Large Language Models are eating the world.

In the last two years, NLP has become the flagbearer of AI with state-of-the-art AI models becoming more accessible and affordable for all.

(A thread) 👇🧵

Commercial release of LLMs by companies like OpenAI, Cohere, and AI21Labs have made it possible for anyone to quickly go from ideas to execution irrespective of their backgrounds.

These developments have changed the very basic perception of AI which conventionally was highly inclined towards the machine learning process rather than building products and services.

It has given rise to a new kind of fully managed "ML-as-a-service" delivered with the abstraction of API and the comfort of a user interface.

Now building sophisticated AI solutions is all about getting your data and a few clicks to deploy your solution.

All of this has helped to make AI accessible for all, but what about the other major pain point which is affordability?

This idea has spurred a new wave of startups focused on making LLMs cheaper and more effective via managed AI infrastructure with open source as the catalyst.

In the last few months, there has been a lot of movement in this space with startups like Goose AI, Spellbook from Scale AI and many others trying their hands on it with value propositions lying around reduction in cost and increase in efficiency.

These startups claim a 20-30% reduction in the cost of AI infrastructure.

It would be safe to say that this is just the tip of the iceberg and ML/NLP-as-a-service will become mainstream in a year or two.

This raises the biggest question of all with all these startups coming up - "what will be the differentiating factor?" 🤔

IMHO, the underlying hardware.

Efficiency and Affordability of hardware will be the key differentiators going forward and to win in this market you need hardware specifically designed for AI.

Check out @tenstorrent to get a glimpse of the future of AI hardware and computing 👉 tenstorrent.com/

@tenstorrent If you are an individual or a startup looking to train or inference large language models, reach out to me for understanding how the next generation of chips designed for AI can make your life a breeze.

DMs are open 📩

@tenstorrent If you enjoyed reading this, two requests:

1. Follow me @Saboo_Shubham_ to read more such content.
2. Share the first tweet in this thread so others can also read it 🙏