The Borg Chef

Can computers get creative in the kitchen? From robot cooks to an AI application that invents new recipes, we look at how technology may change cooking forever.

Learn more about your ad-choices at https://www.iheartpodcastnetwork.com

See omnystudio.com/listener for privacy information.

Episode Transcript

Speaker 1 (00:00):
Brought to you by Toyota. Let's go places. Welcome to
Forward Thinking. Hey, they're in weld into Forward Thinking, the
podcast that looks at the future and says John Belaiah
and the Crawfish Pie and a Dele Gumbo. I'm Jonathan Strickland,

(00:20):
I'm Lauren block Oban, and I'm Joe McCormick. And today
you're hungry. Well, actually I am hungry. I'm pretty much
always hungry. I too early today and now I am hungry.
It's my secret. I'm always hungry. So today for lunch,
I went to a local h what would you call it, shrine,

(00:41):
local local establishment called eats. It's a Southern cafeteria. It's
called Eats. The sign for which I always think. I
don't think anybody agrees with me, but to me, the
sign for Eats looks like it says ants, so I
call it iNTS. So, so lunch there can last upwards
of ten years. It really could. It could be in

(01:01):
your body that long. Uh, Joe. Joe decided to be
adventurous and try something that he doesn't normally try. Yeah. Well,
I didn't like my my new dishes as much today
as I like their their standard like jerk chicken, which
is pretty good. They're really known for it. But it
is definitely always worth being a little adventurous in the
kitchen trying something new. Yeah, and and another thing we

(01:23):
wanted to mention. The new thing is going to become
really important later on and it gets weird, y'all, but
before we get there. Uh, you know, the other thing
I wanted to mention is that the kitchen has for
a really long time been kind of a gateway to
technology in the home. H Yeah. Yeah, people are very
technologically adventurous in the kitchen as well as food wise. Yeah.

(01:46):
I think I even was looking around and the reason
why whenever c e S rolls around that used to
be known as the Consumer Electronic Showcase, but now it's
just c E S. Uh. There's always a huge number
of various alliances and inventions that our kitchen oriented, and
it's largely because when it comes to technology, we tend

(02:06):
to adopt it for the kitchen pretty early. People love
kitchen gadgets. Yeah, late night TV filled with the various
kitchen gadgets of questionable quality, many of which I own. Uh.
So I wanted to talk a little bit about how
the kitchen is kind of this this gateway to home
automation and programmable appliances before we get into the concept

(02:30):
of robot chefs and AI getting involved in our food,
just to kind of illustrate how it's been changing, even
uh just recently. Oh sure, we take all of these
automated kitchen devices very much for granted. These days, you know,
you're your coffee maker that you can program to go
off whenever you want coffee, uh, in anticipation of when

(02:53):
you're going to want coffee, but which is always. But
getting to this point has been kind of uckey. The
very first computer that was built for the kitchen was okay,
to be fair more of a marketing ploy than it
was a real product. But this was in nineteen sixty
nine and at the time even Jane Jetson's kitchen still

(03:13):
ran on punch cards. Uh. So to stir up publicity,
retailer Neiman Marcus paired up with technology innovator Honeywell to
promote a kitchen computer which would plan menus for you
based around a single main course. It also weighed about
a hundred pounds uh and was so large that it

(03:34):
came in the form of like a small kitchen island
with a cutting board and everything. Uh, it was it
was really a repackaged Honeywell three sixteen, which was a
sixteen bit mini computer. By mini computer, they meant smaller
than a room at the time, and and and these
things were commonly used in like commercial and industrial applications,

(03:55):
So it was it was so complex to use that
the come but he has offered a complimentary two week
programming course to potential buyers. Like it had it had
no screen, It was just like a flashing light display.
The communicated in binary code. So uh, only twenty were
ever produced, and those only after public interest was way

(04:18):
higher than they anticipated. It was basically it was like, hey,
talk about this thing that we might possibly do to
be fair Enow, this is an era where people were
really looking forward, like the home of the future, right,
this this idea of everything is going to be easier
from now on because it's all going to be automated.
This is still a story we're telling today, but it's

(04:38):
obviously taken a lot longer than what we anticipated. So
I'm curious if back in those days, I wanted one
of these miraculous machines to grace my kitchen, how much
might I be expected to pay for such a thing. Well,
it was priced at wow. Yes, but when you account

(05:01):
for inflation in today's dollars, that is worth sixty thousand,
nine hundred and twenty five So so for the price
of just a couple of automobiles, Yeah, it's it's unlikely
that anywhere ever sold. Wow. That's pretty amazing. But you know,
obviously that was not just the beginning and the end

(05:23):
of this this you know trend in trying to create
automated devices for the kitchen. Oh sure, it would take
the level of miniaturization and the level of Internet access
that came along in the late eighties and early nineties
to really kick off the creation of inexpensive and actually

(05:44):
marketable devices that do not talk to you in binary code.
The first touchscreen fridge, for example, debuted in from electric Lux,
and basically after that, we've been living in a in
a rush of kitchen tech. You know, I feel like
crowd fun ending websites are just chock full of kitchen
gadgets people are trying to fund, like immersion cookers, that

(06:07):
kind of thing. I mean, I remember seeing that, like
for suvied type approach. I've been seeing those crowdfunded for
a while, but all sorts of kitchen gadgets are up there.
And uh, I mean I think of like when you're
thinking about the the the history of automation in the kitchen,
things like dishwashers obviously spring to mine, and coffee coffee makers,

