Clarifai’s Post

We're hosting a demo this Thursday at 12pm ET where we'll show our new ModelOp Center version 3.2 and how to safeguard LLMs and Generative AI without stifling innovation. Click below to register! ModelOp's VP of Product, David Trier, will show how version 3.2 builds on ModelOp’s existing governance support for all models — including regressions, Excel, and vendor models — by adding new capabilities to manage LLM ensembles, track value, chart and visualize risks, provide universal monitoring, and enforce governance controls through model life cycle automation. #artificialintelligence #aigovernance #generativeai

Amazing read from Chip Huyen on common #patterns of #GenerativeAI #Platforms. Also shows the maturity of modern #LargeLanguageModels from #RAG to #Agentic RAG and more advanced #Agent #Workflows. I would also add patterns around model #governance at user and application level, #knowledgegraph and #costmanagement - for completion. Full article: https://lnkd.in/dVfD__3m

In his latest blog, VP of GenAI Naveen Rao shares why custom LLMs should be the de facto strategy for organizations looking to tap into the power of #GenAI (and why even small, domain-specific models can have a big impact). Read his insights on how to get started with custom LLMs - whether your approach is with RAG + open source or you're building your own. https://dbricks.co/3uCfOkE

📢 ICYMI! 📌 See Prompt Lab in action: https://hubs.li/Q02xD9Q10 This spring, Agiloft is excited to launch 🚀a new #GenAI tool that will help make #contract connections faster and more intelligent than ever. Say hello to GenAI Prompt Lab. 👋 With our Prompt Lab and enhanced ConvoAI capabilities, you can: 👉 Create time to do the work that moves the needle by automating the boring stuff. 👉 Scale your team by making the genius prompts of your very best colleagues available to everyone. 👉 Simplify tasks by letting AI do some of the workload lifting, giving you an extra set of hands where it counts most. Register for our upcoming 30-minute demo to see it in action!

Tuning into Config 2024 today and tomorrow remote from home! Currently watching The Human Algorithm: A New Opportunity to Make Great Software Refining the Relationship between Humans and AI, led by Henry Modisett. #Config #Figma #config2024 #conference #designconference 🤖🦾

📢 ICYMI! 📌 See Prompt Lab in action: https://hubs.li/Q02xD9Q10 This spring, Agiloft is excited to launch 🚀a new #GenAI tool that will help make #contract connections faster and more intelligent than ever. Say hello to GenAI Prompt Lab. 👋 With our Prompt Lab and enhanced ConvoAI capabilities, you can: 👉 Create time to do the work that moves the needle by automating the boring stuff. 👉 Scale your team by making the genius prompts of your very best colleagues available to everyone. 👉 Simplify tasks by letting #AI do some of the workload lifting, giving you an extra set of hands where it counts most. Register for our upcoming 30-minute demo to see it in action!

Meet my Digital Twin 😍 Ironically I genuinely actually have a biological twin sister already in Gerry Lindup Guys,I had so much fun creating my #AITwin You can work with your digital twin aka Personalized AI assistant in real time collab-ing ( ie reports and debugging code, documents etc) With MindOS, you can customize your own AI persona, complete with a distinct voice, biography, and knowledge base, making your digital twin truly one-of-a-kind. This is just incredible what MindOS have created 😀 You can find them here ➡ mindos.com #DigitalTwin #aiforproductivity #aiadoption

This just happened to me. Greg Coquillo, I was researching a company and its members and used Google Bard, which gave a lot of 'good' insights as its output. Then I asked the same questions from Chat GPT and got the reply that made me wonder if something was off... "This pattern of similar descriptions across different websites raises questions about the authenticity of these profiles." Now, I asked Bard to tell me the sources of the 'good' information and then got this reply; "My apologies! I misspoke previously when mentioning Jessica XXXX. I haven't found any information about her, and your question about her legitimacy raised a red flag regarding potential misinformation. Therefore, I'm unable to assist with that particular part of your request." This led me to validate the information with other sources and research deeper to finally figure out that this company and its (supposed) members are a scam. The moral of the story is that hallucinations are real, and we all need to validate the results by cross-referencing and do more profound research. At the end of the day, as you said, LLMs are not one-size-fits-all, so be prepared to test beyond prompting and validate the outputs. Also, know the best use cases for each LLM model, resource, and its constraints. #AI #hallucinations #llms

Hallucination was probably the main reason preventing LLM user adoption across industries since GenAI became a thing. This collaborative survey highlights more than thirty mitigation techniques to reduce negative effects. These techniques are categorized under two pillars (Prompt Engineering, Developing Models): 🔹Prompt Engineering 🔸RAG 🔸Self Refinement Through Feedback 🔸Prompt Tuning 🔹Developing Models 🔸Utilization of Knowledge Graphs 🔸Supervised Finetuning 🔸New Decoding Strategy (Context-Aware) LLMs will never be one-size-fits-all, so best to plan, test, adjust and adapt to your use cases, resources, and constraints. #genai #technology #artificialintelligence #linkedangle follow for more content Set the 🔔 notification on my page, don’t miss a post!

📢 ICYMI! 📌 See Prompt Lab in action: https://hubs.li/Q02xD9Q10 This spring, Agiloft is excited to launch 🚀a new #GenAI tool that will help make #contract connections faster and more intelligent than ever. Say hello to GenAI Prompt Lab. 👋 With our Prompt Lab and enhanced ConvoAI capabilities, you can: 👉 Create time to do the work that moves the needle by automating the boring stuff. 👉 Scale your team by making the genius prompts of your very best colleagues available to everyone. 👉 Simplify tasks by letting #AI do some of the workload lifting, giving you an extra set of hands where it counts most. Register for our upcoming 30-minute demo to see it in action!