Who Are Microsoft's LLM Contemporaries?
While details are scarce, leaks report the new 500 billion parameter model dubbed ‘MAI-1’ could be big enough to rival the largest built models currently available.
Although Microsoft's recent public pronouncements regarding LLMs might seem like a fresh entry, they've been involved in the field for some time. Their GPT-4 integration into Microsoft 365 Copilot, for instance, highlights their application of LLM technology.
The core function of LLMs, regardless of the company, lies in processing and understanding massive amounts of text, and increasingly visual, data.
By analysing this, they can perform a variety of tasks, including generating different creative poems, writing code, emails, and translating languages.
The number of parameters within an LLM, often in the billions, is a metric used to gauge its potential complexity and capability.
While a higher parameter count doesn't guarantee superiority, it suggests the model's ability to handle the relationships within the data it's been trained on.
So with that said, which companies' models would it consider its contemporaries?
Here's a breakdown of some of the leading companies developing LLMs, along with their key features.
1. Google AI
Google AI has one of the longest and track records in LLMs. Google AI has made significant contributions with models like Meena, LaMDA, and PaLM.
LaMDA focuses on generating realistic and engaging dialogue, while PaLM tackles complex reasoning tasks. Their PaLM model, with its 540 billion parameters, is currently considered one of the largest and most powerful.
2. OpenAI
Backed by heavyweights like Microsoft and Elon Musk, OpenAI is one of the most well-known LLM, renowned for its text generation, creative writing abilities; translation and even code generation.
Unlike some companies that keep their LLM technology entirely private, OpenAI believes in a degree of open access. They offer limited access to GPT-3 through their API, allowing researchers and developers to experiment with the model and contribute to its advancement.
3. Meta AI
With its vast user base and data resources, Meta brings a unique perspective to the LLM landscape.
Their Blender model focuses on factual language understanding and information retrieval, useful for the vast amount of data within their social media platforms.
Meta's research involves the LLaMA development, a model which aims to assist scientists and engineers in exploring AI applications.
4. Anthropic
Anthropic, can be compared to OpenAI’s ChatGPT but for business. A rising star in AI, is making a splash with their LLM, Claude.
With an estimated parameter size in the tens of billions, Claude goes beyond creative writing. It can act as a search engine that not only finds information but summarises it, or a programming assistant that suggests relevant code completions, or even generates basic snippets based on your instructions.
While Claude might not be the biggest LLM out there, its capabilities in areas like information retrieval and code generation make it a valuable asset.
The LLM landscape is one of constant evolution, especially now as the AI race is well and truly underway in the tech industry.
But just exactly how the various LLMs, each with varying different focuses, stack up or give the edge over one another, remains to be seen.
******
Make sure you check out the latest edition of AI Magazine and also sign up to our global conference series - Tech & AI LIVE 2024
******
AI Magazine is a BizClik brand