AI advances are mind-blowing. AI has now the capability to convert text-to-image. AI generates a unique image from just one simple line of text. It is a very good translation tool for making visual images. It does this by recognising patterns. If we write a sentence ‘A rose plant wearing a hat and goggles’, we can get that image. If AI can create an image from a line of text, in future it can make a video and a whole movie. There are text-to-image AI systems called DALE-2 from Open AI and Imagen from, Google. The building blocks for AI to start making videos and movies already exist. There is Google’s chatbot called LaMDA that indulges in free-flowing conversations on a variety of topics. Then there are advancements in natural language generation (NLG). Together both these can write scripts. These scripts can be subjected to text-to-image, and you get an AI movie; say in next 2-5 years.

A Google’s engineer called LaMDA a sentient which is aware of its existence and is able to perceive things. There is Google-owned firm DeepMind. Its general AI model is called Gato. Most AI systems are trained to perform one task, whereas Gato does 600 tasks. Gato was trained to execute each of these 600 tasks but if it is confronted with a new challenge, it would not be able to use its learning from the 600 tasks to analyse and solve logically that new problem.

AI engines can now generate their own simulated data for different scenarios, e.g. detecting violence on streams in real-time. Here it is difficult to obtain data, but with new image-video generation, it will not be necessary for engineers to interfere to get the data for training as AI will generate its own training data. It can train another AI model for various purposes.

AI is a double-edged sword. Human beings are prone to amoral activities. Text-to-image capability or AI-created content will have a profound effect on the society. Already, we have issues of fakes and deep-fakes. In future, the problem will be more serious.