Andy Savage’s Post

What an amazing advancement. the combination of Langchain and Groq can provide a significant boost in enterprise applications. Langchain's natural language processing capabilities combined with Groq's high-performance computing can lead to remarkable advancements in various fields, from data analysis to machine learning.

I'm amazed by this. Llama 3 on Groq makes GPT-4 look like a grandpa: I asked both models to list all the prime numbers from 1 to 1000. Llama 3 hit over 830 tokens per second(!) Plus, it generated the entire sequence again whilst GPT-4 was still inferencing. In an arena as competitive as LLMs, speeds like this really set you apart and open up hundreds of use cases. You must give it a go. If you enjoy insights like this, follow me Alex Banks for more on AI.

AN ENJOYABLE GAME - Here is the link to a word game that uses AI powered by Machine Learning- https://lnkd.in/eErddqyG Play around with different types of clues to see what references the AI understands best....

Check out the most accurate AI Search engine! iAsk.Ai

Decent and very concise summary on how to be less of an energy consumer while deriving benefits from AI language models. Bonus on running privately and on the edge as well. "Using Lightweight AI with Small Language Models" by Morten Rand-Hendriksen! Check it out: https://lnkd.in/gFaJM6mc #artificialintelligence, #largelanguagemodels, #llm, #slm

Video here (60 sec) https://lnkd.in/ep75EETg This is SUPIR upscaling technique, based on technique published just a couple months ago, so either the cutting edge of generative AI or since we are talking about AI here...likely obsolete before anything takes to market. Gotta say is good. I kept resolution he same between old and new so its not a typical upscale its more of a denoising effect shown here. When upscaled in resolution afterwards the improvement is amplified due to the improved clarity. These are without resolution change the photo below is 768x768.

Google’s new Gemini AI model is out. The big deal is that Gemino Ultra appears to be the first model to beat GPT-4. The fascinating thing is that it does it by just a tiny bit. For right now, they are only integrating Gemini Pro into Bard. Gemini Pro is a GPT-3.5 class model, so it will not be up to the standards of GPT-4, yet.

Initial reports is that Gemini Ultra is a GPT-4 beater, Gemini Pro is a GPT-3.5 beater. Honestly just being comparable will be huge since that means we can keep all data within GCP and not send to a 3rd party.

In addition to the below, I think the announcement of Gemini Nano and its integration with Android is an interesting move. I think we are quickly approaching the days when every operating system will ship with an integrated LLM that runs locally, this will make basic LLM based features more economical and performant by offloading the cost of compute to the consumer (electricity and device cost) and make it feasible for developers to build interesting applications without ballooning costs. I'm guessing this "Generative AI at the edge" that Google is kicking off here is going to be Apple's long term play since they have been leaning so heavily on privacy and security lately. State-of-the-art models will still require hardware that will be too advanced to run locally for awhile but many of the these smaller models are already good enough that they can provide a lot of value even if they can't compete with GPT-4.

I will be writing more on this new ai!