Normally, when I cook, I’m on my own in the kitchen. Today, I have a sous-chef, and it’s a supercomputer.
The recipe for my lunch has been supplied by Watson — yes, that Watson, the IBM-made, Jeopardy-winning computing system. Watson has swapped its quiz show contestant’s get-up for chefs’ whites, and has been getting down and dirty in the kitchen.
The dish (which is shown above) came in three parts: a sweet potato puree flavoured with celery, ginger and garlic; crispy fried and spiced sprouts; and a gremolata topping. The signs weren’t good – while I love brussel sprouts, I always confront any dish containing sweet potatoes with a mix of fear and suspicion.
As part of a recent initiative called Cognitive Cooking, Watson has been putting its smarts to work by dreaming up new recipes. Which is why I’m now setting to work cooking up a dish of Kenyan Brussel Sprouts Gratin. And no, that’s not a typo.
I’m not the first person to cook up one of Watson’s quirky dishes. At this year’s SXSW festival and IBM’s recent Pulse conference, a team of cooks served up dishes based on recipes generated by Watson and chefs, including Vietnamese Apple Kebab and Belgian Bacon Pudding.
The dishes may seem a little off kilter, but that’s the plan: the system is designed to come up with novel dishes — surprising combinations that might not have occurred to human cooks. IBM’s work was inspired by the idea of trying to make machines think creatively.
“Creativity is often considered as the pinnacle of human intelligence so that sounded like an interesting challenge,” said Florian Pinel, a senior software engineer at IBM.
The Cognitive Cooking version of Watson uses three criteria to devise the recipes:
1. Surprise — How similar the recipe is to others already out there
2. Pleasantness — How good the dish is likely to taste and smell
3. Pairing — How well the different flavour compounds in the dish will work together
The latter is based on the theory that the more flavour compounds ingredients in a dish share, the better the dish will taste, says Pinel. It’s an idea common in Western cooking, but less so in Eastern cuisine, where it’s more common to have flavours that share fewer compounds.
Still in its early stages, the system has three basic inputs users can control: selecting a choice of ingredient, a particular country’s cuisine, and a type of dish. Once the parameters are set, the system will offer up tens of potential suggestions for dishes, which can then be narrowed down according to preference — more surprising dishes could be bumped up the list, or dishes that are less surprising, but have a higher chance of tasting good, could be selected.
Eventually, the system could one day be packaged up and offered to home cooks as an app, able to take information from the user’s Instagram feed or their grocery list, and customise its results accordingly to make meal suggestions. Does every smartphone snap show you’re ordering a burrito? Knowing it’s your favourite food, Watson’s at-home recipes could be customised to give a hundred variations on the burrito, inspired by cuisines from around the world; equally, knowing you’re getting heartily sick of eating the same thing day after day, it could steer you away from recipes that feature Mexican flavours or come wrapped in a flour tortilla.
For now, the system relies very much on cooperation with humans — it may be able to tell you what ingredients should feature in your Honduran avocado pie, but it doesn’t advise on quantities, cooking times or methodology. That’s down to human chefs to come up with.
And, at the other end of the process, Watson can’t cook the dish either: that’s down to the person in the kitchen. While Watson can offer up a recipe with input from smart machines and skilled chefs, the success of the dish is still down to the human in the kitchen, as I was reminded in polite terms when I asked IBM for a Watson-generated dish I could try cooking, to see how the system’s Willy Wonka creativity worked in a home cook’s kitchen.
For me, Watson came up with a Kenyan Brussels Sprout Gratin. On that title alone, Watson knocked it out of the park on the surprise factor, but did it deliver on the taste side of things?
After a few minutes of inhaling the scent coming off the frying puree ingredients, I was starting to think maybe Watson was onto something. By the time I started on the second part of the dish, I realised that brussel sprouts and cardamom were like the Romeo and Juliet of foods – a beautiful pair, destined to be together, kept apart by cruel forces that didn’t understand how right they were as a team. Sure, the gremolata had enough garlic to keep away vampires, and likely most humans for that matter, but I couldn’t help but think that I’d been judging sweet potatoes too harshly once I got to try the finished dish.
