The Evolution of Open-source LLMs and What it Means to Us

According to a recent Verge article, Meta has released the biggest and best open-source AI model yet. Meta claims that the new release of Llama 3.1outperforms GPT-4o and Anthropic's Claude 3.5 Sonnet on several benchmarks. Furthermore, Meta's CEO, Mark Zuckerberg, predicts that Meta AI will be the most widely used assistant by the end of the year (2024).

I mentioned in my previous postingthat these large AI vendors might need more training data. The new Llama 3.1 model has 405 billion parameters and was trained with over 16.000 of Nvidia's ultra-expensive H100 GPUs. The Verge article estimates that the model's training cost hundreds of millions of dollars.

The question that we should pose is what kind of future these open-sourced LLMs will have. Zuckerberg claims that open-source LLMs will overtake proprietary LLM models with time. He compares this evolution to what Meta did with HP in its earlier Open Compute Project, where HP helped to improve and standardize Meta's data center designs. He furthermore believes that the Llama 3.1 release will "be an inflection point in the industry where most developers begin primarily to use open source."

Meta is working with two dozen companies, such as Microsoft, Amazon, Google, Nvidia, and Databricks, to help developers deploy their own versions. Also, according to the article, Llama 3.1 costs roughly half of OpenAI's GPT-4o to run in production. Roughly a year ago, Mikko Peltola from A-CX and I run a webinar in our "Approachable AI" series with some cost models, we have since them come drastically down in pricing even in the OpenAI case. I assume that this trend will continue as it typically does with any new technology introductions.

Meta agrees that there is a growing consensus that the industry is running out of quality training data for models. However, entrepreneurs and developers have a business opportunity with these large LLMs to use LLM models such as Llama 3 to be "teachers for smaller models that are then deployed more cost-effectively."

According to the Verge article, "Meta's own implementation of Llama is its AI assistant, which is positioned as a general-purpose chatbot like ChatGPT and can be found in just about every part of Instagram, Facebook, and WhatsApp. Starting this week, Llama 3.1 will be first accessible through WhatsApp and the Meta AI website in the US, followed by Instagram and Facebook in the coming weeks."

The AI field, especially solutions within GenAI, is moving at an accelerated pace, and organizations, both independent software vendors (ISVs) and end-user organizations, are pondering how to get started and what type of use case to kick off the AI journey. A-CX, a design and engineering firm, is helping organizations to build GenAI solutions with the collaboration of TELLUS International, which has delivered tens of envisioning and business design sessions for both ISVs and end-user organizations. A-CX and TELLUS are partnered to allow an end-to-end scenario for clients, and we would love to hear from you about where you are with your AI journey and see if we can help you get started.

Yours,

Dr. Petri I. Salonen

PS. If you would like to get my business model in the AI Era newsletters to your inbox on a weekly or bi-weekly basis, you can subscribe to them here on LinkedIn https://www.linkedin.com/newsletters/business-models-in-the-ai-era-7165724425013673985/