All articles tagged with #transformer
Amazon's devices chief Panos Panay says the company is not pursuing a new smartphone, citing the Fire Phone's failure and emphasizing a strategy to make devices profitable by boosting use of Amazon services and AI rather than launching another phone.
Amazon’s head of devices, Panos Panay, says the company isn’t necessarily planning a new smartphone, while not fully denying a project tied to a rumored Alexa-enabled AI phone codenamed Transformer; he notes the smartphone form factor isn’t vanishing but is evolving, and any future device may differ from a traditional phone, likely learning from the Fire Phone and Panay’s Microsoft experience.
Amazon is reportedly developing a new AI-centric smartphone, codenamed Transformer, that would integrate Amazon services and AI features (potentially using Alexa as an OS) to boost usage of its AI products and shopping ecosystem; Reuters cites unnamed sources and notes cost, investment, and release details are unconfirmed, and the project could still be canceled, with no official comment from Amazon.
GPT stands for Generative Pre-trained Transformer: a generative AI that creates new text, trained on vast datasets before release and refined with human feedback, using transformer neural networks to turn queries into tokens and vectors and prioritize important information for responses. The article notes these models are large language models prone to inheriting biases from training data, and adds that ChatGPT’s name was a late, simplifying choice before its 2022 public launch (originally being called ‘Chat with GPT-3.5’).
Google unveils Gemini 1.5, the latest iteration of its AI model, built on a Transformer and Mixture-of-Experts architecture. Gemini 1.5 Pro, designed for a wide range of tasks and devices, can handle larger prompts and outperforms its predecessor on testing benchmarks. It features in-context learning and is currently available for trials through AI Studio and Vertex AI, with a waitlist for interested developers.
Large language models (LLMs) like ChatGPT are built on neural networks trained using billions of words of ordinary language. While the inner workings of LLMs are not fully understood, experts have made progress in explaining their functioning without technical jargon. LLMs represent words using word vectors, which are long lists of numbers. These models rely on transformers, the basic building blocks, to process and reason about language. Training LLMs requires massive amounts of data to achieve good performance.