From Deep Dreaming to Election Predictions: Deep Learning Changing the Game across industries in 2024
By Dr. Abhinanda Sarkar, Academic Director, Great Learning
In the progression from Machine Learning to Deep Learning to Artificial Intelligence, 2023 was a landmark year. From discriminative tasks such as sentiment analysis based on text and facial recognition from images, applications became generative as exemplified by ChatGPT and DALL-E. More impressively, training data went beyond one-dimensional text and two-dimensional images to applications like ready-to-execute software code and ready-to-present business reports. So, what does 2024 hold in store beyond natural language processing and computer vision, and what innovations are on the horizon?
Deep Learning is being used to create new ways in which data relates to human imagination and comprehension. An example is “deep dreaming” where creative images can be generated amalgamating imaginative scenarios, often in the realm of fantasy. For instance, dragons can incorporate biological features learned from animal images in existence and they can fly over winter landscapes based on the creator’s vision. While computer-generated artwork is not new (think of your favorite animation movie), data-driven artifacts provide reality to the surreal – and it can all be done in a matter of seconds. Art and entertainment promise to be within reach of all who can dream.
In terms of human understanding and comprehension, we think of school and college. Psychologists have long posited that learners have different kinds of intelligences – visual, logical, etc. The challenge has been to efficiently incorporate these into personalized learning journeys. Deep Learning is showing a way. Assignments can be readily generated in multiple styles and at differing levels of difficulty. Grading can be commensurately automated. All this is already in place. What is next is the individualization of entire learning journeys, not just assessment. What content is to be emphasized can be tailored to the receiving student’s strengths and weaknesses. To each, his or her own curated textbook.
Another dimension Deep Learning excels at is dealing with complexity. This is because a defining feature of the technology is the ability to process different kinds of information and synthesize. It is turning out to be valuable in industries that have such multi-modal information. For instance, finance. Data can be in the form of balance sheets, stock prices, analyst’s reports, fraud indicators, and many more such formats. Predictive algorithms that aggregate all this into market projections and specific risk scenarios have always been valuable. But now they can be quickly generated and be comprehensive in data usage. Despite economic turmoil and financial instability, we should be safer.
In 2024, the two largest democracies in the world – India and the United States – will both have national elections. The complexity of electioneering is fertile ground for Deep Learning. Political parties can process data from, say, social media to see which way winds are blowing and then generate content to retain the faithful and convert those on the fence. Psephologists can analyze demographic trends and historical voting patterns to generate nuanced predictions at granular levels and election outcomes more broadly. But with the great power of Deep Learning comes great responsibility. Separating fact from fake news will become more challenging as well.
All this is clearly pointing to a new skill set built on Machine Learning. Some components are:
(i) the ability to work securely with distributed and federated data,
(ii) comfort levels with model workflows, particularly on the cloud,
(iii) building user-friendly applications, often on mobile and edge devices,
(iv) communicating the methodology and the results, i.e., “explainable AI”.
For the more experienced, working with and creating large language models and their associated tokens, embeddings, and transformers are already opportunities – this will grow.
The new kind of AI that Deep Learning has spawned is already changing the landscape of work and much is being written about this. What seems to be undeniable is that professions once thought of as creative and non-technical (literature and music included) are impacted. Professionals in all fields are likely to need a level of familiarity with Generative AI, much like email and cell phones. A broad level of basic upskilling seems to be in the air for 2024.
