Llama 3: How AI is scaling fast? Mark Zuckerberg tells us about the platform and the AI of the world, and how it affects GPT-4
The goal is for Meta AI to be “the most intelligent AI assistant that people can freely use across the world,” CEO Mark Zuckerberg tells me. We feel like we are there with Llama 3.
He says the goal is not to compete with the open source models. “It’s to compete with everything out there and to be the leading AI in the world.”
A key focus for Llama 3 was meaningfully decreasing its false refusals, or the number of times a model says it can’t answer a prompt that is actually harmless. An example Zuckerberg offers is asking it to make a “killer margarita.” Another is one I gave him during an interview last year, when the earliest version of Meta AI wouldn’t tell me how to break up with someone.
Llama 3 is a good example of how quickly these AI models are scaling. The biggest version of Llama 2, released last year, has 70 billion parameters, while the upcoming large version has over 400 billion. Llama 2 trained on 2 trillion tokens (essentially the words, or units of basic meaning, that compose a model), while the big version of Llama 3 has over 15 trillion tokens. The number of parameters in GPT-4 has not been publicly confirmed by OpenAI.
The Metaai Assistant: Online Chatbot for More Popular Social Networks (with a Survey) and a Comparison to Stories and Reels
The Metaai assistant is being integrated into the search box of several popular social networks. It will be seen directly in the main Facebook feed. If you want to chat with it in the messaging inboxes of Meta’s apps, you can. And for the first time, it’s now accessible via a standalone website at Meta.ai.
There’s a comparison to be made here to Stories and Reels, two era-defining social media formats that were both pioneered by upstarts — Snapchat and TikTok, respectively — and then tacked onto Meta’s apps in a way that made them even more ubiquitous.
While it has only been available in the US to date, Meta AI is now being rolled out in English to Australia, Canada, Ghana, Jamaica, Malawi, New Zealand, Nigeria, Pakistan, Singapore, South Africa, Uganda, Zambia, and Zimbabwe, with more countries and languages coming. This wider release gets Meta AI closer to reaching the 3 billion users that the company has and is far cry from what MarkZuck said about a truly globalai assistant.
I don’t know of a chatbot that combines real- time search results from both Bing and Google in order to answer a prompt. High-resolution images are now generated on the fly when you type, and an upgrade to the image generation has made it possible to create animations. Meta wants a panel of prompt suggestions when you open a chat window to show you what a general-purpose chatbot can do.
The evaluation set contains over 1,700 prompts that cover a number of use cases, including asking for advice, classification, closed question answering, coding, creative writing, and persona, according to Meta.
It should also be noted that benchmark testing AI models, though helpful in understanding just how powerful they are, is imperfect. The datasets used to benchmark models have been found to be part of a model’s training, meaning the model already knows the answers to the questions evaluators will ask it.