Iván Martínez Toro

Is the end of traditional search engines near? There is a fear that conversational AI-based tools like ChatGPT, Gemini, and Llama will eliminate the need for search engines, affecting the communication of organizations that rely on search engine optimization. I’d like to share a few thoughts on this. The business of search engines is to deliver relevant and high-quality content as quickly as possible. To achieve this, they promote the production of original and specialized content through their algorithms. Additionally, they consider factors such as site usability, geographical proximity, and content publication date. If producing content did not generate a benefit in attracting visitors, organizations would not invest time in creating it. Similarly, if the information provided by conversational systems were not relevant or original to users, these systems would lose their utility and stop being used. Therefore, the generation of relevant content will continue to be rewarded, as this is the essence of their business model. In any case, we must continue to focus on providing value, regardless of the platform users use to access information. Having said that, I believe that in the near future, innovations in generative AI will integrate complementarily with traditional search engines, as is currently the case with rich snippets and frequently asked questions that already enhance the user experience. https://lnkd.in/dYQzVrgq

9

2 comentarios

OLMoE is a fully open, state-of-the-art language model that leverages sparse Mixture-of-Experts (MoE). OLMoE-1B-7B has 7 billion parameters but uses only 1 billion per input token. It was pre-trained on 5 trillion tokens and further adapted to create OLMoE-1B-7B-INSTRUCT. These models outperform all available models with similar active parameters, even surpassing larger ones like Llama2-13B-Chat and DeepSeekMoE-16B. Various experiments on MoE training were conducted, and the routing in the model was analyzed, showing high specialization. All aspects of this work, including model weights, training data, code, and logs, are open-sourced. Weights https://lnkd.in/evb_dTRZ Data https://lnkd.in/eZAukC8U Code https://lnkd.in/entrE2B6 Logs https://lnkd.in/e_qB5C8g OLMoE-1B-7B-0924--Vmlldzo4OTcyMjU3 #OLMoE #Llama2 #DeepSeekMoE #MoE #LLM #HF #dense #FFN #feedforward #NN

5

1 comentario

LLaMA 3 is the new talk of the town, some of you might be curious what it brings to the table. This new series of models in the LLaMA series by Meta comes in different sizes such as 8B and 70B, with an even bigger model with 400B parameters coming soon! With a larger model than LLaMA 2 and better results than Mistral and Gemma, the model has also been observed to beat GPT 3.5 in terms of performance. In the following blog published on Tune AI, we explore the extended hype behind this new series of models and what it holds in the future for Researchers and Organizations! Head over to the link in bio to custom train and check out the model for yourself on Tune Chat and Tune Studio! https://lnkd.in/ekaw7QB4 #llm #llama #meta #chatgpt #llamadeploy #llama3 #llama3chat #llama3meta

19

1 comentario

😲 Wow, hold the phone! 📱 I just stumbled upon something that's got me grinning from ear to ear. Guess who's sitting pretty at the **top spot** on GitHub's contribution rankings? Me! Yes, you read that right. I'm as surprised as you are – it's like finding out I've won the lottery without buying a ticket! 🎟️🥇 Link or is fake: https://lnkd.in/eUJ6G85J So here's a massive shoutout to all the night owls 🦉, early birds 🐦, and everyone in between who's ever forked, cloned, or starred a repo. We're in this together, and I couldn't be more stoked about what we've achieved. Let's keep this code party going – who knows what surprises are around the next commit? 🎉 #GitHub #OpenSource #SurpriseNumberOne #CommunitySpirit #CodeParty

4

4 comentarios