Todd Nist’s Post

MARKTECHPOST writes about an open-source solution trying to replicate what Perplexity does https://lnkd.in/eP7H6gmf. #perplexica #perplexity #opensource #search #generativeai #artificialintelligence #marktechpost

We might not be seeing a significant change at Google after all!! I can answer my question, no they are not halting the new SGE because everyone in SEO is mad. There is a way better reason! This week has been a rollercoaster of news updates about AI coming to the forefront of Google in the user search engine results. Google is keeping SGE in it's lab indefinitely! That's a big anxiety point for me that has been lifted off of my shoulders. But why? After so many announcements about how SGE is going to shake up Google, now they aren't launching. There's actually a really good reason. Let's dive in: 1. Google already has a long history in using AI in the background of Google's search results engine. Starting in 2015, all of their AI is running in the background to solve different problems within search as features, sometimes as a part of the larger Core Algorithm. 2. It was likely that Google would have never had the intention on bringing in AI into the foreground of the SERPs. Microsoft's integration of ChatGPT into Bing forced Google to take steps to add AI in a more foregrounded way with their Search Generative Experience (SGE). 3. One of Google's guiding principles for the use of AI is to only use it once the technology is proven to be successful and is implemented in a way that can be trusted to be responsible and those are two things that generative AI is not capable of today. There are at least three big problems that must be solved before AI can successfully be integrated in the foreground of search: 👉LLMs cannot be used as an information retrieval system because it needs to be completely retrained in order to add new data. . 👉Transformer architecture is inefficient and costly. 👉Generative AI tends to create wrong facts, a phenomenon known as hallucinating. In simple terms, what happens is that in a regular search engine, adding new webpages is a process where the search engine computes the semantic meaning of the words and phrases within the text (a process called “embedding”), which makes them searchable and ready to be integrated into the index. In contrast to current search technology, LLMs cannot add new webpages to an index because the act of adding new data requires a complete retraining of the entire LLM. Google is researching how to solve this problem in order create a transformer-based LLM search engine, but the problem is not solved, not even close. On a scale of 1-10, how happy are you that you won't have to scramble around changing your SEO strategy in 2024?? #SGE #googleseo

Super interesting article about retrieval-augmented AI. The core of the problem is to build an effective search engine, and there's much more to that than just computing embeddings and nearest neighbors. First of all, you need to be able to assess the effectiveness of the search engine. I.e., if you're looking for a specific piece of content, how easily can you (or an AI) retrieve it using the search engine? How well does the search engine distinguish it from irrelevant or misleadingly similar content? Futhermore, search engines based on text similarity alone do not perform well for most queries. You need filters, advanced search tools, keyword search, and sometimes even a recommendation system to support effective search for a given user. An effective search engine is a complex system that combines multiple signals and processes to provide relevant results and exclude irrelevant ones. Hamza Farooq I'd be interested in hearing your thoughts on this! https://lnkd.in/eBi6BfmJ

Search has become such an integral part of our lives we barely even think about it. But when you do, you realize it's not that great an experience. Honestly, you don't want a Search Engine, you want an Answer Engine. Enter Brave Software. With their Answer with AI tool, you type in what you are looking for and in addition to getting links, you get a direct answer (with citations) to your question. It's brilliant and will change how you engage with the internet. Check it out: https://lnkd.in/eBm756Kp Petamber Pahuja, Shezmeen Hudani, Ryan McFadden Sean Bohan, Vijay Chattha, Vinodh Bhat #ai #search

The future of web search will be driven by advancements in natural language processing and machine learning. This will enable search engines to understand complex queries and provide more accurate and personalized search results. But even so, there are potential problems and ethical considerations associated with these advances, such as privacy issues and algorithmic bias 💻👨💻 #tech  #news  #ML  #web https://lnkd.in/dRBEzcxd

