in MS word."
No, you just... don't click that button.
The print button doesn't persist across all sections of the ribbon, the Copilot button does. They decided people weren't using Copilot enough so they carved out a permanent place for it on the window. If the print button started elbowing other UI elements off the screen to occupy more and more prominent places then you could absolutely bet users would complain.
The last time Microsoft went all-in on an annoying UI element like this? Clippy. How'd that go?
If someone is really upset about seeing the button, I guess they could ask an AI to make their browser filter it out
You don't think users, paid users at that, shouldn't have a say in product development? Good lord no wonder the public distrusts the tech community so much.
Yeah, I think the more broad issue here is that NYT should know better that public sentiment on AI is incredibly negative.
I do find this somewhat paradoxical to the fact that clearly a lot of people use it.
It seems like one of those things where culturally it’s become virtuous to be against it and to say you don’t use it, but privately most people are using it.
Still, I think NYT’s subscriber demographic is likely pretty educated and specifically paying for higher quality content. They don’t want haphazardly tossed together recipe blogs.
AI sentiment is negative enough that all companies who build in AI features need to include a toggle that hides them completely. Otherwise you get user backlash like this.
So negative that the first I heard about this was with the title "New York Times adds glue to its Cooking App. Users aren't thrilled" (presumably from someone who read the niemanlab article before I did)
I think there are prob more paid users who want the feature and are willing to pay than those who don't want it, which is why the team built it. They might even believe it brings in new users. It's polarizing but all ai features are lately.
TFA says that over 70% of the comments were negative with the remainder being split between positive and neutral. I'm sure the Times thinks it will rope in new users, but I'm sure the Times is factually wrong.
If it's not enabled by default, "disabled" is the natural state, no?
Or is this more intrusive than I imagine?
How would the NYT Cooking website even look like if it was entirely modularized into pieces that could be turned on and off to customize it to perfection for every taste?
This AI thing is not even active nor activated unless you use it. The complaints here sound more reactive to the word AI than anything else to me.
It's a big ask to be able to toggle whole features lol. Just don't use it if you don't like it
“It blows my mind at how allergic AI bros are to consent.“
You make it sound like you have to use it.
Would you demand your mobile device manufacturer to rollback the features you don’t like and don’t use? Did you consent to it?
The allergy against AI is understandable. But that’s because companies tries to shoehorn it in everywhere. This however sounds like a perfectly fine use case, which is also optional to use.
You make it sound like NYT has to include slop in the app and that the only way to do ingredient substitution is with AI.
“You will do as you are told. The market demands it.”
The New York Times has approximately 6,000 employees, of which over 2,000 work in engineering. In the AI era, each engineer can output 5 to 10x as much as before, and there are more employees than work to be done. Hence employees work pet projects like this one, that few users are asking for.
> In the AI era, each engineer can output 5 to 10x as much as before
Citation needed
There's no way anyone knows where the limit is of a feature set that is valid for AI enhancement. Substitutions and changes in amounts across a dataset that is not fully pre-taxonomized seems like a pretty good use of AI to be honest.
We've used it for modifications to enjoyable effect a few years ago https://news.ycombinator.com/item?id=46131280
So it's a matter of execution etc. to see which features stick.
Few users are asking for it but many are running similar workflows themselves separately. It's a huge value add imo and exactly the right thing an app dev team should be working on.
You rarely get users asking for your next big leap and you often get many users asking for small tweaks (all slightly different).
That is not how/why this came about.
I don’t want to be the one to go there but I think it’s ridiculous that NyT has 2000 engineers. GTA 5 had listed 150 SWE in its credits.
I think we are getting ahead of ourselves if we are to believe the 2,000 number without verifying it.
Advertising and marketing tech increases the bloat, I agree. I too think it's insane that Meta has 20,000 engineers for their entire product suite but when you develop for a malware featureset the bloat comes with it.
I do this exact workflow with random internet recipes and ChatGPT or Claude pretty frequently. In fact modifying a recipe on the fly has been pretty handy.
The base recipes are still vetted by the staff right?
> Say you’re cooking braised white beans and greens with parmesan, NYT Cooking app in hand. But you notice that you’re out of chard and only have spinach. Cooking veterans know that chard is a rough-and-tumble green that has some backbone, while spinach is a comparative wimp that will wilt the moment you look at it funny. If you need to sub in spinach for chard, how would the recipe change? Would you, for instance, add the spinach later than you would the chard to account for its pathetic stamina? That’s the sort of thing the new AI feature could account for.
Frankly I can see how this would be useful to some people. At least it would help prevent the ridiculous "I didn't have eggs so I added applesauce, the cake came out all flat IDK what's wrong" type of substitutions.
I wasn’t going for any of this until your comment.
Now the inane two star reviews will only come from people who are in both categories: (no cooking skill) and (no AI).
>has previously implemented AI features... <to convert volumes to weights and scale quantities for recipes>
(Paraphrased)
Seriously??? You need AI to multiply the quantities in a recipe for more or less servings? Something I was taught to do by hand as a child and can be done with the most trivial of software since the day software was invented?
I think i do want off mr. sama's wild ride
It seems to be a non-invasive assistant to augment the human-created, human-curated recipes they give.
I think NYT Cooking recipes are sometimes... not that great anyway, I think some of their cooking techniques are sub-par.
As for this quote from a mad user: “Anyone using AI for cooking has no sense of taste and no desire to work out the intricacies of cooking and baking”
What a horrible take! I use AI to remind me of certain simple tips I haven't written down, to give substitution advice, the exact kind of stuff NYT app says it's going to do. I learned to cook through Youtube chefs, some book recipes, and I cook for the fun of it and to eat good food. How judgemental to say that someone using AI to learn isn't interested in food!