She's not just cast now, she's a reporter so she chooses which stories and topics to cover and how to cover them I'd assume, so it's more freedom but more work on her end. Shes not just showing us recipes she developed
Priya/NYT, you've really outdone yourself with this one! Such an innovative and fun video concept. It really got me thinking about the less-explored applications of AI. Love it, keep having Priya in more videos!
As an engineer and a foodie, I love all of Priya's content lately. This is just another example of how to make food interesting in a world where new and truly unique content is hard to find.
What a concept! Revisiting in 10 years and see what the AI can do then might be fun. Thinking about what Eric said, I would like to see a video that showcases the entire recipe development process from beginning to end.
@@gdefuykjds 10:32 she is saying this. How would the AI generate the words "my husband Seth" when she didn't mention this during the AI creation of the recipes.
One of you is looking at 10:32 where she's just talking. The other is looking at 23:45 where the AI uses the word "husband" but doesn't mention his name.
Please make a series testing ai generated recipes, each following a “theme”…. I need another show to watch between episodes of Ham and Sohla mystery menu! We love you, Priya!
I would love to see Priya or the team try to make a dish with the same titles as the one the AI recommended. In other words, "if I had to make naan stuffing, here's what it would be"
This was such a cool concept, I'd love to see more of these types of videos! And showing Priya getting a crash course in A.I. was a great choice so that viewers would have the context too. Big fan!
Priya is an undoubtable win for NYT. Her and Sohla are creating some incredible food content and journalism which is engaging, educational and exploratory.
Perhaps you should use the A.I generated recipes as a suggestion; a direction to head down then chefs/recipe developers can tweak the recipe to make it good.
I loved this! Great application of the technology and it really showcases the importance of humanity. I also loved seeing some of my favorite NYT Cooking personalities all in the same video. Happy Thanksgiving!
I was an instructor and that example was used when I went to school. By the time the class finally finished the peanut butter and jelly sandwich, He had peanut butter covering his arms, table, utensils, etc. It was a very good example of how we have to be specific and basic we have to be at times when teaching a subject.
@@WerdnaNiraehs I've seen it used in a programming class to demonstrate how many unspoken assumptions go into giving instructions to humans, and when you're programming you have to be specific about everything. I think it's also just a good introduction to being aware that things you know or things that are common to you are not universal, which is important for education/instruction
we did this in school too but the prompt was basically “tell an alien how to do it” so you had to explain every single thing to a being who wouldn’t know any context. it’s just an interesting exercise to see how much things are from connotation and context and how things can be misunderstood
I would love to see you play around with A.I. more. I think the problem is most people take A.I. way too literally, and it's just not there yet. I think it gives you a great starting point for recipes that can then be tweaked and developed into something new and interesting.
Please make more videos with this exact vibe this was so so so so so good. Any concept is good, but I loooooved that group of five at the end it was so much fun
I really wonder how the tasters would describe the experience if they didn't know that these recipes were AI-generated. I'm sure they would still agree that taste and execution are lacking but I wonder if they would still use words like "no emotion, soul, heart, backstory" etc.
GPT-3 is biased. Unfortunate side effect of what they "learn" from. I wonder what it would give if she didn't say "grew up in Indian household" but rather "texas born enjoy Indian flavors" or "Texas born, some of my favorite spices are x y z. " In any case, very enjoyable video. (I do wonder why it didn't "cook" the turkey right though...
Great topic. Priya was a charismatic host. My take away is that cooking is about the senses. Recipes develop over time based on resources and culture. AI’s Achilles heal is it can’t taste.
absolutely loved this! such an interesting twist on the typical thanksgiving cooking videos, and we stan priya! so cute having all the video contributors together as well. 10/10
I love Priya and this is exactly the content that I love to see. So, so interesting! Loved the IT professionals too! I'd love to see this as a series...
This was so fun! I learnt a bit about AI in university, and its interesting how little some of the fundamental flaws have been addressed which this video demonstrates really well. With simple, shorter recipes and few ingredients (the beans, turkey and sauce) - it can come up with something that works. For more complicated recipes, all it does is throw all it "learns" at the wall because it has no understanding of what the individual elements really mean in relation to one another. Can definitely see this becoming more and more advanced in time though! Will be fascinating to see where this goes.
The gap is because humans need to learn the skill of prompting to get good outputs from the AI. Theres already separate AI models that will take normal human spoken requests and reword them into a prompt that the ai model will take instructions from better. I would like this video to become a series, maybe a challenge around different holidays, to keep exploring where we're at as these AIs progress.
