We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
cover of episode From Technical Perspective

From Technical Perspective

2024/1/13
logo of podcast AhbarjietMalta

AhbarjietMalta

Shownotes Transcript

From Technical Perspective

Today's Amazon Deals - https://amzn.to/3FeoGyg

—-

From Technical Perspective

In technical overview, it’s already discussed that there are large complex models available than GPT which are more capable for performing in various other tasks than what ChatGPT does. Some of such high-level models are:

BLOOM is an autoregressive Large Language Model (LLM) that has been trained on vast amounts of text data using industrial-scale computational resources. This impressive ChatGPT alternative is designed to continue text from a prompt and is capable of outputting coherent text in 46 different languages and 13 programming languages. The text it generates is so advanced that it is often difficult to distinguish it from text written by humans. One of the key strengths of BLOOM is its ability to perform text tasks that it hasn’t been explicitly trained for. This is achieved by casting the task as a text generation task, allowing BLOOM to use its vast knowledge base to generate text that is relevant and coherent. This feature makes BLOOM an incredibly versatile tool that can be adapted to a wide range of applications and industries.

PALM or the Pathways Language Model, one of the key innovations of the Pathways architecture is multimodal training, which trains models on various types of data, including video, picture, and text. This sets it apart from text-based models like GPT. Another significant contribution is the use of sparse activation, which involves using only a subset of neurons for a given task, leading to better performance and lower running costs.

OPT models have remarkable capabilities for zero- and few-shot learning which have been shown by large language models, which are often trained for hundreds of thousands of compute days. These models are difficult to replicate without significant capital, given their computational cost. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters. It is shown that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop.

There are many more complex model’s with heavy computational loads, and finally Google’s newest addition BARD, joined the list and being the one of the closest competitors of ChatGPT. Fresh, high-quality responses are provided by Bard, which combines the power, intelligence, and creativity of large language models with the breadth of the world’s knowledge, illustrating information from the web. Now google’s models and cutting-edge AI technologies — like LaMDA, PaLM, Imagen and MusicLM — are building on this, creating an entirely new advanced AI ecosystem leveraging all the different kind of modalities taking into the account.