This Year India Wished Mass-Diwali Greetings with Artificial Intelligence

With help of Artificial Intelligence digital mimicry, India wished mass Diwali greetings

The state of Maharashtra recently delivered personalized Diwali video messages to almost 4 crore people as an example of how technology can be leveraged to reach a bigger audience. Eknath Shinde, the state's chief minister, and Devendra Fadnavis, the deputy chief minister, sent individual Diwali greetings to more than 4 crore residents of the state. These emails consisted of 42,000 messages addressed to distinct names in total. So, among the 4 crore people who received customized video messages in Marathi from Shinde and Fadnavis this Diwali were Chitanya, Manish, Kokila, and Champabai.

Such achievement was made achievable with the aid of a blockchain-based artificial intelligence-generated digital imitation created by the Delhi-based business Messagebyte. A distributed ledger known as blockchain copies distributes transactions among the network of computers involved in the blockchain. The customized video messages, which were transmitted by SMS and WhatsApp, gained enormous traction with their intended demographic. Each personalized film showed Shinde and Fadnavis asking the person's name to whom the welcome was sent.

AI research is focusing on a method for solving numerous issues that previously required fragmented or specialized approaches: building massive machine learning models on enormous volumes of data to carry out a variety of tasks for which they were not specifically created. To describe the significance of this tendency, a team of academics from Stanford came up with the provocative term "foundation models," however we might prefer the more neutral term "big pre-trained models," which broadly refers to a family of models that share several crucial properties. They receive their training by self-supervision, or without the aid of humans manually categorizing the data, and they are capable of learning new tasks on their own. With GPT-3, Open AI set the road for 2020. More recently, a flurry of other enormous models, like PaLM from Google, OPT from Meta (previously Facebook), and Chinchilla from DeepMind, followed.

Furthermore, without making any modifications to the underlying architecture, simply increasing their size and training data has shown to be an efficient way to increase their capabilities. As a result, engineering prowess rather than ground-breaking theoretical discovery has been the primary force behind much of the recent advancement in AI. Some substantial pre-trained models are just trained on text. These language models have demonstrated an amazing capacity to write meaningful paragraphs, decipher jokes, and solve mathematical problems in just a few years. All of the major tech firms at the forefront of AI research have already invested a significant amount of money in training their own sophisticated language models.

Voice assistants like Google Assistant, Siri, and Alexa, which are available in contemporary smartphones, already make use of AI digital mimicking. These systems operate by assembling words and sentences from prerecorded voice files. In addition, artificial intelligence (AI) digital mimicry can be accomplished by teaching the system to mimic a person's voice by listening to speech recordings. Machine learning, a form of artificial intelligence that enables self-learning from data and then applies that learning without the need for human involvement, encompasses the process of training a system.