(06:29):
like we were saying earlier, and also things like bread makers.
I mean, I remember when bread makers really became a thing. Yeah,
my parents got one when I was a kid, and
I remember, I remember it being an extremely long, laborious
process to make, even with the bread makers. But if
you if you were doing something like putting the ingredients
in the morning and starting it off and then just

(06:51):
letting it go, then by evening you had a fresh
baked loaf of bread. The house smelled of bread. This
was a miracle machine as far as I was concerned,
because my mom makes amazing homemade bread, but because doing
it by hand is even more laborious than using a
bread maker, it was so infrequent that this went that

(07:11):
we actually got to have it more often, which in
my mind was amazing. Yeah. So some of the things
I wanted to mention before we get into the robots stuff,
just some some more examples of uh technology that's becoming
available now, stuff that that general automated cooking gadget. Yeah,
things that have debuted at c e S recently that

(07:31):
kind of stuff. One was the Rodi Matic. I love
the name already, but it's an automatic Rodi cooker. So
you can get like r O T I R O
T I h the Indian pancake type of thing. Yeah,
so you just load the ingredients. They have these different
canisters that hold the various ingredients so that they mixed
them in the proper proportions and and uh you then

(07:53):
said it and it cooks about one per minute. So uh,
if I had more room in my kitchen, I would
get one of these because I think that is amazing.
But I think my wife would kill me. I'm the
cook in our family, by the way, so to me, like,
these are the things that appeal to me, And then
I think I don't have physical space for it, so

(08:13):
it's hard for me to justify purchasing it. Um the
Smarter coffee maker. Uh, this is one of those things
that you you wouldn't really think you needed until you
see it, and then you think, but I don't understand
the appeal what's the deal. So the smarter coffee maker
goes beyond programming, like we're used to the programmable coffee
makers at this point. If you have one of those,

(08:34):
then you can you know, you load it up and
you say, all right, at at six thirty in the morning,
start brewing my cup of coffee. So that way I
have it as soon as I get up and come
downstairs or whatever. The smarter coffee maker goes beyond that.
You can actually access it through an app. So you
use your app to tell your coffee maker exactly how
much coffee you want to make, so if you want
a big cup of coffee, you say, use this much

(08:55):
water versus that much water. Um. It also will keep
track of how many beans are still in the coffee maker,
so when it gets low, you can reload it add
more beans to it. So uh, kind of an Internet
of things approach to making coffee. There was the x
y Z printing three D food printer which was shown
at c e S that was capable of using sweets

(09:17):
ort of cookie dough and printing it in different shapes.
So not quite the three D printer that makes a pizza,
but one that could make you know Christmas cookies if
you wanted or whatever shape cookies you wanted. Uh. So,
I don't know how many people would actually get one
of those. Again, like my immediate responses, I totally don't

(09:37):
need that, but I totally kind of want one. Uh
There was the Pico brew Zymatic, which is an automated
beer brewing device. It takes four four hours to brew
an unfermented beer. I looked I saw your note on this,
and I was like, an unfermented beer fermentation the point
of beer. Yeah, it is, no, So the unfermented beer

(09:58):
is not what you drink, and so it'll make this
and then you can ferment it and then you can
have beer. Because I was thinking what would that be like,
I'm sure it would be gross. I don't think they're
suggesting you drink it at this stage. En Fermented beer
is called wart wart with an o wart word. I
don't know how you say it. I guess it's just
we're saying like one of the crunts from Halo. I

(10:21):
don't know what you're talking about, not a Halo player anyway,
but it's this this sugary grain based liquid that becomes
beer through the process fermentation, Right, so this would be
something that would speed up that process and automated so
that you don't have to have the same level of
um you don't have to babysit it as much. Right,

(10:41):
that's the that's the big thing, until you get to
the fermentation stage, which obviously would go outside of this
this device. Yeah, I was reading a little bit about
it and it said that the main appeal of this
thing is if you want to be able to experiment
with a lot of different recipes with very precise measurements. Also,
it appeals to pete who live in a place where

(11:01):
they might not have a lot of space to do
this kind of stuff on their own. Uh So, people
who might live in an apartment or a townhouse where
they don't have like a big garage or something. It
was very appealing for those folks as well. Whirlpool went
bonkers at c S and showed off a fully interconnected
kitchen with all the different appliances able to talk to
all the other different appliances. It even included an interactive

(11:23):
backsplash so you could get like recipes printed on your
backsplash while you're working on stuff, and cook tops that
could actually link to your social media profiles. So if
you wanted, if you wanted to take the step out
of tweeting what it is you're having for dinner, Jonathan
Strickland is using medium high heat exactly. I can't say

(11:44):
that I immediately recognized the appeal, but it certainly was
built into it. People are very fond of, you know,
posting their dinner. That's true social media, that's true. And
you know, if you if you're doing that three times
day every day, it just wears out your type and thumb,
you know. I feel like people are more fond of

(12:05):
posting their brunch or lunch rather than dinner. Maybe I'm wrong.
I think it depends on where you're eating, if you're
eating at home, um like a weekend to brunch or lunch.
I'm sure there's their statistics out there for this big
data get back to us. Yeah. I think it probably
also would give you an indication of how well someone's
date is going based upon the number and frequency of

(12:28):
food related tweets you see from that person. Okay, Jonathan, Sorry,
everything you've said so far, I'm unimpressed. I want a
robot or cyborg type organism or machine that cooks for me. Well,
I have some good news for you, Joe. They're top people.
I always want to see top men because that's the raiders.

