AI4Value’s Post

💡 Now reading:VLMs are blind 🚀 Abstract: Large language models with vision capabilities (VLMs), e.g., GPT-4o and Gemini-1.5 Pro are powering countless image-text processing applications and scoring high on existing vision-understanding benchmarks. Yet, we find that VLMs fail on 7 visual tasks absurdly easy to humans such as identifying (a) whether two circles overlap; (b) whether two lines intersect; (c) which letter is being circled in a word; and (d) counting the number of circles in an Olympic-like logo. The shockingly poor performance of four state-of-the-art VLMs suggests their vision is, at best, like that of a person with myopia seeing fine details as blurry, and at worst, like an intelligent person who is blind making educated guesses. 🌐 https://lnkd.in/ekABJdDs 📄 https://lnkd.in/ewBHnRGH #vlm #deeplearning #machinelearning #visionlanguagemodels #artificialintelligence

The Rise of Large Language Models: What's the Fuss About? Artificial Intelligence has surged in recent years, and at the heart of it are large language models (LLMs). #AIcomparison #AIinbusiness #AImodels #AIrealworldapplications #GPT4 #GPT4usecases #largelanguagemodels #MachineLearning #Reflection70B #Reflection70Bperformance

Prompt Engineering Challenges with Large Language Models (e.g., GPT-4 Turbo) While longer prompts seem logical for complex tasks, they can lead to issues like: 🔥 Focus on beginning & end: The model prioritizes info at the start and finish, neglecting the middle. 🔥 Diminishing returns with examples: Adding too many examples can overwhelm the model. Discover strategies to overcome these challenges and optimize your prompts for large language models - LINK in the comments now! 🔗👇 #LLMs #PromptDesign #AIBestPractices

OpenAI Release Powerful New Language Model: GPT-4o  https://buff.ly/4dESJQ3

Looking to cut costs when working with large language models? This super-simple trick cut costs by over 20% without any statistically significant impact on Accuracy. “New research shows that adding "be concise" to chain-of-thought prompts shortens Al responses by 50% without a statistically significant impact on accuracy, saving more than 20% on the cost of API calls. Chain-of-thought prompting tells the model to provide its step-by-step reasoning before giving final answer, which can improve accuracy on complex problems. The research tested on 1,000 multiple-choice questions across 10 topics, using @OpenAI's GPT 3.5 and GPT 4.”

Selective State Spaces (SSMs) offer a promising alternative to large language models like GPT, addressing key challenges such as computational inefficiency with long sequences and high energy use. SSMs efficiently process long data sequences and focus on the most relevant information, making them faster and more resource-efficient. This adaptability makes SSMs ideal for tasks like legal document review, e-discovery, and legal research. Models like Mamba, which leverage SSMs, are looking like they outperform traditional models in handling extensive data, offering significant improvements in performance and practicality for legal tech applications. Have a deeper read about SSMs in my recent blog post: #ssm #llms #legaltech #artificialintelligence

Hierarchy of Attacks on LLMs

This post shares prompting strategies and tactics for getting better results from large language models (sometimes referred to as GPT models) like GPT-4, Geminini or Mistral. The methods described here can sometimes be deployed in combination for an even greater effect. We encourage experimentation to find the methods that work best for you. Are you already familiar with some of them? What is your favorite technique? Let us know in the comments! #Digitalbricks #aieducation #changemanagement #Prompting #promptengineering

If you wonder how Large Language Models (LLMs) work and aren’t afraid of getting a bit technical, check out this step-by-step walk-through of a GPT large language model complete with an animated and interactive 3D block diagram of everything going on under the hood.

Hey all, I'm excited to share that our recent paper titled "Crafting Clarity: Leveraging Large Language Models to Decode Consumer Reviews" has been accepted in the Journal of Retailing and Consumer Services (ABDC - A, SSCI, Impact factor: 11.0). This is one of the interesting works I have published so far. In this study, we utilized 1,031,478 consumer reviews and trained four prominent large language models (Falcon-7B, MPT-7B, GPT-2, and BERT) to assess their performance in understanding consumer sentiment. Our findings show that Falcon-7B performed the best, while GPT-2 performed the worst in decoding consumer emotions. Additionally, we performed topic modeling on the reviews using both LDA and Falcon-7B. By comparing the results, we emphasize the necessity of a powerful deep learning model that can understand semantic relationships to comprehend consumer perceptions, compared to a Bayesian probabilistic approach which relies on the likelihood of observing specific words within topics. To the best of my knowledge, this study represents the first comprehensive evaluation of large language models' ability to decode consumer reviews.  Thanks to my co-authors Pranshav Gajjar Ashutosh Dutt for working along with me in this project.