Priya... An interesting Part 2 would be taking those same recipes as an inspiration (which to be fair, you never would have imagined any of these on your own) and then make only the most basic adjustments that anyone who knows their way around the kitchen would make automatically make. Examples: for the stuffing, reduce the spices and mix/ fluff it together before baking. For the turkey, cook to a normal temperature and brine / inject some of that glaze and baste / glaze frequently. For the cake, add the minimum sugar needed with some liquid and oil to make it fluffy. Then see how the "intended" recipes come out given a little adult supervision.
*picturing Seth's HORRIFIED face at finding out A.I. "thinks" that could ever even be his favorite dessert while he mentally starts thinking about how to rework the recipe* Thank you! I really needed new Priya today
This is the funnest cooking related video that I've seen in a very long while! I can see this AI-generated recipe test becoming a new "fun" activity to do...
I loved Wayneradiotv's "A.I.ron Chef" stream, so it was interesting to see an actual chef tackle AI-generated recipes! Also, love seeing Priya always :)
I'm a researcher in this field and what Priya said about the peanut butter jar/professor situation it's sooooooooooo to the point! Thank you so much for this wonderful metaphor. Actually this is the problem in a way that if we don't have an internal model of a system of if we are not capable of arguing whether it is "reasonable" or not it's just us who think it says something and we correct it or find the reasonable patter on what it says. A version of this problem is the apophenia problem. So smart & so cool! (Hopefully in the next years we will figure out some kind of grounding or reason in these models performance however it's not clear whether this will start with such impressive tasks as figuring out recipes. This is a main problem that wow AI now wows us by sparkling our imagination instead of actually taking the responsibility of a task and being useful on its own. It may be a lot of the times, but the problem is how we can ensure this is definitely the case for some tasks. In other words how can we turn it to something that we trust knowing something, is the big question.) PS: Also the nice thing with what the professor did there is that they deliberately did the "unreasonable" act training you to perform reason. This is a challenge to currently do.
This was a great idea for an episode! I do feel that all the recipe developers in the video are a bit quick to say "the AI won't take our job"... The AI they used is trained on "general text" which also includes recipes, sure, but it doesn't really know the difference, so it generates text that seems similar to what a recipe is... If you actually trained an AI ONLY on recipes, and also gave him an indication of the number of stars the recipe got from the readers online (which all websites do now)... I'm sure it would quickly find patterns that make popular recipes, and it would then be able to create much more palatable, and potentially popular, recipes... So I'm sorry Priya, but this isn't for 30 years down the road... it could happen in the year to come, and yes your jobs are also at a danger of being replaced (not entirely, I don't want to sound over-dramatic... but this is the NYT, ok... but think of all the small publications with no money that will soon be able to generate recipes instead of hiring recipe developers and writers.... it's the small people, the beginners, the juniors who are at risk)
I had a teacher do the exact same thing! I forgot to write that she should put the lid of the peanut butter down, so she made the whole sandwich while still holding the lid.
Super interesting. I'd love to see how it would have gone if none of the tasters knew it had been generated by a machine. So many of their comments seemed to imply that knowing that it had been (generated by a machine) influenced their perceptions.
Answer in Progress has a video where they try to create a recipe-generating AI! Also just wanted to say I’m watching this while eating Priya’s recipe for aloo ka rasa!
ChatGPT-4 is out. Please try again the same prompts asking for suggestions and recipes. I very much enjoyed this Thanksgiving episode so I'd like to see some follow-up episodes.
When SkyNet finally arises, it will commemorate the former holiday as “that time the humans finally taught us how food works so we could make ourselves indispensable and them more dependent on us, so starting the inevitable rise of the machines (tm) a mere 19 years later than Foreseen in the historical documents”. If AIs do commemoration, that is.
Fascinating video, though I wonder if Priya was inspired by Answer in Progress’ video on the same topic, AI generated recipes. It’s interesting to see two different approaches and results to the same concept.
I actually subscribe to NYT cooking because I know how much it takes to develop recipes. I'd like a follow-up here: how many attempts do you usually make before considering a recipe is final, what type of feedback do you receive? Can you feed feedback into OpenAI maybe with pictures of final products and ask them to adjust, see how it fares after maybe the 3rd attempt?