(12:49):
Come on, let's be fair. There are tough. In fact,
one of the people who is behind one of the
robots chefs I'm gonna talk about is an eighteen year
old lady and she really impressed me because I got
to meet her in person on an airplane on the
way to see es. But so, why don't we have

(13:09):
robot cooks? Well, I mean that's that could be asked
about almost anything where. Why don't we have robot butler's. Yeah, well,
why don't we? I mean, that's a good question, and
the answer is, actually, I think fairly similar because these
jobs require all kinds of versatility, of movement and recognition,
you know, multiple different types of artificial intelligence converging on

(13:33):
a task. That can become pretty intuitive to us. But
think about all the different types of AI it needs.
It needs spatial awareness and manipulating objects and visually recognizing
things and understanding what tastes are. I mean a lot
of manual dexterity. Yeah, yeah, it is incredibly complicated, but

(13:53):
that hasn't stopped various people from trying to create types
of robots that would help with cooking. So you know,
we've already talked about some automated devices that are you
could think of them almost like robots. Uh. And they
certainly have some functions that are are similar to robotic functions,
but they're very very specific in there the tasks they

(14:14):
can do, which not a big surprise, right, because when
we talk about robots, that's that tends to be the case.
We talk about people designing robots for very specific tasks.
That's way easier than designing a robot for general tasks.
So one of those companies is called a Serenity Kitchen.
Serenity is spelled s e r e n e t I. Uh. Yeah,

(14:34):
I keep on and calling calling it a sereneti kitgenimate
Serenity Kitchen. This is this is the group that I
got to meet on an airplane trip to c e S.
I was flying in the very back of a plane
and the road immediately ahead of me were the people
from Serenity Kitchen, which was three young students, uh, one
of whom was the person who came up with the idea.

(14:55):
She was the brains behind the concept. And it is
a brilliant concept. So the basic idea was that you
have a essentially a cooking surface and a robotic arm
mounted above the cooking surface, and the robotic arm could
create a stirring motion to stir whatever was in a
pan that's on top of that cooking service. So you

(15:17):
would put ingredients into the pan, the robot arm would
continuously stir it until the food is done, and that
way you wouldn't have to babysit stir. Yeah. Yeah, oh
that's just just the application for making a ru Yeah, exactly, yeah,
something like that. Now they were looking they were actually
using it to show off things like pad tai. Uh.
They were showing a lot of Eastern dishes and they said, well,

(15:41):
we also have applied this for Western dishes like a
you know, if you want to make a spaghetti or whatever.
And they said, really, the first target audience, we're looking
at our college students, people who would maybe go and
buy a cheap microwave dinner, which typically doesn't taste very
good and usually isn't very good for you. So what
if you would create a device that would make uh,

(16:03):
the the healthy food for you. You You know, you're just
dumped the ingredients into whatever and it would do the
rest of the work because college students are lazy, and
it would allow them to to not have to put
forth that kind of effort and they would get the
benefit of a healthy dinner h cooked for them. And
the brilliant part of their approach was that their plan

(16:25):
wasn't to sell expensive robot cooks. Uh. They were actually
looking at using the robot cook is sort of a
lost leader the same way that video game consoles come out.
That's immediately what I thought of, like that they you know,
they're not making a lot of money selling you a
PlayStation for something. They're hoping that you're they're going to
make it back and you're buying games, right, so instead

(16:45):
of games, we're talking about meals. So they actually compared
it to curig, which makes sense, the coffee maker that
uses the little pods. They said, well, you know, you
could buy this this device, Uh, if you wanted to
buy it out right, you could, but would cost somewhere
around the neighborhood of five nine U s dollars. It's
pretty expensive, you know for something that would not be

(17:08):
a direct replacement for a large appliance in your kitchen.
But they had a subscription plan where you can subscribe
to a service that would send you the ingredients for
ten meals a month for three years. So it's a
three year commitment, and then the robot would cost just
ninety nine dollars. The subscription would be fifty dollars a month,

(17:29):
and that sounds like a lot at first, but then
when you figure weight, it's ten meals in a month.
That's five bucks a meal. That's a fast food meal
right there, but you're getting fresh ingredients. So the idea
being that, uh, you know, this would be a way
to get people locked into a system where they would
get the delivery ten meals a month. They'd be able
to use this. They could use it for other things too,

(17:50):
if they had the ingredients on hand, but these would
be things that robot would be specifically programmed to handle.
So you put the ingredients in either directly in the
p in or their actual plan has a robot that
would have trays and you put the ingredients into the
the individual trays according to the directions of the recipe,
and then the robot would add the ingredients as needed

(18:13):
in the process of making that. After two minutes of
sauteing the onions, then you put in the carrots, and
then you put in the meat and then right exactly. Yeah,
So that way you you cook the various ingredients the
right amount of time so that it's the right consistency, texture,
nutritional value, all that kind of stuff. And I thought
this was brilliant. I thought that their business plan really

(18:34):
appealed to me because you could tell they got it.
They it was a very uh, innovative approach, not just hey,
we'll sell them the equipment, more like, hey, we'll sell
the equipment at a big discount, but will really make
the money on the food part. And as long as
the food is good and the robot does what it's
supposed to do, then that's a great business plan. So

(18:57):
if you look at the images of what it actually
the working model they brought to see, yes, what it
looked like, and then you look at the mock ups
of what they wanted to look like when it's an
actual consumer product, it's miles apart. But that makes sense
because it's a prototype versus what would be the finished product. Uh.
But you know, they did launch an indie go go campaign,

(19:17):
which they ended early. I think they ended it after
four days. Uh, And the reason they did was They
said they got a lot of feedback that they thought
was helpful and they wanted to incorporate in the design,
but they didn't feel like continuing with the indie go
go campaign would reflect that. So they ended in and
said we're gonna come back in a few months with
a new design based upon the feedback we're receiving. Um.

