 |
| By Jon Markman |
Last week, OpenAI — the AI developer funded by Microsoft (MSFT) — and Alphabet’s (GOOGL) Google debuted wildly upgraded versions of their large language model agents.
Their multimodal AI chatbots can now process text, voice, video and still images in real time.
This makes ChatGPT 4o and Gemini 1.5 seem sentient, like science fiction.
However, the reality is that AI chatbot technology is still all about scale and reach.
And that honor goes to a third-generation Llama.
I’ll explain what that means in just a moment.
First, let’s look at how LLM is learning its way into our daily lexicons.
It’s LLM-ementary
Large language models became part of the zeitgeist when OpenAI debuted the original ChatGPT in 2022.
From the get-go, ChatGPT could answer rudimentary questions and write primitive essays, poetry and glitchy code.
The output was amateurish. But the writing was on the virtual wall that computer interfaces were forever changed.
During the live demo last Monday, it was clear that ChatGPT 4o is orders of magnitudes better than previous versions.
It carried on natural conversations, complete with human-like voice inflection and a sense of whimsy.
The AI agent seemed to understand text, audio and images. And it returned responses in milliseconds, as opposed to waiting several seconds for the previous incarnation.
GPT Stunned on Monday.
Gemini Twinned on Tuesday
The next day, Google used its I/O developer conference as a coming-out party for Gemini, its most advanced AI agent.
Like ChatGPT 4o, this software feels human-like in every way. Gemini can even see the world around it and creatively draw conclusions.
Moreover, Gemini is coming to every part of the Google ecosystem — from Google Workplaces, Music and Photos to Android smartphones.
There’s no question that ChatGPT 4o and Gemini are onto something big. However …
Why GPT & Gemini Could
Get Outpaced by a LLaMa
If you use Facebook, Instagram or WhatsApp, you might soon encounter its Llama-powered Meta AI technology.
This is part of a bigger strategy by Meta Platforms (META) to dominate the AI conversation.
CEO Mark Zuckerberg pledged in April during a conference call with analysts to spend $100 billion on AI infrastructure.
Meta’s AI push across its platforms will presumably result in a digital goldmine of data. One that can better train Llama, now in its third generation, to draw conclusions.
Like Chat GPT 4o and Gemini 1.5, Llama 3 can perform multistep tasks. This makes them more adaptable to complex operations.
However, some humans believe Meta has a notable edge over the competition.
For starters, Meta made the whole LLM open source.
That means it’s widely available to developers for research and experimentation.
This Is an Android Moment.
And It’s a Really Big Deal.
Google was able to capture 80% market share in the mobile operating system because Android was robust, open source and free to use.
Developers and enterprises tend to be reluctant to share their digital data.
Llama 3, and all of the versions to come for the foreseeable future, are Android, except for AI.
Investors should not dismiss the power of this move. Analysts sure aren’t.
AI Spending Set to
Jump by MiLLions
An Andreessen Horowitz report from March noted that enterprises are tripling their investment spend on LLMs — from an average of $7 million in 2023 to $18 million through the end of 2024.
Also, some 82% of respondents said they are working with open-source models. Or they plan to move to open source in the future.
Meta is making LLMs a commodity, then giving them away for free to save costs on inference.
The improved models can then be continuously deployed across Facebook et al.
This isn’t material to Google, Apple (AAPL), Microsoft, Amazon.com (AMZN) or even Tesla (TSLA). Executives at these firms are gatekeepers of huge digital platforms, where proprietary models make sense.
However, all other LLM developers — including Open AI — will be disrupted. It is hard to compete with open source and free.
Most investors are blind to this AI masterstroke. They are too focused on cutesy ChatGPT responses … and the mighty scale of Google.
Investors should be laser-focused on buying Meta. The company has the biggest platform … and the best chance to develop a low-cost LLM that can be deployed at scale.
All the best,
Jon D. Markman
P.S. AI is even more powerful once you move away from human interactions and LLMs. It is, after all, a new way to mine and deploy data.
Here at Weiss Ratings, we know this firsthand. Click here for more. But be sure to hurry. We will be pulling this page down in just a few days.
