Revolutionizing Retail & Dining with Generative AI: A Conversation with Markus Stripf of Spoon Guru

Are you ready to revolutionize the way you shop, eat, and live? Discover how generative AI is transforming the retail industry in our insightful discussion with Markus Stripf, co-founder of Spoon Guru. Markus explains how AI-powered chatbots can provide more intelligent, personalized experiences for consumers and shares an example of planning a dinner party with specific dietary needs.

But what about the challenges faced by open-source models like ChatGPT in ensuring accuracy and reliability? Markus reveals how Spoon Guru and Google's partnership aims to overcome these challenges by providing transparency, accuracy guarantees, and evidence-based advice. Don't miss this fascinating conversation exploring the future of Gen AI and its potential impact on the food world.

Transcript

Phil: 0:15

Welcome to Lost in the Supermarket. There's no doubt that what you've seen is a lot of controversy over what's called ChatGPT or Gen AI. Well, today we've got an expert with us, like a strip, who's the co-founder of Spoon Guru and his association with Google is about ready to launch something that's going to change supermarket buying and supermarket shopping and supermarket recipes forever. So, Markus, it looks like just about every day all the headlines are about ChatGPT, about Gen AI, about all this new artificial intelligence that's going to change the world. Number one, tell us about what it is. And number two how is it going to affect the food world?

Markus: 1:02

Great to see you again, Phil. Yes, so Gen AI and ChatGPT, and Bart, i've been dominating the conversations in the last few months. Really, since ChatGPT launched in earnest at the end of last year, it has been front of mind for consumers, but also for enterprise applications. It's a really exciting space. Whether you're in technology or not, it really is transformative. It's going to change pretty much every aspect of our lives. We can use it. You've seen the press coverage. It's being used to write articles, it's been used for research, it's been used for analysis, for coding, for all kinds of things.

Markus: 1:47

The generative bit in AI is really interesting because AI has been around for a long long time. The generative bit means that the AI, the computer basically can interact with you, and that opens up a huge number of use cases specifically for retail, because until recently, and even with Google search results, you ask Google a question and it returns a number of results And then you pick the most relevant to you. What generative AI allows you to do is you can keep going back with additional questions, and I can give you an example. In fact, one of the scenarios we are working on for grocery retailers is to enable chatbots to be more intelligent. So imagine a scenario where you go to your retailer of joys and you ask the chatbot a certain day and say hey, i'm planning a dinner party tonight, but three of the people who have invited have some kind of requirements One has an allergy, one is vegan and one doesn't like mushrooms. If you punch that type of requirement into chatbot, most chatbots will already fail to respond with a relevant option, with a relevant answer, and generative AI will allow us to do is. It will not only empower the chatbot to find suitable results. You can then keep going back.

Markus: 3:19

So in this example I've just given, imagine the chatbot then says OK, i understand an allergy. One of them is vegan, one of them doesn't like mushrooms What kind of allergy is it? And then you can say that person doesn't like gluten, and then it starts returning results, and then you can continue to interact with the chatbot. You can say well, that's great, but I've got some garlic and some spring onions in the fridge I'd like to use, and the chatbot will then again use that information to hone the response it comes back with, and I think that type of scenario will be transformative for the industry. I think it's what we as consumers want. We want more relevant and personalized experiences, and I think it will transform how retail itself will interact with consumers and empower consumers to have more rewarding and satisfying experiences.

Phil: 4:16

But with all the press that it's received with this. There's also been a lot of negative press that says that the results have not been that great. In fact, if we look at eating disorders, the Eating Disorders Group dismantled their live human helpline and they used a generative AI and actually it's called TESA. They had to take it down because it was giving out wrong information. So, in the partnership between Spoon Guru and Google, what are you doing to make sure that the information is 100% accurate?

Markus: 4:57

It's a great question And one of the big, big challenges with open source models like OpenAI and chat GPT is based on OpenAI. One of the big challenges they have is and users have using that services is you don't know where the data is coming from. It's open source. Chat GPT has been trained until the end of 2021 with billions and billions of scripts and articles And it uses that library to respond to questions, and when you see a response, it may sound credible and impressive, but unless you're a subject matter expert, you don't actually know how accurate that response is. And that is a big problem, especially in a nutritional context or in a medical context, because the wrong information can have some serious consequences for individuals. So the way we are approaching this on the Google framework is we are a verified data source. We will ensure, first of all, we will ensure that there's 100% transparency, so you know, if you give your advice on products or recipes, that you can understand how we've arrived at that piece of advice. We will credit the source if it's required. We will also give you an accuracy guarantee. We will ensure that any advice we give is evidence-based and is based on sound scientific principles. With Chat GPT I mean, you've just mentioned it, there have been many documented examples where nutritionists and dietitians tested the service and they found some alarming results which could potentially have severe consequences for people who are affected by allergy or intolerance, for example. So that is one of the big challenges.