It strikes me that the AI recipes could be useful as a source of ideas - combinations of flavors one might not think of, but that then must be "tuned" by a human cook to remove the rough edges, as it were. For example, one obvious correction for the turkey recipe is not to bring it to so high a temperature in the oven.
I feel like the food experts missed the most important point. The AI actually does have some grasp of what foods go together. But it has no concept of quantity and doesn't really understand technique or cooking time. The cake was under-sweetened, the pumpkin spice chaat had way too much spice etc. If you could tweak the numbers and leave the ingredients the same, most of these recipes are salvageable.
This is a great, fun demonstration of the current state of AI, and I think it's a great idea to invite a technology correspondent to a cooking show, but I think you could've used him or a neutral expert much better to explain some of the behavior that you observed in the AI in order to educate on what artificial neural networks really do. It's not by accident that the algorithm put Priya in a box of ethnic identity, the network-axiom of homophily explains this quite well: Homophile artificial neural networks try to identify a dataset in its similarity to others, therefore it relates individuals to a group-identity based on similarity. While this is harmless and maybe even creativity-inducing when it comes to cooking recipes, it is really problematic in areas such as crime prevention, where similar algorithms are being used. While understandable, the most important questions around AI aren't "Will AI replace us, will it be able to do the exact same things as humans do, but better", but" "What is AI doing differently then humans, and what challenges might result out of that?"
While it was cool to see, it was interesting to note in the early part, how all the dishes that was supposed to seem inspired by his "Chinese heritage" were American Chinese-inspired like General Tso's chicken. While I don't think we should dismiss the culinary contributions of the Chinese diaspora earlier in American history, unless it says "American Chinese", I would expect mainland dish inspirations like a tea smoked turkey or dry fried green bean casserole or mantou stuffing or something. Which makes me wonder about systemic biases in the input data. Don't get me wrong, they could be delicious, but just more like something a white suburban American in the Midwest would think represents Chinese cuisine.
That’s such a great point. I wonder if maybe the American aspect is baked in by virtue of thanksgiving being largely an American holiday so it just assumes American + Chinese = American Chinese. Whereas American Indian 🇮🇳 as a food term isn’t as baked into the lexicon so the AI was able to apply Indian ingredients to American dishes.
i think this is a good starting point. I think the naan stuffing is a good idea but it just needs to be tweeked. same with the cake. it just needs some sugar. lol
It's funny to look at the disparity between the comments on this video and the comments from its NYT article. The folks on the article (presumably, mostly older) were not very fond of the video, affirming our knowledge that AI most certainly cannot replace human-made recipes at this point in time, and definitely didn't view it to be as fun as we did.
Also people forget that a couple years ago, AI couldn’t write anything at all, in just a couple years, it was able to write nearly indistinguishable from humans, in 2 or 3 years at most, this kind of AI will be writing entire books in the fly. People have no imagination.
Nice video. Who/what is to blame when the meal is not what was expected and disappointing, the cooker or the AI? Remember that what you get from two persons following the same recipe are generally not the same.
The actual algorithms for neural network AI is rather straight forward (if you know linear algebra). It's the necessary computing power that makes it complex. The more powerful your computers are, the more effective the algorithm becomes.
I feel like sohla & ham are basically just human ai chefs. They come up with the most random things and they seem generally like hits but occasional huge misses
The thing I want is a three way showdown: AI + Novice. Expert only. AI + Expert. Challenge is to create a fully novel crowd pleasing recipe. Metrics are grade of final result + time taken. The question answered is: How much expertise can the AI make up for if some basic human feedback is in the loop and how much can an AI help an expert.
I'd love to see a second video where recipe developers try to take these AI-generated recipes and actually make them palatable.
Oh that's a great idea. Maybe explaining what they look for when writing recipes at the same time.
I totally agree! The stuffing seemed like it could be really tasty if it was just adjusted properly.
I agree, I bet most of these recipes could be adjusted to be good!
Yes! I'd like to see how the sausage is made (maybe not literal sausage though...)
I was thinking the exact same thing.
Okay, NYT, you’ve got a gem in your jewelry box with her. Why don’t you feature her more often?
She's not just cast now, she's a reporter so she chooses which stories and topics to cover and how to cover them I'd assume, so it's more freedom but more work on her end. Shes not just showing us recipes she developed
This must be an AI generated comment
@@22ukiuki Precisely! I love Priya, but clearly there’s more to NYT than just appearing on their RUclips channel, their smallest media outlet.
