.png)
Hollywood has long painted vivid and often terrifying pictures of artificial intelligence (AI). From the blinking light of HAL 9000 in "2001: A Space Odyssey" to the menacing cyborg in "The Terminator," and the nearly human android in "Ex Machina," our view of AI has grown increasingly human and, simultaneously, more frightening. These sci-fi depictions, coupled with our human-centric language around AI — terms like "neural networks" and "training"—have only fueled our fears. But what is AI really? Should we be scared?
Machine Learning Engineer Cole Franks, PhD, and Molten Cloud Head of Product, Ethan Shvartzman, help demystify AI, explaining what it is and what it is not.
.png)
Contrary to its name, AI is neither artificial nor intelligent. At its core, AI is a sophisticated algorithm that relies entirely on the data it is fed. It excels at analyzing datasets and predicting what comes next based on that information. Cole Franks draws on the example of a line of best fit — a predictive model familiar to many of us — to explain this concept. Just as a line of best fit can be used to predict someone's weight based on their height, AI models, including advanced ones like ChatGPT, perform predictions on a far more complex level. Franks argues that a line of best fit is as much AI as ChatGPT is. While ChatGPT generates text in a highly sophisticated manner, the underlying principle is the same: making predictions based on existing data. The term “intelligence” is somewhat of an overstatement. AI is an advanced predictive model, but it remains fundamentally a tool for data analysis and pattern recognition. Whether it's text-to-image, image-to-text, or text-to-speech applications, AI operates by processing and predicting based on its dataset.
Like all AI, the core of generative AI is a predictive model. When AI writes an email, an essay, or generates an image, it doesn't create something new from scratch. Instead, it makes educated guesses based on its training data. This generative process involves taking the average of its data set to predict what comes next. For instance, when AI generates text, it predicts the next word or sentence based on the context of the previous words. This principle underscores AI's limitations: it cannot produce genuinely novel ideas but can only extend and recombine existing data. Cole illustrates this idea with a prop replicating 2001: Space Odyssey’s Hal. A giant switchboard on stage is connected to Gridspace’s voice bot, Grace. Cole explains that every time a switch is flipped, another 5% of Grace’s parameters are set to zero, essentially wiping out a portion of her data set each time. Cole and a colleague walk us through the demonstration, keeping Grace in conversation while flipping more and more of her switches until her half of the conversation is reduced to gibberish.
This exercise demonstrates how even an instance of AI as (seemingly) humanistic as a text or speech generating bot, is a predictive model at its core. Without a data set, Grace has nothing to predict the next word of best fit, and cannot produce language.
While AI's most visible impacts might seem flashy, some of its most powerful applications are quite subtle. In the visual effects (VFX) industry, AI is transforming tedious and impractical tasks into manageable ones. VFX expert Jory Federighi notes that AI tools can drastically reduce render times for 3D modeling, automate the creation of 3D textures, and generate fake album art for props. These applications streamline workflows, empowering human artists to focus on creative direction, and increasing the scope of what’s possible to accomplish in a practical timeline.
Similarly, in media operations, AI is revolutionizing backend processes in a powerful way. Molten Cloud’s recently released AI Rights Management System is an exciting example of how AI is being used to reduce or remove tedious human workflows. In this case, AI is automating the generation of rights data from long-form contracts, extracting and mapping data to Molten Cloud’s powerful existing platform for rights and contracts management. It may not sound flashy, but if you’re a distributor you know just how tedious manual tasks like these can be. Molten Cloud’s AI is transforming back office operations, cutting onboarding time into the platform by months, allowing major media companies to digitize their rights in a short time frame, straight from their source of truth (their contracts). The impact on business is astronomical, creating scalable infrastructure and eliminating hours from daily and weekly workflows.
Understanding AI beyond the myths and fears shaped by Hollywood is crucial as we integrate this technology into various facets of our lives. AI is neither an independent thinker nor a creative force; it is a powerful tool designed to predict and generate based on existing data. This predictive nature, while not as sensational as sci-fi depictions, is what makes AI incredibly useful for practical applications. From enhancing the efficiency of visual effects workflows to revolutionizing backend media operations, AI is transforming tedious tasks into manageable ones, allowing humans to focus on creativity and strategic decision-making.
By demystifying AI and recognizing its true capabilities and limitations, we can better harness its potential to augment human efforts. AI's power lies in its ability to analyze vast amounts of data and generate useful predictions, making it an invaluable asset in fields ranging from media production to business operations. As we continue to develop and refine AI technologies, our focus should remain on leveraging these tools to enhance human capabilities, streamline processes, and drive innovation in a responsible and informed manner.