Maybe when it comes to working out what veggies go well with each other, a huge computing system has an advantage on a mere human and its fallible palate?
A new generation of foodies
Technology has always played with our food. When chrome-clad homes of the future began turning up in newsreels of the 1950s and 60s, technology was the avatar of progress – what a brave new world awaited us if the cutting edge of science could even revolutionise the humble kitchens.
Today in the West, cooking is seen as part chore to be automated, part symbol of affluence or otherwise. Still, technology plays its part in how we prepare, eat and enjoy our food. Instagramming your lunch al desko being perhaps the most obvious (and most derided) example of where food and everyday technology meet.
While we may have long been happy to hand over the grunt work of food preparation to machines — food processors have been around since the 40s, and their predecessors the food mills for centuries more — soon, it looks like they’ll not just be doing the dull stuff we don’t want to, but helping us execute our more creative culinary visions.
As prices for 3D printing kits drop, companies are beginning to see potential for the technology in consumer, as well as industrial, markets. Chocolate was one of the first foodstuffs to get attention from researchers and academics looking into 3D printing possibilities, and now machines targeted at home cooks who want to print their own sweet stuff are hitting the market.
With price tags running into the several thousands of dollars, for now they look to be the preserve of chocolatiers looking to give themselves an edge with sweet design, or marketing companies creating bespoke, and edible, collateral.
While not cheap — it will be priced around $1,300 — the Foodini 3D printer is aimed at home cooks as well as neighbourhood cafes, and can print more than just sugar fixes. Once pureed and put into Foodini capsules, food can be extruded into the shape of the user’s choosing. The idea behind the Foodini is thatusers can print out sweet and savoury foods, including pasta, pizza, biscuits, and quiche, but using fresh ingredients.
“You might ask, why would you want to print that? You can get that in the supermarket,” said Lynne Kucsma, cofounder of Foodini’s maker Natural Machines. “Well, that’s the point. If you ever read the ingredients list, it’s not exactly healthy, there’s a lot of preservatives, and even if you go for the organic version, there’s still a lot of salt and oil, and you have preservatives because it’s a packaged food, so it’s much better to print your own. Even though crackers and breadsticks might sound simple, have you ever made those things by hand? They’re not simple to make.”
The first generation Foodini will be rolling off the production line later this year, and for now is chiefly a 3D printer that can help you with your food presentation rather than one that will do away with the labour of making a meal. In the future however, the company is hoping to expand the features of the machine so that it can cook the food it prepares. In the first generation of the machine (shown printing pizza below), printed pasta still needs to be cooked, and printed marinara sauce would still need to be warmed up in a saucepan. Potentially, later versions of the machine could deliver both elements, fully formed and steaming hot, on a plate, inching one dish closer to a real life Star Trek replicator.
RFID chips, too, could be added to capsules in future: if a user inadvertently put the wrong capsule in the wrong slot — pasta dough where the filling should be and vice versa, for example — the machine would be able to read the chips and correct its operations before accidently turning out frankenpasta. Presumably, it could also be used for health and safety purposes too, like scanning the chip to make sure that the capsule is still in date.
One day 3D food printers could go even further, swapping data with connected fridges in users’ kitchens to adjust recipes according to what’s available at the time, or placing the ingredients for a planned meal into your online grocery shopping basket ahead of time. (Such capabilities have been talked about by electronics manufacturers for some time, but their clunky implementation means they’re more theoretical for now.)
Connected devices could also offer users the chance to make restaurant food at home — as 3D food printers get more advanced, they could potentially download recipes devised by local eateries or celebrity chefs, order the right capsules, and have the whole lot ready to go by the time you get home from work. Takeaway schmakeaway.
Both download-a-chef 3D printing and IBM’s Cognitive Cooking experiment raise interesting questions about the nature of creativity in cooking.Talk to the providers of the tech and they’ll tell you that chefs are enthused about the possibilities that technology can open up, that it will be a complement to their skills and that, even in a technology-heavy restaurant industry, their creativity will still be needed.
