Making soup together
Imagine you’ve just walked into in a warm kitchen, and you’re immediately surrounded by the delicious aromas of simmering ingredients. Your nose is drawn to a big pot of steaming hot soup that’s sitting on the stove. Just like that pot of soup, the world of AI has been bubbling away, evolving from simple data and algorithms into a sophisticated, tasty creation.
Cooking is an art of transformation, where humble ingredients become something extraordinary through knowledge, technique, and time. While anyone can prepare a basic meal, it takes time, effort, and experience to make something sublime. Like master chefs perfecting a complex dish, the creators of AI have been carefully developing their recipes, and their skills, combining ingredients in increasingly sophisticated ways. And now that the results are starting to taste pretty good, everyone's curious about what's in the pot.
Generative AI has been simmering like a soup for a long time. In the early days, back in the 90s and even up until the early twenty-teens, the outcomes of AI were quite bland and lacking in flavour. The basic ingredients were there – neural networks, pattern recognition, statistical analysis – but they hadn't been combined in ways that created anything particularly appetizing. Sure, AI could read postal codes and detect fraudulent transactions, but the early algorithms were like a simple soup that was only tasty to a few experts: technically interesting, but not the kind of thing that appealed to the average palate.
Over time, the chefs of the AI world came up with better recipes, adding key ingredients like deep learning, and refining their skills with improved training techniques. When they added powerful computational processing it acted like a pressure cooker. AI mastered games like chess and go, but it was like a high-end consommé – technically perfect, admired by connoisseurs, but still a bit intimidating for the average person. All the while though, more ingredients were being added to the broth – natural language processing, computer vision, reinforcement learning (which is how AI learns by trial and error, like a child figuring out how to ride a bike) – and the flavors began to meld in unexpected ways.
As these different "flavours" of AI started combining and reinforcing each other, something magical happened – in the same way combining separate ingredients can suddenly create an entirely new taste. Years of experimentation, refinement, and unexpected discoveries led us to today, where AI has gone from being good at playing games to creating art, writing stories, predicting how experimental proteins will fold, and solving real-world problems. All of a sudden, it seems like we have hundreds of great soups to choose from.
But what went into those soups and how were they prepared? With AI, the algorithms are like the recipes – the computational equivalents of temperature, timing, and technique. Most use the same basic recipes but each employs special little touches, like a chef's secret spice blend. These basic recipes are often shared as academic papers, so others can learn and improve upon them. Yet not all parts of the recipes are shared, and some techniques remain closely guarded secrets. This is why you can try the same prompt in five different LLM chatbots and get very different responses.
The quality of the ingredients matters too. With AI today, there's increasing scrutiny about what went into training the models. Who gathered the ingredients, and how careful were they to ensure nothing toxic got added? Did the chefs get permission to use all the ingredients, or were some of them stolen by their suppliers? Beyond just avoiding harmful ingredients, there are questions about nutritional value – does the AI provide genuine substance and utility, or is it just empty calories? For most of these questions, we have to trust the chefs. Just because it tastes good doesn’t mean it’s good for us.
At this point, the soup tastes pretty darned good. And while some people are rightfully concerned about the ingredients, most are just happy to enjoy the warm, delicious results. Looking ahead, we can imagine even more sophisticated recipes being created – perhaps AI systems that truly understand rather than just combine, or ones that help solve our most pressing global challenges. The kitchen is getting warmer, the aromas more enticing, and hopefully there’s a seat at the table for everyone. And as we sit at this table of technological transformation, we should be asking ourselves: where is this meal taking us next?
If you’re still curious about the soup being served and haven’t tried it yet, now is a good time to grab a ladle and start tasting. Explore generative AI tools, understand the ethics of these ingredients, and think about how you might contribute your own spices to the evolving recipe. The wonderful thing is, just as some soups taste better the longer they’ve been on the stove, the AI we’re enjoying today is still getting better. So I hope you’re hungry. The best dishes have yet to be served. 🍲
Images created with Midjourney. Editing assistance provided by Claude 3.5 Sonnet and ChatGPT-4o with canvas.