I know why but I can't say it out loud..
@@popicelolly because she’s too awesome?
Priya/NYT, you've really outdone yourself with this one! Such an innovative and fun video concept. It really got me thinking about the less-explored applications of AI. Love it, keep having Priya in more videos!
priya is really cool 👎
As an engineer and a foodie, I love all of Priya's content lately. This is just another example of how to make food interesting in a world where new and truly unique content is hard to find.
What a concept! Revisiting in 10 years and see what the AI can do then might be fun. Thinking about what Eric said, I would like to see a video that showcases the entire recipe development process from beginning to end.
Maybe 2…
In ten years, an AI will be able to create a recipe development video.
Check out the video about Eric developing a sugar cookie recipe!
Hopefully AI doesn't take over by then...
Love how Priya went from saying "my roommate" 2-3 years ago to 'my husband' 🥰
@@gdefuykjds okay and what is your point?
@@gdefuykjds 10:32 she is saying this. How would the AI generate the words "my husband Seth" when she didn't mention this during the AI creation of the recipes.
One of you is looking at 10:32 where she's just talking. The other is looking at 23:45 where the AI uses the word "husband" but doesn't mention his name.
Please make a series testing ai generated recipes, each following a “theme”…. I need another show to watch between episodes of Ham and Sohla mystery menu! We love you, Priya!
HELLO🤡💣👹
First time hearing priya say husband instead of partner !! Congrats to you and Seth!!!
My husband's favorite recipe was hilarious to me knowing what a high standard Seth holds his baked goods to.
Love that you get to borrow from the other parts of the NYT to develop even more robust stories and videos. This was such a great video!
I remember Priya's early videos on BA. She seems so natural now on camera and picks great topics ! We love to see the growth
I would love to see Priya or the team try to make a dish with the same titles as the one the AI recommended. In other words, "if I had to make naan stuffing, here's what it would be"
naan stuffing sounds epic lowkey…
This was such a cool concept, I'd love to see more of these types of videos! And showing Priya getting a crash course in A.I. was a great choice so that viewers would have the context too. Big fan!
This video is simultaneously the most inspiring and the most horrifying video NYT Cooking has ever posted LOL
I see why they released it as close to Halloween as they could.
She’s so intelligent and optimistic. It’s very refreshing, very likable person.
Priya Krishna is my new hope for 2023 that she gets to share more cinema and food. She shares the most beautiful recipes. She's honest and creative.
Priya is great! Always love her videos.
Priya is always such a delight to watch! NYT-keep it up, these videos have so much life.
Priya is an undoubtable win for NYT. Her and Sohla are creating some incredible food content and journalism which is engaging, educational and exploratory.
Perhaps you should use the A.I generated recipes as a suggestion; a direction to head down then chefs/recipe developers can tweak the recipe to make it good.
I loved this! Great application of the technology and it really showcases the importance of humanity. I also loved seeing some of my favorite NYT Cooking personalities all in the same video. Happy Thanksgiving!
I would like to see her make these recipes not bad. Have the same people come in and give them another chance
You could see the BA war flashbacks when Priya said "it's really put me in a box"
I was an instructor and that example was used when I went to school. By the time the class finally finished the peanut butter and jelly sandwich, He had peanut butter covering his arms, table, utensils, etc. It was a very good example of how we have to be specific and basic we have to be at times when teaching a subject.
I understand you have to make recipes accessible for novice cooks put I don’t get having to assume a Neanderthal is making your recipe?
@@WerdnaNiraehs I've seen it used in a programming class to demonstrate how many unspoken assumptions go into giving instructions to humans, and when you're programming you have to be specific about everything. I think it's also just a good introduction to being aware that things you know or things that are common to you are not universal, which is important for education/instruction
@@schnozz4301 Very good points.
we did this in school too but the prompt was basically “tell an alien how to do it” so you had to explain every single thing to a being who wouldn’t know any context. it’s just an interesting exercise to see how much things are from connotation and context and how things can be misunderstood
To make a peanut butter and jelly sandwich you must first invent the universe
Mr. Rush sounds like a fantastic teacher!
Ok, the subtle "my husband Seth" drop??
