Link: Things we learned about LLMs in 2024

The breakthrough of the GPT-4 barrier showcased significant advancements in Large Language Models (LLMs) during 2024. With models like Google’s Gemini 1.5 Pro introducing capacities like video input and increased context lengths, multiple organizations have now surpassed OpenAI's initial GPT-4.

LLM capabilities have expanded onto personal devices; models that were once datacenter-bound now operate on regular laptops. This leap in accessibility is bolstered by dramatic reductions in operating costs, thanks to improved efficiency competing in the market.

A marked rise in multi-modal models has made LLMs capable of processing and generating based on audio and video inputs. This progress toward audio and live camera modes indicates a merging of LLMs with elements of everyday technology.

Prompt-driven application generation has transformed into a common and accessible tool, demonstrating the practical utility of LLMs in software development. The ease of creating interactive applications speaks to the significant strides in technology utility.

Despite broader access to advanced models for a period, the landscape has shifted towards more restricted, subscription-based access to premium models. This gating of technology signifies a move towards monetization and controlled access.

Meanwhile, many challenges persist, such as the environmental impact of datacenters and the ethical dilemmas surrounding AI. These issues emphasize the ongoing need for critical engagement with LLM advancements and their societal implications.

 #

--

Yoooo, this is a quick note on a link that made me go, WTF? Find all past links here.