View the session recording below.
Chief Analytics Officer at OneMagnify, Jonathan Prantner, opened his keynote with a prediction that “looking back, this will be known as the time of the AI revolution.” Although, it comes as no surprise to him because “we were doing this back when they referred to it as statistics.” He described artificial intelligence (AI) as a system that combines data, analytics, and technology to make decisions and act upon them. AI can currently be categorized into three different models: machine learning, computer vision, and generative AI.
Machine learning mimics the process of differentiation and can be traced as far back as Johannes Kepler, who used this model to predict the orbit of planets around the sun in the 17th century. In 2014, extreme gradient-boosted models were introduced and “changed almost every aspect of how vehicles are manufactured.” With this, companies could predict things such as how many vehicles are likely to sell, how much factories will need to produce, when parts will need to arrive, and even down to the optimal temperature at which something will need to be welded at.
The computer vision model was first introduced in the 1960s and has become exponentially more popular since. It transitioned AI from primarily machine learning-based to deep learning, meaning it can mimic the perceptual mapping of the brain’s neural networks. It is capable of object recognition, image segmentation, scene understanding, pattern recognition, and anomaly detection, which enables it to make connections level upon level to identify images present and predict what will be present. This is especially useful in the automotive and mobility industry as it applies to the function of autonomous vehicles, a vehicle’s ability to tell if the driver is looking at the road, and a backup cam that can detect obstacles.
Generative AI is unique because it is being led by sales and marketing organizations that can use the technology to analyze their brand’s presence and audience response and replicate new work that is not only brand-consistent but also speaks to what the audience wants. It is an expert at understanding language and focuses on “sensory familiarity,” using transformer architecture to translate a question into a vector that gets compared to the information it was trained against.
Prantner expects the next phase of the AI revolution to be focused on image production, where an image is “worth” 16×16 words. Although, there is still much progress to be made when it comes to translating language into visuals. “We know that when we speak in language, it’s not the same thing that we see,” said Prantner. To simplify and enhance the processing, engineers are now creating cross-modal models that use text, language, images, video, and audio all at once in the encode and decoder transformers to create solutions.