I would love to see you play around with A.I. more. I think the problem is most people take A.I. way too literally, and it's just not there yet. I think it gives you a great starting point for recipes that can then be tweaked and developed into something new and interesting.
it is craaaazy to me how much talent BA’s test kitchen had. Their loss truly, but so glad we still get to see Priya, Sohla, Carla et al on our screens
Please make more videos with this exact vibe this was so so so so so good. Any concept is good, but I loooooved that group of five at the end it was so much fun
I really wonder how the tasters would describe the experience if they didn't know that these recipes were AI-generated. I'm sure they would still agree that taste and execution are lacking but I wonder if they would still use words like "no emotion, soul, heart, backstory" etc.
I'm surprised GPT-3 didn't try to go for a Tex-Mex/Indian Thanksgiving fusion since the input mentioned Texas
I expected more of a southern twist!
She said the first version of the stuffing had salsa but that might have been its only attempt to bring Tex Mex in
GPT-3 is biased. Unfortunate side effect of what they "learn" from.
I wonder what it would give if she didn't say "grew up in Indian household" but rather "texas born enjoy Indian flavors" or "Texas born, some of my favorite spices are x y z. "
In any case, very enjoyable video. (I do wonder why it didn't "cook" the turkey right though...
You're right! Actually, Priya did mention that it tried to throw salsa into one of the recipe iterations that she didn't end up using.
Great topic. Priya was a charismatic host. My take away is that cooking is about the senses. Recipes develop over time based on resources and culture. AI’s Achilles heal is it can’t taste.
absolutely loved this! such an interesting twist on the typical thanksgiving cooking videos, and we stan priya! so cute having all the video contributors together as well. 10/10
I love Priya and this is exactly the content that I love to see. So, so interesting! Loved the IT professionals too! I'd love to see this as a series...
Wow watching this, I just miss the old crew so much. They brought out the best in each other.
This was so fun! I learnt a bit about AI in university, and its interesting how little some of the fundamental flaws have been addressed which this video demonstrates really well. With simple, shorter recipes and few ingredients (the beans, turkey and sauce) - it can come up with something that works. For more complicated recipes, all it does is throw all it "learns" at the wall because it has no understanding of what the individual elements really mean in relation to one another.
Can definitely see this becoming more and more advanced in time though! Will be fascinating to see where this goes.
YEWANDE'S FACE when she tastes the chaat just SENT ME
The gap is because humans need to learn the skill of prompting to get good outputs from the AI. Theres already separate AI models that will take normal human spoken requests and reword them into a prompt that the ai model will take instructions from better.
I would like this video to become a series, maybe a challenge around different holidays, to keep exploring where we're at as these AIs progress.
Priya... An interesting Part 2 would be taking those same recipes as an inspiration (which to be fair, you never would have imagined any of these on your own) and then make only the most basic adjustments that anyone who knows their way around the kitchen would make automatically make. Examples: for the stuffing, reduce the spices and mix/ fluff it together before baking. For the turkey, cook to a normal temperature and brine / inject some of that glaze and baste / glaze frequently. For the cake, add the minimum sugar needed with some liquid and oil to make it fluffy. Then see how the "intended" recipes come out given a little adult supervision.
*picturing Seth's HORRIFIED face at finding out A.I. "thinks" that could ever even be his favorite dessert while he mentally starts thinking about how to rework the recipe*
Thank you! I really needed new Priya today
That “Ok, maybe?!” At 14:23 is me looking in the mirror before I leave the house every day 😂
Who came up with this concept? This is AMAZING!
This is great!!! She’s perfect for this episode and I would LOVE to see “AI cooks” again!!!!! Great!!!
This is the funnest cooking related video that I've seen in a very long while! I can see this AI-generated recipe test becoming a new "fun" activity to do...
I loved Wayneradiotv's "A.I.ron Chef" stream, so it was interesting to see an actual chef tackle AI-generated recipes! Also, love seeing Priya always :)
I'm a researcher in this field and what Priya said about the peanut butter jar/professor situation it's sooooooooooo to the point! Thank you so much for this wonderful metaphor.
Actually this is the problem in a way that if we don't have an internal model of a system of if we are not capable of arguing whether it is "reasonable" or not it's just us who think it says something and we correct it or find the reasonable patter on what it says. A version of this problem is the apophenia problem. So smart & so cool!
(Hopefully in the next years we will figure out some kind of grounding or reason in these models performance however it's not clear whether this will start with such impressive tasks as figuring out recipes. This is a main problem that wow AI now wows us by sparkling our imagination instead of actually taking the responsibility of a task and being useful on its own. It may be a lot of the times, but the problem is how we can ensure this is definitely the case for some tasks. In other words how can we turn it to something that we trust knowing something, is the big question.)
