This year, Artificial Intelligence reached new heights. Lot of new innovations were seen in 2022 in the field of AI. Among them, a few made headlines and became prominent this year. Cicero, DALL-E 2, ChatGPT, Stable Diffusion and AlphaFold have topped the list this year.
ChatGPT
Open AI announced ChatGPT on the last day of November 2022. It is a chatbot that is based on the GPT-3 language model. It was made available for free on the OpenAI website for it to gather feedback and data from the public on the model to offer more accurate and less harmful results. The CEO of OpenAI, Sam Altman tweeted after five days of the launch that ChatGPT has reached over 1 million users. People used it do4 programming tasks that include simulating a Linux console session, writing poetry, generating recipes and a lot more. GPT-3 is being provided since 2020 but ChatGPT brings the best of it.
Cicero
Meta announced Cicero in late November 2022. It is an AI agent that can beat humans at Diplomacy which is a strategy board game that is played online on webDiplomacy.net. This game requires cooperation, persuasion and negotiation with other players. A bot was developed by Meta that would fool humans into thinking that the other player is also a human. Cicero was trained to gather negotiation skills and develop strategic components, predict the move of other players and react accordingly. It was trained on transcripts of 40,000 games played by humans that are available on the internet and on webDiplomacy.net.
Stable Diffusion
Stable Diffusion 1.4 was made available by StabiltyAI and CompVis in August 2022. It is an image synthesis model which is similar to OpenAI’s DALL-E2. Stable Diffusion is an open-source project that is complete with source code and checkpoints. It allows unrestricted generation of any synthesized content and unlike DALL-E 2 it can be used privately and locally with a good enough GPU on their PCs. But a lot of critics claimed that with the introduction of AI into art, artists will get replaced. A technique called Dreambooth could fine tune the Stable Diffusion model easily.
AlphaFold
AlphaFold AI model was announced by DeepMind in July 2022. It predicted the shape of all the known proteins in organisms on Earth. It was announced in the summer of 2021 originally and predicted the shape of all human proteins then. But a year later it was able to expand its database to over 200 million protein structures. It was made available to the public by DeepMind via the European Bioinformatics Institute at the European Molecular Biology Laboratory (EMBL-EBI). This allows researchers to make use of the data to relate it to medicine and biological science globally.
DALL-E 2
DALL-E 2 was announced by OpenAI in April 2022. It is a deep-learning image synthesis model that has the ability to generate images from text prompts. It has been trained on millions of images from the internet and DALL-E 2 used latent diffusion to make novel combinations of imagery. Soon Twitter was filled with images of teddy bears wandering in ancient Egypt, astronauts on horsebacks and other photorealistic works. Only 200 beta users were allowed to use DALL-E 2 due to concerns regarding misuse. Content filters were used to block violent and sexual prompts. Gradually, over a million users were allowed to use the trail and then in September 2022 it was made available to everyone.
ALSO READ | From Quantum Computing to 5G and IoT, 5 biggest tech trends of 2022 | Tech Bytes year-ender
ALSO READ | Year-Ender | Top world events no one expected in 2022 but made headlines and changed the course of history