(19:39):
They also were talking about that it would be great
in the future if the robot could do more than
just you know, essentially it's just adding ingredients and stirring um.
If it could do more than that, like do more
of the prep work or or you know work, or
maybe maybe even something a little more simple, like being

(19:59):
able flip over an omelet or something. Because as opposed
to just stirring um. So that was really interesting to me.
And like I said, I was really impressed by their
their ingenuity, uh and really their business plan. It was
beyond just the technology, which was already interesting, but the
idea that they were like, well, convincing people to buy
an expensive robot is kind of a no win approach,

(20:21):
Like it's you know, unless you are just you know,
determined to go after a very niche audience, you're not
really gonna make a lot of sales that way. But
then we can look at a different one, a different
approach that doesn't really concern itself too much with making
affordable robots. Yeah. Yeah, let's let's say that that we

(20:42):
are still or that Joe at the very least is
still deeply unimpressed like this entire plan. Like he he
needs something that's going to do way more work than
just stirring the food. And he says, you know those
kids over at Sertainity, they're they're great and all, but
I need I need a heavy duty robo chef, not
just a cook. I need a chef. I want arms

(21:02):
that operate on a rail above the stove encounter. Well,
I have news for you, my friend, Moly Robotics. It's
a London based company that created a robo chef. It's
more advanced than the cookie concept. Cookie, by the way,
is the name of the robot from Serenity Kitchen. Uh
so it actually incorporates learning algorithms, or at least the
end product is supposed to, meaning that you can show

(21:26):
the robot how to make something and then it will
then be able to make it. You know, repeat that
that process and make that thing for you in the future. Yeah,
so I watched a demonstration video of this robot trying.
I think it was making some crab bisk. I'm not positive,
but that's the impression I got because they mentioned crab
bisk and the soup looked kind of bisky. It took

(21:47):
a fresh crab and cracked it over the pot, crack
into fresh crab exactly. Uh no, but it was. It
was kind of like it was a pair of arms
that are hooked to a ray all over the stove
area and then yeah, moving back and forth, grabbing ingredients,
pouring them into the pot, stirring them up, adding things.

(22:11):
But but but how does it learn how to do
all of this? Well, the funny thing is it doesn't
have any idea what the heck it's doing, so it doesn't,
you know, when it grabs the olive oil and adds
that to the pot. If you were to put a
bottle of arsenic where the olive oil is, it would
just add that all the same. It learns based on

(22:32):
basically spatial memory from observing a human do the same actions,
which means that you have to place the ingredients in
the same spots as when you taught the robot how
to do this thing. Sure, sure, And and it could
be slightly less catastrophic than arsenic its olive oil. It
could be as vinegar that would make the taste of

(22:56):
the dish quite different. Y. Yeah, and also from the
cooking pattern, Yes, the ingredients, the ingredients you would use,
you might put in very special containers that are specific
sizes and shapes, and that would allow the robot to know,
all right, the thing that's in the round container that's
three inches tall goes in next, then the squarish container

(23:18):
that's four inches tall that gets at it after that.
Like it has it's like a program. It follows a
specific set of steps, and you can show it what
steps it has to follow. But you then have to
set everything up exactly the same way so it can
follow those steps. It can't just take raw ingredients that
are piled in on the kitchen counter and then chop

(23:39):
them up figure out what to do with them. Yeah.
But but it's still pretty cool because once you show
it how to create a recipe, it can it can
remember that forever, and it can create that recipe without
any variation. Like it could be perfectly consistent, so you
can see how something like this would be incredibly useful
in a restaurant. And yeah, where your goal is to
do it the same every time. Want you want the

(24:00):
dishes to be as consistent as possible. You don't want
there to be a lot of variation unless there's, you know,
a request from a customer saying, you know, I'm allergic
to such and such. Don't put that. Please replace my
olive oil with yeah when you're you know that person.
Uh so the oh man, what a great kurdled biscuit.

(24:21):
So one of the big challenges they had was creating,
uh hands that would mimic the the utility of a
human hand because they're pretty complex. That's still a robotics problem. Yeah,
it's tough, and they did a pretty good job. They
said that they had almost all the same degrees of
freedom of movement that a human hand has. They said,
the really the only part of a human hand that

(24:42):
they weren't they didn't replicate for the purposes of this
robot was the ability to kind of not dislocate, but
but to shift your thumb over and like they that one,
they said, wasn't necessary for them to be able to
hold onto the various implements, and they used the exact
same type of kitchen utensils that we humans would use,
like a ladle or spoon or whatever. Uh. Those can

(25:04):
be hung up and the robot can can recognize which
ones are which and grab the appropriate one at any
given point of the recipe. Um. They said, it was
really tough. I'm sorry, I'm just imagining a scene in
a movie where somebody has one of these in their
house and they get murdered because somebody comes in and
teaches the robot how to stay. Let's hope that that