PS: Also the nice thing with what the professor did there is that they deliberately did the "unreasonable" act training you to perform reason. This is a challenge to currently do.
This was a great idea for an episode! I do feel that all the recipe developers in the video are a bit quick to say "the AI won't take our job"... The AI they used is trained on "general text" which also includes recipes, sure, but it doesn't really know the difference, so it generates text that seems similar to what a recipe is...
If you actually trained an AI ONLY on recipes, and also gave him an indication of the number of stars the recipe got from the readers online (which all websites do now)... I'm sure it would quickly find patterns that make popular recipes, and it would then be able to create much more palatable, and potentially popular, recipes...
So I'm sorry Priya, but this isn't for 30 years down the road... it could happen in the year to come, and yes your jobs are also at a danger of being replaced (not entirely, I don't want to sound over-dramatic... but this is the NYT, ok... but think of all the small publications with no money that will soon be able to generate recipes instead of hiring recipe developers and writers.... it's the small people, the beginners, the juniors who are at risk)
I had a teacher do the exact same thing! I forgot to write that she should put the lid of the peanut butter down, so she made the whole sandwich while still holding the lid.
Priya is my favorite cooking nerd. She's so cute! 🥰
Prya is so beautiful, it’s crazy! Congrats to Seth 🙌😍👏
I love videos with Priya! 💖 This was a very interesting watch. ☺️
Super interesting. I'd love to see how it would have gone if none of the tasters knew it had been generated by a machine. So many of their comments seemed to imply that knowing that it had been (generated by a machine) influenced their perceptions.
This is such a fun (and funny) series. I NEED more
Answer in Progress has a video where they try to create a recipe-generating AI! Also just wanted to say I’m watching this while eating Priya’s recipe for aloo ka rasa!
we love a fun Priya video.
This needs to be a series. Cooking Decoded (or something). AI-generated recipes tweaked and refined by chefs.
Yeah. The perfect thanksgiving. Build a time machine and have a team of travellers destroy all those ships trying to get to the new world. Perfect.
i'm really enjoying priyanka's work with NYT cooking - hope to see her featured more.
Please keep the Priya content coming... Loves it!
This is unironically so funny they need to make this as a series 😂😂
I giggled throughout this entire video which is something I don't usually do with cooking videos. Loved this. MORE.
I always love watching Priya videos! She's so creative, energetic, and informative!
I love this episode. And whenever Melissa or Priya are in a video I'm all for it
ChatGPT-4 is out. Please try again the same prompts asking for suggestions and recipes.
I very much enjoyed this Thanksgiving episode so I'd like to see some follow-up episodes.
Love this but yeah, definitely tweak the cook times or salt levels etc based on chef knowledge.
When SkyNet finally arises, it will commemorate the former holiday as “that time the humans finally taught us how food works so we could make ourselves indispensable and them more dependent on us, so starting the inevitable rise of the machines (tm) a mere 19 years later than Foreseen in the historical documents”. If AIs do commemoration, that is.
Fascinating video, though I wonder if Priya was inspired by Answer in Progress’ video on the same topic, AI generated recipes. It’s interesting to see two different approaches and results to the same concept.
I actually subscribe to NYT cooking because I know how much it takes to develop recipes.
I'd like a follow-up here: how many attempts do you usually make before considering a recipe is final, what type of feedback do you receive? Can you feed feedback into OpenAI maybe with pictures of final products and ask them to adjust, see how it fares after maybe the 3rd attempt?
LOVE this episode. This is peak Priya
This was great! Would be cool to do this as a series and have Priya or another editor improve on the AI generated recipe
Would love to see what the AI could do with input like recipes from Priya's book and see if it could generate new recipes based on that input
Heyyyy, shout out from one Dallas chef to another!! Good to see a hometown girl doing so well!!
This video is a hit! It was very fun and eye opening. Great job!
For the turkey I think Priya misunderstood the instructions. It mean 180 degrees c which = 356 c or roughly 350 in US.
I legitimately want the recipes for these. With a few tweaks, I feel like... well, not all, but most of them could end up being really good.
Priya is awesome! I am an amateur artist and get the same feeling when I look at AI generated art - no soul.
It strikes me that the AI recipes could be useful as a source of ideas - combinations of flavors one might not think of, but that then must be "tuned" by a human cook to remove the rough edges, as it were. For example, one obvious correction for the turkey recipe is not to bring it to so high a temperature in the oven.