Been spending a bit of time experimenting with "AI-powered" searching. By this I mean two things: (1) AI search: finding results based on semantic matches, and (2) Q&A: AI-generated answers to questions based on semantic matches to the question Here are some things I have found. 1. Don't let anybody tell you "keyword" searching is dead. These things need to be considered side by side. Keyword searching is the best way of tracking down a document where you know a bit about its contents but not much about its location 2. AI search and Q&A have different use cases. AI search helps you find needles in the haystack, like a less precise but further reaching keyword search. Q&A helps you synthesise materials so you don't have to read them all cover to cover 3. The natural next step in an AI search (which relies on semantic relationships rather than keywords) is "why". What made you bring back these results? Which bits of the results were semantically relevant? 4. When using AI in a "Q&A" capacity (i.e. respond to this question instead of find things related to these words), it does so based on a synthesis of semantically similar documents ("RAG" or "retrieval augmented generation"). Knowing which documents were used to produce the summary is key. You'll also want to know which bits of the documents were used 5. Wrong or incorrect answers can still be produced by RAG if the wrong source material is synthesised or if the answers seem to come out of nowhere 6. Everything is improved by having better quality content. Chucking an entire DMS at things is probably not helpful 7. Everything is improved by having better organised content. Chucking an entire DMS at things is probably not helpful My overall conclusion is that as you start to explore these things, you open up can of worms after can of worms. There are so many complexities to work through here. The future is promising, but not simple. Don't let anybody tell you otherwise.

THE PULSE ON AI - CSPARNELL Reinventing Search: Perplexity AI's Online LLMs Challenge the Status Quo In a groundbreaking move, Perplexity, a San Francisco-based AI startup, has unveiled two innovative online Large Language Models (LLMs) named pplx-7b-online and pplx-70b-online. These models are not just new additions to the AI landscape; they represent a significant shift in how we approach online search and information retrieval. What Makes These LLMs Stand Out? The pplx-7b-online and pplx-70b-online models are the first of their kind to be publicly accessible online via an API. This access marks a pioneering step in the world of AI and search technology. Unlike traditional offline LLMs, such as Claude 2, Perplexity's models leverage live internet data. This unique feature enables them to provide real-time, up-to-date information, a capability that traditional models lack. Real-Time Information at Your Fingertips Imagine needing the latest sports scores, stock prices, or current news updates. These online LLMs can tap into real-time internet data to deliver just that. This ability to access and provide the most current information makes them exceptionally suited for answering time-sensitive queries, setting a new standard in the realm of digital information access. A Step Towards Knowledge Without Limits Aravind Srinivas, CEO of Perplexity, aptly describes these new LLMs as "the first-ever live LLM APIs that are grounded with web search data and have no knowledge cutoff." This statement underscores the revolutionary aspect of these LLMs – their capability to continuously update and expand their knowledge base, breaking free from the limitations of knowledge cutoffs typical in many existing AI models. The Potential Impact The introduction of pplx-7b-online and pplx-70b-online is not just a technical achievement; it's a potential game-changer in the way we search for and access information online. By offering up-to-date, accurate, and factual information, these models could challenge the dominance of traditional search engines like Google, potentially reshaping the landscape of online search and information. Conclusion Perplexity's launch of these online LLMs is a bold step towards a future where real-time, accurate information is readily available at our fingertips. It's an exciting development that promises to enhance our online experience and possibly redefine the way we interact with digital information. #AIRevolution #PerplexityAI #OnlineLLMs #RealTimeInformation #InnovativeSearch

2024 is going to be very exciting 🌊! What am I expecting, apart from AI finally taking over the world, are advances in Search tecnology. In my latest article, I share my findings on new technologies that may affect Search, and will likely manifest in 2024. Follow me for the best news in tech, or even better, connect! Thank you Choicy.co for publishing! https://lnkd.in/gqPryMtj #datascience #searchenginetechnology #aisearch #aisolutions #ai

Infini-attention is Google's new technology that enables transformer-based models to process huge context lengths. What makes this important to SEO is that because it's plug and play Google can add this to to any model and scale up their AI. Search Engine Journal https://lnkd.in/ejVchFyK

Understanding the difference between traditional search & vector search is the basis of 𝐮𝐧𝐝𝐞𝐫𝐬𝐭𝐚𝐧𝐝𝐢𝐧𝐠 𝐭𝐡𝐞 𝐦𝐚𝐠𝐢𝐜 𝐨𝐟 𝐚𝐝𝐯𝐚𝐧𝐜𝐞𝐝 𝐀𝐈 𝐚𝐩𝐩𝐬 with vector databases - great job explaining that Dr Anastasia Ragulskaya 💪 #vectordatabase #vectorsearch #AI #databases