In one restaurant in China, however, man is already starting to be displaced by machine. In the Robot Restaurant in Harbin province, dishes are prepared by robot chefs and delivered to the table by robot waiters.
Not only does it draw in the punters but it also helps cut down the restaurant’s wage bill, according to its manager. The robot workers may be a fun gimmick, but they’re more like the 1930 visions of robot butlers than autonomous or creative chef-replacements — humans are required to prepare the raw materials that the robots cook, and diners have to take their meals from their robotic servers themselves, as they lack the motor skills to move the plates from the tray their carry.
Elsewhere, however, more conventional technology that’s likely to have an impact on staffing levels is making its way into the food industry.
Tablet table service
What looks to be the biggest rollout of tablets in the food industry was announced last year by US chain Applebee’s, which has over 1,800 outlets in the country.
By the end of 2015, the company expects to have 100,000 tablets installed in its restaurants, where diners will be able to add extra dishes to their orders, play games, and pay for their meals via the Intel-powered devices on their tables.
The move was prompted by a need to cut wait times. “Let’s face it, everyone who has ever been to a restaurant has been frustrated by waiting for their check,” Applebee’s president Mike Archer said when the news was announced in December. “Starting out, our goal was to create a way for guests to control when and how they pay their check. What we learned after nearly two years of testing is we can provide much more. The Presto tablet will deliver our guests a robust slate of offerings for not only transactions, but entertainment, social interaction and more, moving forward.”
According to the company, a pilot of the tablets led to “significantly reduced transaction times for guests.” Applebee’s is now planning to rollout the tablets “aggressively” over the year, though there’s no hard details on the deployment timetable yet.
And it’s not just chains that are investing in the technology. Asian-fusion restaurant Inamo, which has two locations in London, has had a human-free ordering system since it opened in 2008. Diners can order food and drinks by way of a mouse-pad built into the tables. Using overhead projectors, guests can scroll through pictures of dishes, place their orders, request the bill, and even play a game of Battleship against their dining companion.
Restaurants are using tablets to assist waiters with ordering, tab payments, and bar games.
Adding such ways of ordering will presumably have other benefits in the longer term as the technology becomes more advanced — fewer slips-ups with the orders, an ability for those with food allergies to check the contents of a dish, and advice on expected wait times for those in a hurry.
But it could equally lead to fewer wait staff in restaurants. If the ordering and payment is handled by machines, the workload for servers could be cut to just bringing food from kitchen to table and collecting empty plates. Bad news for the huge numbers of serving staff employed around the world, over 2.3 million in the US alone.
Restaurants often pitch the addition of technology as a way to “enhance the dining experience” or process more orders in a given time, particularly in the fast food sector, rather than as a way to cut headcount. However, for chains as well as small restaurants, the appeal of minimising the wage bill is likely to be undeniable, particularly as diners appear to favour fewer dealings with staff. Applebee’s reported that in the pilot phase of its tablet experiment that diners reported a “better experience overall” with the tablets than without.
In the future it’s quite possible that restaurant staff will mirror the transition of those in supermarkets where self-service tills have been introduced, overseeing several diners placing their orders electronically, and stepping in in the event of problems. An entirely server-free restaurant, however, is unlikely — not everyone is a fan of touchscreens over the human touch, particularly those who are traditionally not huge adopters of tech, such as some older diners.
Many organisations are targeting the demographic nonetheless, including the Performance project, which recently won European Commission funding to help develop food which has had its structure altered in such a way to allow people with chewing and swallowing problems, normally only able to eat mashed food, to eat a more visually appealing meal.
The genesis of the project was a concept developed by Germany company Biozoon, called smooth food. Typically, for those with mastication or swallowing problems, eating means a succession of pureed food.
“The Presto tablet will deliver our guests a robust slate of offerings for not only transactions, but entertainment, social interaction and more, moving forward.”