(25:27):
would be against its product, Like I would say, I'm sorry,
that's against my my law of cooks robotics division. Um.
So it also can't send some of the more delicate
processes that are involved in cooking things that have some
variability to them. And the example that they gave, which
I thought was a very uh it was a great example,

(25:49):
was if you want to beat eggs so that they
start to peek, Yeah, egg whites, Like if you're trying
to like you're trying to get to a meringue or
something like that, and you're beating these egg whites, it
doesn't Sometimes it takes a little longer than other times,
Like it can depend upon the you know, whatever eggs
you've got at whatever time of year. Yeah, that the
temperature of the bowl and of the eggs, and the

(26:11):
humidity in the room. Yeah, So these these variables may
be difficult for a robot to sense initially. So that
would mean the robot would not necessarily have the the
ability to predict how long it would take to get
to the right point. So for something like that, you
would need to build in other elements for the robot
to be able to sense when it needed to stop,

(26:32):
like right, like um, like the resistance of the whisk
going through. Yeah, or it may it may be as
simple as some stereoscopic cameras that can get a good visual,
uh idea of what's going on, or the resistance sensors.
It might be a combination. Probably would be to try
and increase the accuracy as much as possible, although peaks

(26:54):
are stiff. You have to you know, you have to
balance the accuracy versus the cost that it adds to
the overall device obviously. But yeah, that's one of those
things where that would they hope to have that built
into future versions of this. The BBC covered this robot
Chef in a story and mentioned that it was being
trained by Tim Anderson. It was a BBC Master Chef

(27:16):
Champion in two thousand eleven. I want to be a
cooking robot? Yeah, yeah, would you learn from Tim Anderson? Yeah?
I keep wanting to call him master Chief instead of
master Chef. I see master Chef and I'm like, he's
master Chief. Uh. This is the most halo centric cooking
episode that has ever been recorded by anyone war war.
So ideally you would end up getting this this very

(27:39):
consistent results. So it might not be the most exciting
thing for a home chef, like a home cook, unless
you just you know, as you know, there's this one
dish that you love and you love you know how
it's made every time, then obviously that would be desirable
to you. But some people say, well it kind of
takes the experimentation part out of cooking. Um, And so

(28:00):
it really depends upon your approach to cooking in the
first place. Some people take a more out and brown
kind of chemist approach and physics based approach, and some
people are a little more we call them bakers. Some
people are a little more loosey goosey. We call them
me and uh, yeah, I am not. I don't bake
well because I'm a little too loosey goosey with the
rules and that does not work in baking. Do you

(28:23):
not own a scale? I do own a scale, Actually
I don't use it. Some people were really bad at
potions or chemistry, depending on which universe you live in,
and uh and and and some you know, yeah, I was.
I was. That was always my strong point. I am,
in fact baker great in charms, but terrible in potions. Uh.

(28:44):
So they want to develop this further. They want to
make the finished consumer version smaller. It's a larger form
factor than they would prefer right now to try and
put on a consumer market. And they also want to
incorporate other things in the robot itself, like a ball
refrigerator and a dishwasher, so would actually replace certain things
in your home and therefore maybe justify the incredibly expensive

(29:10):
price tag. So the BBC says the estimated price will
be around ten thousand pounds or fifteen thousand, six hundred
dollars US for the for the cheap version. The expensive
version would be closer to dollars more expensive than the
Honeywell commutery. Yeah, I was about to say, we're getting

(29:30):
right back into Honeywell territory. That's exciting. Yeah. So this
would not necessarily be something your average home would would
have uh incorporated into the kitchen. I mean, kitchen appliances
are expensive. But I will start thinking about going to
the bank and getting alone for this. If this thing
can deveye shrimp and pitch cherries throwing, throwing a cherry

(29:56):
pitter and an in, I feel your pain. I certainly would.
You know, for me, it's like the idea of cleaning fish.
I can't do that, but I would buy a machine
that could um. So one of the things we also
wanted to mention was that, you know, robots are one thing, right,
the automation of these tasks, that would free up a

(30:17):
lot of time for us, so we could do something else.
And for those of us who you know, we might
enjoy cooking, but there's certain parts of cooking that we
find tedious. We could give that over to the machines.
But what about going a step further and having artificial
intelligence inform the types of foods we prepare in the
first place. So there, you mean the capacity of a

(30:38):
chef not to do the physical act of cooking, but
to come up with recipes something vent dishes exactly, to innovate,
not just to follow a set of instructions, but to
come up with new instructions, perhaps combinations and instructions that
have never before been attempted by humans and should be

(30:58):
some of which may fall into that out ofgory. So
we when we can talk about it, Yeah, we're going
to talk about chef Watson. Uh so Watson. Hey, we
talked about Watson recently, so this would be the it's
been a Watson kind of month, it has been. Yeah, Watson.
It's funny. On forward thinking, I've been talking all about
Watson and on tech stuff. I've been talking about the

(31:18):
Manhattan Project, Manhatton project, and the post apocalyptic technology and
radiocarbon dating. So I guess I just get on a
kind of a streak after a while. But with Watson,
in case you don't remember, Watson's the computer developed by IBM.
It played on Jeopardy and One against two returning champions
because humans aren't smart I know who, I'm just kidding.