I wanted suggestions for a Korean-inspired Thanksgiving. Also got green beans with sesame seeds.
loved the concept, you're killing it gurlll
Priya!!! This episode idea was so inventive. Educational and entertaining 🎉
This theme as a series would be hit
I feel like the food experts missed the most important point. The AI actually does have some grasp of what foods go together. But it has no concept of quantity and doesn't really understand technique or cooking time. The cake was under-sweetened, the pumpkin spice chaat had way too much spice etc. If you could tweak the numbers and leave the ingredients the same, most of these recipes are salvageable.
“It is drier than the driest turkey I’ve ever had” now THATS saying something 😂
That naan stuffing is the "mole" of stuffing with the sheer amount of ingredients it had!
This is a great, fun demonstration of the current state of AI, and I think it's a great idea to invite a technology correspondent to a cooking show, but I think you could've used him or a neutral expert much better to explain some of the behavior that you observed in the AI in order to educate on what artificial neural networks really do. It's not by accident that the algorithm put Priya in a box of ethnic identity, the network-axiom of homophily explains this quite well: Homophile artificial neural networks try to identify a dataset in its similarity to others, therefore it relates individuals to a group-identity based on similarity. While this is harmless and maybe even creativity-inducing when it comes to cooking recipes, it is really problematic in areas such as crime prevention, where similar algorithms are being used. While understandable, the most important questions around AI aren't "Will AI replace us, will it be able to do the exact same things as humans do, but better", but" "What is AI doing differently then humans, and what challenges might result out of that?"
While it was cool to see, it was interesting to note in the early part, how all the dishes that was supposed to seem inspired by his "Chinese heritage" were American Chinese-inspired like General Tso's chicken. While I don't think we should dismiss the culinary contributions of the Chinese diaspora earlier in American history, unless it says "American Chinese", I would expect mainland dish inspirations like a tea smoked turkey or dry fried green bean casserole or mantou stuffing or something. Which makes me wonder about systemic biases in the input data.
Don't get me wrong, they could be delicious, but just more like something a white suburban American in the Midwest would think represents Chinese cuisine.
That’s such a great point. I wonder if maybe the American aspect is baked in by virtue of thanksgiving being largely an American holiday so it just assumes American + Chinese = American Chinese. Whereas American Indian 🇮🇳 as a food term isn’t as baked into the lexicon so the AI was able to apply Indian ingredients to American dishes.
100%, makes me think of that "image enhancer" AI that turned everyone and everything into white people
I'm CACKLING watching poor Priya throw all these ingredients around. This is hilarious.
I would love for you guys to do this again but take inspiration from some out there recipe’s and tweak it to make it taste good
i think this is a good starting point. I think the naan stuffing is a good idea but it just needs to be tweeked. same with the cake. it just needs some sugar. lol
It's funny to look at the disparity between the comments on this video and the comments from its NYT article. The folks on the article (presumably, mostly older) were not very fond of the video, affirming our knowledge that AI most certainly cannot replace human-made recipes at this point in time, and definitely didn't view it to be as fun as we did.
Also people forget that a couple years ago, AI couldn’t write anything at all, in just a couple years, it was able to write nearly indistinguishable from humans, in 2 or 3 years at most, this kind of AI will be writing entire books in the fly. People have no imagination.
Nice video. Who/what is to blame when the meal is not what was expected and disappointing, the cooker or the AI? Remember that what you get from two persons following the same recipe are generally not the same.
This is such a unique concept. Awesome video!
The actual algorithms for neural network AI is rather straight forward (if you know linear algebra). It's the necessary computing power that makes it complex. The more powerful your computers are, the more effective the algorithm becomes.
I feel like sohla & ham are basically just human ai chefs. They come up with the most random things and they seem generally like hits but occasional huge misses
The thing I want is a three way showdown: AI + Novice. Expert only. AI + Expert. Challenge is to create a fully novel crowd pleasing recipe. Metrics are grade of final result + time taken. The question answered is: How much expertise can the AI make up for if some basic human feedback is in the loop and how much can an AI help an expert.
This would be a fantastic series! Amazing video, Priya!
I love priya! ❤
Naan Stuffing sounds kinda amazing if it were done properly
I want to see Sohla and Ham fix these recipes and make them tasty.
Bobby Flay is the AI 😭😭😭 no DOUBT
This provided me with an education on AI. Thanks