MIKE ARCHER, APPLEBEE’S PRESIDENT
“In nursing homes, people that suffer from mastication and swallowing problems get served mashed food, and it’s unappealing looking. This is suitable for them to eat, but they get this kind of food everyday and the joy of eating is not there anymore,” Sandra Forstner, R&D Project Manager at Performance, said.
Biozoon developed a texturising material that could be added to the mashed food. Once the purees were put into silicon moulds and cooked, they resemble the food they originally came from, breaking up the monotony of texture and appearance for diners.
However, for nursing homes where a higher number of residents have special eating needs, making tens of individual dishes can take too much manpower to be efficient.
“We thought maybe we can somehow industrialise [production] and, what’s more important in this case, personalise it. You can’t force someone to eat more than they want but you can increase the calorific value — maybe someone needs smaller portions sizes — but we can enrich it somehow with fats or supplements. That’s where we said, ‘Okay, how can we produce personalised meals simply?'” Forstner said.
Performance, a coalition of companies of which Biozoon is a member, looked to 3D printing, and is now working with a company that’s adapting its existing food printers, usually used to print thin layers onto flat surfaces (think icing on biscuits), to print 3D structures out of the smooth food. The smooth food will eventually be packaged up — a mix of vegetables, meats and carbohydrates — and sold as microwaveable meals that can be delivered to nursing homes or people at home.
It’s hoped when the meals go on sale in the next few years, they’ll be able to be tailored to a customer’s individual requirements: the portion size they need, any additional micronutrients they might require, how robust a texture they can handle — perhaps even adjusting flavours themselves to suit a particular palate.
From farm (to cloud) to fork
An increase in 3D printed food tailored to consumers on a very precise level could also be beneficial to individuals’ health in a more simple way: by allowing them to keep tabs on what they’re eating.
Data from 3D printed food could enable users to build up a day-to-day picture of the micronutrients and calories they’ve taken in, helping with monitoring of health conditions like diabetes as well as offering a tool for weight control. Until the technology becomes both more sophisticated and more widespread however, there is no shortage of companies offering apps for the same purpose.
The SCiO food scanner connects to a smartphone and shows data about the makeup of food.
Image: Consumer Physics
Currently, they’re fairly simple affairs, often just a smartphone-based replacement for writing down what you ate all day. Food intake is recorded through a text diary or photo journal feature. The rub with food monitoring apps is that much relies on manual input — the user taking down the ingredients for the meal they’ve consumed — with the app then calculating the calories. For a single food like an apple, that’s not so hard. For something with more ingredients, like a black forest cake or a mixed salad, the caloric and nutritional range can be wider, and would require entering each ingredient manually and guessing the amount of each element that the dish contains.
One solution to the problem could be greater use of image recognition software — building in features that would let users snap their meal, and then have the app work out the fruit it’s looking at, its size and calorific value. However, there are naturally limits to what photo recognition can cope with: a vegan cake would look much like a dairy-laden equivalent, oven chips would resemble fried chips.
There are several startups that are already turning to higher-tech ways of answering the problem. Consumer Physics for example, has created a food scanner that uses near-IR spectrometry. The scanner, called SCiO, just raised over $2 million on Kickstarter.
The handheld scanner works by directing a beam of near-IR at a piece of food. Each functional group in the food molecules absorbs a different part of the near-IR spectrum to a different extent. Water would have a different absorption pattern to sugar, for example. The near-IR that bounces back to the scanner is then analysed to work out the composition of the food it’s pointing at, and the nutritional details displayed on an mobile app.
What all of these systems, low and high tech alike, have in common is that they’re pitched to potential customers on the grounds that, by getting a greater insight into eating patterns, individuals will be able to make better decisions about what they choose to eat.
For those just hoping to keep their weight under control or eat more healthily, does knowledge really equal power? Does knowing a pint of beer is 500 calories and a kebab 1,500 mean we won’t have one after several of the other?