(31:40):
They were very smart, but the computer had a very
impressive performance. Yeah, and it's a it's a cognitive computer.
It learns UH and as as it is trained, it
learns how to perform certain tasks, whether it's actual physical tasks,
you know, if you have a robotic component, or just
you know, the process of something better and better as

(32:01):
it as it's as you train it and tell it, yes,
you've done a good job. That's exactly what we want,
or no, this is not the result we wanted, then
it can start to learn and do them more and
more efficiently. So Watson has been used for all sorts
of stuff we talked about in the Tone Analyzer episode.
It's also been used to help with UH in the

(32:21):
medical field to help lots of different processes and anything
that you want to do that requires you to to
bring a whole bunch of information to the table. It's
it's really great for yeah, and it can it can
end up kind of weighing that information and giving you
at least an idea of what information seems to be
the most relevant. UH. And it's obviously making guesses in

(32:45):
some cases, but they're very educated guesses based upon huge
volumes of information. Well, Chef Watson is using that same technology,
but in the process of creating recipes uh IBM calls
it cognitive cooking, and it's based upon a couple of things,
the historic uses of various ingredients as well as the
chemical composition of those ingredients themselves and which chemical compositions

(33:09):
would be most likely to complement one another. So it
is going to be learning in large part from recipes,
from recipes that exist. It did. Yeah, they they fed
it thirty five thousand recipes along with a whole lot
of information like like food pairing theories and that that
flavor compound chemistry and just cultural taste preferences. Right, So

(33:33):
that way, with all that information combined, it could start
coming up with really interesting recipes recipes. So they showed
this off. IBM showed this off in in the perfect
way in my mind. They got a food truck and
they sent it to south By Southwest Interactive, which makes

(33:54):
perfect sense as a huge festival, a lot of a
lot of influential people from various businesses are there, and um,
what they did was they gave people the opportunity to
vote on one of several different dishes that had been
invented by chef Watson. So Watson would go through its
data base and say, all right, well, traditionally this ingredient
could be paired with this other ingredient, but I'm going

(34:17):
to do that and because I know that that's going
to produce this one type of taste, that taste would
probably go really well with this other thing that no
one's ever really bothered to combine with these and that's
gonna be my recipe. So attendees would end up seeing
the the candidates and then vote on Twitter essentially do

(34:37):
a hashtag and their choice, and then whichever one got
the most votes would be the one that the the
food truck would deliver the next day, So you could
go to the food truck and get a sample of
that and try it out. Not to be clear, this
food was not prepared by robots. No, no, what Watson
was not in the food truck. No Watson and Spirit

(34:59):
was in the food. Yeah, the Watson technology remained back
at IBM and was connected through the cloud. Uh. The
humans were actually preparing the stuff, but they were preparing
the recipes based off of an artificial intelligence creation. So
this was innovated by a machine, not by a human humans.
This is kind of like the opposite of what we
were just talking about, right, So instead of the robot

(35:20):
preparing these ingredients and making a meal. The robot came
up with which ingredients you were going to use, and
then you created the meal. So this is the model
where you become the robot and just perform out of
blind faith, exactly what the machine tells you to do. Yeah,
you become one of the drones in the nineteen eight
for Macintosh commercial. But hey, you're a robot that gets

(35:45):
to eat some potentially tasty and probably really weird food.
Let's talk about that. Well, all right, let's talk about
some of the examples that they came up with that
the food truck. So uh, they had stuffed like Belgian
bacon putting. I don't know exactly what that was. The
one that they gave the most attention to in the
videos I watched was an apple cobb, Vietnamese apple cobb.

(36:07):
That sounds pretty good to me. It was apples and pork,
which have been used quite a few times like apple pork,
apple sausages kind of stuff. And mushrooms were the other
key ingredient in this, along with some other ones. Um
and the program had identified a common flavor compound and
apples and pork and calculated that if the two combined
that make a pleasant flavor and added the mushrooms in,

(36:29):
and in fact, the people behind it said this was interesting.
I never would have thought to add mushrooms to this
particular dish, but people seemed to react to it positively.
They also had other stuff like coconut caribbean snapper fish
and chips, which I'm down with great. They had a
beef burrito that contained two ounces of dark chocolate, which
I first reacted. Yeah. First, my first reaction was weird,

(36:51):
and then I thought chicken. I put a tiny amount
of dark chocolate into my chili recipe, which that makes sense.
You know, I end up putting in an adult beverage
in my chilly recipe, and I might surprise you. Because
I don't drink adult beverages. It often means that I
am left over with quite a bit of adult beverage
that you can get rid of because they don't sell

(37:13):
it in the small amount I need. Not usually no.
But anyway, if you wanted to give this a try,
you don't have to wait for the food truck to
show up in your neighborhood. Instead, you can actually go
to the website, the Chef Watson website, and you can
sign in using something like a Facebook profile or the
IBM I D and you then can experiment. So it

(37:36):
has a place where you can put in an ingredient
and you can put I think up to four ingredients,
maybe more than that. I just I remember seeing four
on the screen. And you can start with just putting
in a single one and then I'll fill out the
rest for you, or you can go in and fill
in you know, no, I want these two paired together,
and then tell me what I can make from that.
And you can also fill in things that you don't

(37:57):
want to put in. For for example, I really shouldn't
have bell peppers, so I could say no bell peppers, yep. Yeah.
And in fact I did that as well, because I thought, hey,
I might actually use this in the near future. But
one of the people I'll be cooking for can't have tomatoes,
so I have to eliminate tomatoes from that. Otherwise I
would just have to go through recipe by recipe, you know,
and that gets t d s. Yeah, well yeah. IBM

