We’re passing the torch to our AI friends: the supersized ones are slowly dying out and the tiny ones are taking over the top spot. In turn, Meta intends to launch a lite version of its Llama language model so that users can get a feel for the power before the good model comes out by the summer. This focus of the AI community on building power-efficient models is a common trend due to their fast, affordable, and well-use process.
Just like you have to choose between a five-course meal and just one tapa. They may not be able to deal with complex things that require blocking out long sequences, but they have their right place – narrow. Think of them as summarizing all the documents; Simple code lines or even more complex tasks that involve coding. Here comes the advantage, the thing that sets them apart from their bigger cousins: they are not power consuming which will hinder your computing power, thus they do not deplete devices like desktops or cell phones and laptops.
This direction on accessibility confirms its society compliance which is happening rapidly in AI. Meta is not unique but the competition is also fierce as other business entities like Google and Mistral have also come up with their compact models. “AI tapas” are closer to such tastes that users will not need much power and the burden of a huge language model for the concentration of their needs.
The main course is still on pending delivery status. Llama 3, a meta with a higher grade than Llama 2.0, should be expected to launch in just a few weeks. It can do everything the Llama 2 can’t. It may also be able to answer complex questions that Llama 2 has no use for. This means highlighting the plethora of AI ranges, Meta will be providing its full-course meal along with appetizers of such favorite easy choices. Big possibilities and much more!