“It’s not enough to collect this data… if you’re writing down everything you’re tracking on your weight, if you’re not regularly looking at how much you’re taking in and thinking about ‘Wow, what happened that day why did I get so much more calories on that day?’, if you’re not regularly doing that, you’re not going to be successful in changing your behaviour,” Dr. David Cooper, a psychologist with the Mobile Health Program at the National Center for Telehealth and Technology, told a Self Spark lifehacking event recently.
“Food safety concerns are also driving the technologisation of food.”
MANDY SAVEN, STYLUS
Gathering that data is one thing, but using it as a prompt to change behavior is another. Technology can only provide you with the information, but it’s still down to the user to review it, make changes, gather more information, and repeat the process. “It’s all about feedback and generating that feedback loop,” Cooper said.
For those with a food allergy, however, gathering very simple data can still be hugely useful. Another startup TellSpec is working on similar food scanning tech in the form of a handheld spectrometer and combined with a mobile app, giving users an insight into the nutritional make-up of their food — the calories, sugar, and fat content — along with the presence of allergens like gluten. Their feedback loop is more likely to be a short one: I have a nut allergy. Does this have nuts in it? Better not eat it then.
But for the allergy-free, tech can still be used to improve food and drink safety. Companies are working on food labelling that can give consumers further insight into the history of their food as it travels from farm to fork (or factory to fork).
Whether it’s high or low tech, food safety will be driving the need to put more technology on our plates.
“Food safety concerns are also driving the technologisation of food,” Mandy Saven, head of food, beverage & hospitality at Stylus, said. “Towards this end, we see restaurant chains utilise technology to convey provenance and authenticity of ingredients. Harney Sushi restaurant, in the US, actually places QR codes – produced using soy-based ink on edible paper – on pieces of sushi, thus enabling diners to immediately validate the sustainability credentials of the food via their smartphones.”
Other startups are working on food safety tech, mostly aimed at manufacturers and retailers themselves, but over time as mass production ramps up and costs come down, such labels could end up on food at item level, allowing consumers to see for themselves if their groceries have made it from producer to plate in the conditions it should have.
Thinfilm Electronics, based in Norway, uses printed electronics to create labels that provide details on whether the item it’s attached to has been transported outside of certain parameters. For example, if boxes of milk were stored at temperatures that were too high to prevent the growth of bacteria.
A food delivery robot shows off some goods at a robot restaurant in China.
Image: Cui Jian/ChinaFotoPress
The printed screen on the label will display details on how far out of a designated temperature zone the product went — whether that’s too high, too low, or both — over a period of time. The data can then be exported for analysis, the theory being that that can provide more detail than the traditional chemical-based labels that will change colour once a threshold has been breached. Spanish startup Laserfood meanwhile is offering technology that allows details to be inscribed into fruits or vegetables — sell by dates, brands, or the producer’s name — with the aim of increasing traceability.
Other tech tools aiming to alert consumers to areas of concern around food safety can range from the prosaic to the futuristic. On the former side: supermarket loyalty cards. By scanning the shopping histories provided by consumers every time they pick up the groceries, supermarkets can target food recalls very precisely — a recent recall by a UK supermarket involving jars of olives suspected of containing particles of glass was sent out to all loyalty card holders whose histories showed they had bought the product in question.
The future of meal time
Other more familiar technologies are also encroaching into our meal times. While cafes have been quick to offer Wi-Fi, restaurants are being encouraged to do the same by food industry analysts — a SOP perhaps to couples on an awkward first date, Instagram-addicted diners, or teenagers who’d rather not actually speak to their parents at meal times. It’s also a reflection of how meals are conducted in people’s houses.
“At home, we find that pervasive use of technology is impinging on the family meal,” Saven said. “An always-on mentality, spurred on by smartphones, means people are engaging with content on their digital devices even when breaking bread together. There is a disconnect here. Mealtimes are inherently social occasions but technology can isolate those even in close proximity to each other.
“But equally, technology can also be an enabler and connector when it comes to connecting remote family members. We’ve seen the emergence of Skype or Google Hangout dinners, where remote friends and families cook and eat together – virtually.”
When it comes to food, everything old is new again.