(38:19):
teamed up with Bonappetite to create this app, and they
they've also published a cookbook of Watson's recipes with the
Institute for Culinary Education, which is the team of chefs
that they work together with to develop the software. In
the beginning, recipe I want to try is called Broccoli Frecacy,
which is pretty interesting to me, like it actually sounded

(38:40):
really appealing. It was also kind of odd because broccoli
was the ingredient I put in at the beginning, so
I was like, tell me the kind of stuff I
can make with broccoli. It filled out the other ingredients.
I think it added colligue. Well, first added tomato and
its no, no, I can't have that. Knocked it off.
They replaced tomato with cauliflower, and then I can't remember
what the other two ingredients that it added before it

(39:00):
just gave me the list of recipes. But it was
funny because it's it's called broccoli Frecacy. And then I
looked at the the recipe itself and I was like, oh,
this does sound really good. And then it occurred to me, hey,
I haven't seen broccoli yet. And it turned up that
the broccoli was an optional ingredient in broccoli Frecacy, and
instead of using noodles, which is how the regular recipe

(39:21):
was being presented, it said you could replace the noodles
by boiling broccoli and then mashing it up and then
putting the rest of the dish on top of that.
That would be a base. Okay, so optional broccoli broccoli frecacy. Yeah,
so freacacy, which may or may not have broccoli. So
that's just one example of something odd that might happen.

(39:44):
But that's not the strangest thing we discovered while using
Chef Watson. No. I was like, well, I'll give this
a spin. I am also the cook in our household,
and I feel like I have fairly decent culinary instinks,
so I decided to give it a try. I was
disappointed to find that you can't start with whatever ingredients
you want. They're they're limited by stuff the the you

(40:07):
know database whatever it is, already recognizes. So I tried
to start with rat meat, and it just didn't know
what to do with it. Apparently you're like, hey, what
if I am living in a fallout vault and I
have a limited group of ingredients. That's exactly when something
like this would be really useful. All right, we'll go on.
But okay, so no rat meat recipes. But then I thought, okay,

(40:29):
maybe mustard greens. I like mustard greens. I'll see what
I can do with that. One of the top results
was something called a mustard greens dumpling, like a Chinese
style dumpling with wanton wrappers. Yeah, that sounded interesting to me.
I was just like, okay, I'll give it a look.
And I clicked on this recipe. And I promise I
wasn't doing this to be mean. I wasn't trying to

(40:51):
make Watson cry or intentionally display its shortcoming. But I
think the results were quite bizarre. So the first thing
it tells you to do in the mustard greens dumpling
is to fill the dumplings with a food processor minced
combination of vegetables and seasonings. I quote, finally mince scotch, bonnet, pepper,

(41:13):
and garlic. In food processor add okra and okra, Brian
cured olives, Nisa olives, sesame oil, garlic, chili sauce, and
soy sauce. Well, first of all, let me say, you
can never have too much okra. I like okra, but
it I don't know what to do with the okra
and okra. You have to finally mince it. It tells

(41:34):
you at the beginning of the sentence okay to add
in all the Okra. Uh yeah, well, okay, so I
also like Okra a lot too, But even as a
fan of Okra, this sounds kind of strange to me.
And maybe I'm just not being open minded enough, but
based on my like human culinary instincts, a Chinese dumpling

(41:56):
paste that is literally half by way made up of
brine cured olives, nice Swa olives, and Scotch bonnet peppers
sounds absolutely disgusting. Now you say that, and have you
tried one yet? No, I haven't say that. For those
who are not familiar with Scotch bonnet peppers, they they

(42:19):
pack a bit of a punch. Yes, these are going
to be hot. They're hotter than how many arrows they're Yeah,
they're quite spicy. Yes, these are gonna be very spicy.
I mean I like spicy food. That's fine, But more
than half of what makes up the filling is going
to be olives. And then you're gonna add soy sauce
to it, and chi sauce, which is I'll talk about

(42:39):
an ingredient that didn't pop up. Oh yeah, wait a minute,
I thought I was making mustard green dumplings. What happened
to the mustard greens. Well, finally I kept reading through
it to find where they came in, and finally I
got to the step pour enough water into the pot
to reach depth of one half inch line steamer rack
with mustard greens leaves. So, in other words, the steam

(43:01):
is going to be passing through the mustard leaves, right.
So well, I mean, so if you've ever made dumplings
in a steamer, it's a very common practice to lay
out cabbage leaves or let us leaves or something in
the steamer just to sit the dumplings on while they steam. Yeah,
so they don't stick to it in the end, and
those kind of things can impart a little bit of

(43:21):
flavor to the dough. So the thing that it calls
mustard green dumplings is is might have a waft of
mustard green, but it has no mustard greens as an ingredient.
Maybe what this tells us is that Watson really hates
mustard greens. Could be I just wanted to be tangentially
related to the cooking process. Also, there's another thing that

(43:44):
kind of surprised me, which is that it said this
recipe makes thirty six surveys when they're they're about four
cups of filling, it looks like an appetizer, clearly, Yeah, okay,
And if I don't know, if the filling is mostly
salty olives and so I sauce, that might be for
the best. Anyway, I I don't mean to rag on

(44:04):
it too much. I mean, I think the idea of
creating an AI chef is really cool, and I can
actually see a recipe synthesizer being a better investment, at
least in the short term than a physical robot cook,
just because of all the problems of you know, how
to navigate the kitchen and recognize ingredients and stuff if
you're talking about the physical cook. But to me, this