Markus: 7:00

The other big challenge with large language models, which is the technology behind Chat, gpt and BART, is that it is a strange phenomenon, it's called hallucinations, and sometimes they go off. You know, these large language models or these generative AI tools, they go off and they seem to make up stuff. I mean, even we tested that on our own company, spoon Guru. We asked to give us a history of my company And it started to come back with some very credible information on the founders And then, all of a sudden, it started making up things about us, about our qualifications, where we used to work, and it is crazy because there's no reference, you know, no cross-reference at all. It's just utterly made up information. So, again, if you rely on these types of services for adequate advice, you need to be able to rely on it. So transparency, accuracy and, of course, convenience are absolutely key, and that is how we are approaching gen AI use cases, using the Google framework to ensure that consumers can have 100% confidence in the accuracy and in the quality of the responses that we receive.

Phil: 8:19

So when I again pick up the newspaper or look at TV, it seems like just about everybody who's in tech are saying wait a minute, we need to slow down on AI because we don't have the controls in place yet. Are these the kind of controls that they're talking about? About accuracy, About traceability? Is that the big challenge? And it seems that Spoon Guru and Google have overcome that challenge with what you're describing.

Markus: 8:51

Yeah, i think it's important that you architect a solution that is based on verified data sources, especially in an enterprise context. If you're a retailer and you use this type of technology in an enterprise context, they see a risk or error or they see a room for error, because the potential damage to your brand is just too big. So the idea to have a framework that consists of verified data sources so you know exactly how the response or how the answer is being constructed, is really important. For example, would you want an Instagram influencer to give you medical advice or nutritional advice? Probably not, or at least you should know.

Markus: 9:35

Okay, here's the answer Generative AI Software has come back with. At least you should be able to know. Okay, it's based on some scientific Journal that was published on Some on some medical advice, or it wasn't Instagram up, but at least you want to know where that piece of information has come from. And that's why we're excited about developing our capability within the Google framework, because it it does put guardrails around our information is being processed and ultimately Made available and exposed to, to consume us and uses.

Phil: 10:16

So I understand the example that you gave us for a recipe Putting in. You know the food allergies putting in, you know all, all the attributes that people are looking for. What comes next? look into your crystal ball as we look at Gen AI Where is? where is spoon guru gonna be? Where's Google gonna be? Where are we, as consumers, gonna be as we do our supermarket shopping in a year or two years or three years from now?

Markus: 10:45

I Think, you know, i think personalization has been such a buzzword for over ten years now within retail, and I think we will see through personalization coming into play, where you can have highly complex requirements as an individual or as a household and you will get a highly curated and tailored experience and not just online, wherever you are. And I think, again, that's what we want as consumers, because it is just too complicated. You know, i, my household, for example, i don't eat meat, my wife has a gluten intolerance and There isn't a single retail on the planet that provides, you know, a tailored, curated experience for our needs, and that is frustrating. On top of that, we now have health preferences, health goals, requirements, where we need to be able to help people reach their goals and make better choices for their own individual health or the health of the planet, and I think AI and generative AI will Allow us to build capabilities that support consumers and retailers really to provide much, much better and enticing and exciting shopping experiences.

Markus: 12:05

Fundamentally, i think the three pillars will be accuracy, transparency and convenience. I think that's what we want. I want the experience to be accurate, obviously, because I don't want my wife to have an allergic reaction. I want transparency. I really want to understand what basis the advice is given, but also I want convenience. So that example I gave earlier about recipes once I start interacting with a chatbot and it has found the perfect recipe for my dinner guests tonight, i want to say make it shoppable, add it to my cart and deliver it to me within the next few hours so I can get cooking. I think those types of scenarios, which I've been discussed for a long, long time, as you know, will find and become reality for shoppers around the world.

Phil: 12:56

Well, Markus, as always, i love your crystal ball, i love what you're doing and can't wait to be able to log on to my favorite retailer and see what Spoon Guru and Google has come up with. Thanks for joining us today on Lost in the Superm arket.

Markus: 13:11

Thank you, Phil, great talking to you.