Aboitiz Data Innovation (ADI)’s Post

Personification is an area of “Ethical AI” that doesn’t get enough attention. Infusing the design of your AI product with “personality” (scripting responses to sound human and make the experience “friendly”). For example, using the pronouns “I, me, my, or we” in textual responses is ethically risky because we’re (unfairly) leveraging millenia of evolutionary instincts and cognitive bias to subtly influence human users to take certain actions, make information seem credible, or reveal more personal information than they might otherwise share. On Alexa there was a design principle that “Alexa is a computer, not a person, IT only speaks when spoken to”. Its an odd principle to adopt, but I think it embraces the idea that healthy boundaries in the product are necessary to maintain customer trust.

"Our Head of AI, Kalaivani K G, shares her experiences with Generative AI, tackling challenges like AI 'hallucinations' and practical ways to address them. Dive into her insights on separating GenAI fact from fiction." #exafluence #realworld #usecases

AI isn't about replacing humans, it's about amplifying our capabilities. Together, we're rewriting the rules of what's possible by merging human intuition with AI precision. #AIandHumans #LimitlessInnovation #FutureForward"

AI distils all published knowledge into a single answer. But often there isn't a single right answer. From this perspective you could say that AI is the ultimate echo chamber.

Hypothesis: The outputs of generative AI tend toward high stylization and verbosity because high correlation can only be achieved with high detail. Is this the most effective way of stating my hypothesis? How can I even start going about proving it?

Great blog from the Head of AI, Kalaivani K G at Exafluence. Fully admitting my own bias - it is great to see a #female voice speaking up in this space. Kalaivani K G makes the important point here that #AI systems do not understand context, nuance or complexity - they are #statistical systems predicting the most likely response. Looking forward to hearing more from Exafluence! #AI | #MachineLearning | #LLM | #WomeninAI

The Art of Case-Based Reasoning: How to Apply it to Real-World Scenarios Case-Based Reasoning uses past cases to solve new problems, improving AI and business decisions with efficiency and adaptability. https://vist.ly/3mha9q9

"In this blog, our Head of AI, Kalaivani K G, shares insights on her journey with Generative AI use cases. She delves into the challenges of AI 'hallucinations'—the unintended, misleading outputs—and offers practical strategies for managing them effectively. Discover how Kalaivani’s experiences help bridge the gap between GenAI potential and real-world application, shedding light on what’s fact versus fiction in this rapidly evolving field." #genai #hallucinations #applications #exafluence

Gen AI Fact Vs Fiction - Real world tips, tricks and insights on Gen AI for Enterprises

Kreasof AI has a very interesting project to challenge the status quo of the current Transformers-based landscape. We present an idea called "Internal Latent Loop" few months ago, and now we polished the idea with some stabilization technique so we called it "Stable Latent Reasoning". We challenge the perspective of most AI practitioners that think the Transformer model can't do more than System 1 thinking (fast thinking). We present a way to scale up the inference in the Transformers model so they theoretically can do an infinite amount of calculation for the same prediction. We add some stabilization techniques, borrowed validated approaches such as Diffusion models, ResNet, and Relaxed Recursive Transformers. We know this is a radical and bold perspective, but we have enough theoretical analysis to prove our point. For a detailed explanation and analysis, kindly check our explanation: https://lnkd.in/gNAHqCqu