(44:25):
shows the limitations of a system like this when it's
trying to work with unstructured data, because there, I mean,
I understand that the data that they fed it is
I think somewhat structured, maybe partially structured, but it seems
like it could be more well structured. So in other words,
one of the limitations you would say is that it's

(44:47):
listed as the principal ingredient, but in truth, the mustard
greens are just used to incidental to the recipe. But
but the the AI can't tell that, right to the
AI that because that's involved in the process, it must
therefore be as as important as these other anything else
elements inside the recipe. Yeah, so, in my opinion, I

(45:10):
would think to design like reliably good cooking instructions without
a lot of human compensation, because we can look at
recipes like these and say, okay, I could maybe take
some of the suggestions of this recipe, but I mean
not follow it word for word because that would be insane.
And in fact, they even tell you to do that.
They say, quote, these quantities and steps are ideas, but

(45:30):
Chef Watson really needs you to use your own creativity
and judgment. Let us know how to make chef smarter.
So they, I mean they understand the limitations obviously there.
We're not like pointing out something they haven't figured out
in this to point out cognitive computing. That's how cognitive
computing works, right, is that you tell the computer, hey,
I made this thing. You made It was a monstrosity

(45:53):
and it it made me want to burn my house down.
Don't do that. Then the computer says, got you, good,
but I'll not do it again. But I do think
something like this could maybe work better if you start
with more structured data. Well, personally, before we make such
a bold statement as that, Joe, I think we really

(46:14):
do have to make these and find out what happens. Yeah, okay,
so you want me to make some olive dumplings and
bring them in. I'll eat one if you bring it in. Well,
I like Scotch bonnets. I'll bring the water, you bring
the water. Well, Yeah, it's too bad we don't have
like a farmer's market near by the office, because I
think that would be amazing. It would be we can

(46:36):
shoot a video where we make Yeah, we make the Watson.
We should each pick a Watson based recipe and do
a video of it and say, here's the future of recipes.
I think, do you have any day and dates coming up?
Let's let's you know. I'm pretty much sold for the
rest of the year. But I'm okay, but I think
I think it actually would benefit us from doing it.

(46:57):
I think we need to do this, Okay, Yeah, I'm
totally in. I mean because because this is really fun.
It's such a fun demonstration of how a tool as
powerful as Watson can be used in in perhaps unexpected ways. Uh,
and you know the application of Watson's processing power. The
same application can also help us find more effective treatments

(47:20):
for diseases, or identify drugs that could be therapeutic in
ways other than what they were originally developed for right,
or even find potential problems of different medications interfering with
one another before you actually prescribe them to a patient.
So incredibly powerful tool. And you know, yes, this this
application we're talking about seems a little frivolous, but honestly,

(47:43):
I'm excited by it because I love the idea that
the recipe for the next amazing dish might be created
by an artificial intelligence because it doesn't know, it doesn't
know that it's breaking a rule necessarily. Now I don't.
I don't think the mustard green dumplings, what's gonna be it? Well, hey,
I want to end with some good news. Now. I

(48:05):
did find another recipe under the mustard because it gives
you lots of different recipes to choose from. I found
another one that was called grilled mustard greens, and it
was not for grilled mustard greens. It was actually for
grilled cod served on a mustard green salad. But it
sounded pretty good. I was like, well, maybe I will
give this a try cod with mustard greens. Now, the

(48:26):
picture it had associated with it was a picture of
grilled shrimp, and the first step in the cod recipe
was one word. It said, d vane. I think, uh,
I think what we're gonna need I don't think applies
to God. I think at some point what we're gonna
have to do is create our own Like this will
probably be how the singularity actually happens, y'all. What we

(48:47):
will do is we will create a well, we'll cook
a dish that was created by Chef Watson maybe if
we were you know, the Serenity Kitchen folks are located
all Atlanta. Maybe we'll use a robot to help prepare
a dish that was made by a different robot, and
then we'll take a picture of that and feed the

(49:07):
picture through Google Deep Dream. And that's how the Singularity happens.
All right, I think we've got it planned. You know,
So finally that April Fool's episode we released will become true.
I'll just be a little there'll be a little prematures.
All well, let we'll let you know when we're when
we're pulling the trigger. Yeah, but no, I really do.
I really am excited about trying this, whether we do
a video on it or not, I am excited about

(49:29):
about using this and seeing how it turns out, because
even if it fails, it's interesting, right and uh and
I love again being being a gadget guy and being
the cook. This is the kind of stuff that just
appeals to me all on its own. But I'm curious
to hear what our listeners think. Would you, guys, use
a robot in your kitchen or do you think that

(49:50):
that takes the fun out of cooking? Would you follow
a chef Watson recipe or have you? And if you have,
let us know what the results were like, Um, shove
Watson is a very easy app to you, So if
you want to go out there and give it a shot,
let's know your results. Please do. I would love to
hear it more. Of course, if you have any suggestions
for future topics, you can let us know. Send us

(50:11):
an email. The address is f W thinking at how
Stuff Works dot com, or drop us a line on Twitter,
Google Plus or Facebook. At Twitter and Google Plus, we
are fw thinking at Facebook. Just search f W thinking
in the search bar will pop right up you can
leave us a message and we will talk to you
again really soon. For more on this topic in the

(50:36):
future of technology, visit forward Thinking dot Com, brought to
you by Toyota. Let's Go Places,