Artificial Intelligence

From discussing the latest in machine learning algorithms to sharing insights on neural networks and natural language processing, this section is your gateway to the realm of AI.

41 Topics
16 Replies

Anbu VExplorer

posted in Artificial Intelligence

A Deep Dive into LLM Guardrails

Image generated using Microsoft Copilot In the rapidly evolving landscape of artificial intelligence, Large Language Models (LLMs) have emerged as powerful tools with far-reaching implications. As these models become increasingly integrated into our daily lives, the need for robust safety measures has never been more critical. Enter guardrails - sophisticated algorithms designed to act as digital sentinels, safeguarding the interactions between humans and AI.The Evolution of AI SafetyIn the pre-LLM era, AI safety primarily relied on white box techniques. These methods focused on creating transparent and interpretable models, allowing researchers and developers to understand the internal workings and decision-making processes. However, the advent of LLMs has shifted the paradigm towards more complex, black box systems. This transition has brought about new challenges in ensuring AI safety. The focus has moved from inherent model transparency to post-hoc strategies, notably the implemen

jrgosalvezCommunity Manager

published in Artificial Intelligence

To GPT4o or Not to GPT4o?LLM

OpenAI released GPT-4o a conversational chat-bot last week.It is integrated into the online chat-bot as well as an API for integration with 3rd party systems. Early findings illustrate it is twice as fast and half the cost, with higher rate limits for better performance. https://vimeo.com/945586717 The multi-group query decoder architecture performs well based on zero-shot prompts, which are plain prompts that do not provide context in the prompt to help the LLM improve its response. The API leverages the existing OpenAI chat completion API, which requires an API key to load and interact with the model. Many additional parameters, such as temperature, can be tuned. from openai import OpenAIOPENAI_API_KEY = "<your-api-key>"client = OpenAI(api_key=OPENAI_API_KEY)response = client.chat.completions.create( model="gpt-4o-2024-05-13", messages=[ {"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "Who won the world series in 2020?"},

jrgosalvezCommunity Manager

posted in Artificial Intelligence

GPUs in the Himalayas

A recent video post by Akash James of SparkCognition shows how he was able to get the GPU he needed to work on his multi-modal LLM called VERsatile ONline Intelligent Cognitive Agent (VERONICA) by accessing compute he owned in India from the Himalayas. Join the Z by HP Boost webinar Oct 4th @ 9 am PST to hear how Akash is accessing his GPU remotely to create AI.

ArgajitExplorer

asked in Artificial Intelligence

Suggest Workstation under 30 Lakhs(₹) for Ai and deep learning

I'm a Researcher, We want to work on Artificial intelligence, Deep learning and Machine Learning algorithms on healthcare databases. We have a budget of about 30,00,000/- Lakhs Indian rupees to spend. Please suggest a Workstation and an NVIDIA GPU under this budget.

curtisburkhalterHP Employee

published in Artificial Intelligence

AI Evals Are a Piece of Cake!

Now that I have your attention with that suspiciously simple title, let's be honest: evaluating AI systems is anything but a piece of cake. However, thinking about AI evaluations like a layered cake can help us understand the various levels of assessment needed for a robust evaluation framework.Just as a master baker carefully constructs a wedding cake with distinct layers - each contributing its own flavor and texture to the whole - AI evaluation requires multiple layers of assessment working in harmony. From the foundation layer of basic accuracy metrics to the delicate frosting of user experience evaluation, each layer serves a crucial purpose.Think of it this way: You wouldn't serve a cake without testing each component. The base needs to be sturdy, the filling must complement each layer, and the frosting should tie everything together. Similarly, your AI system needs evaluation at every level - from individual model responses to end-to-end system performance.By the end of this art

Anbu VExplorer

posted in Artificial Intelligence

Training AI Agents to Self-Correct: A Deep Dive into Agent-R’s Theoretical Foundations

Imagine teaching a robot to boil water. It might first search for a pot, fill it with water, then place it on the stove. But what if it mistakenly turns on the microwave instead of the stove? Traditional AI agents often struggle to recover from such errors, leading to cascading failures. This is the problem Agent-R—a novel framework for training self-reflective language model agents—aims to solve.In this article, we’ll unpack the theoretical backbone of Agent-R by focusing on how this framework redefines error recovery in AI.Why Error Recovery Matters in Interactive EnvironmentsMost AI agents learn by mimicking expert trajectories (e.g., perfect step-by-step guides). But real-world tasks are messy. Errors are inevitable, and waiting until the end of a task to correct them is like letting a typo in a sentence propagate into a garbled paragraph.The Core Challenge: Timely InterventionProblem: In multi-step tasks (e.g., crafting items in Minecraft or navigating a virtual lab), errors early

Anbu VExplorer

posted in Artificial Intelligence

Shared Recurrent Memory Transformer

Multi-agent systems have become increasingly important in solving complex problems through distributed intelligence and collaboration. However, coordinating multiple agents effectively remains a significant challenge, particularly in the field of reinforcement learning. In this article, we'll explore a novel approach called Shared Recurrent Memory Transformer (SRMT) that aims to enhance coordination in decentralized multi-agent systems.The Challenge of Multi-Agent CoordinationImagine a group of robots trying to navigate through a crowded warehouse. Each robot has its own goal and can only see what's immediately around it. How can they work together efficiently without bumping into each other or getting stuck? This is the essence of the multi-agent pathfinding problem.Traditional approaches often struggle with this task, especially when the environment is complex or the number of agents is large. They may rely on centralized control, which doesn't scale well, or on explicit communicatio

jrgosalvezCommunity Manager

published in Artificial Intelligence

DeepSeek for Local AI Creation

DeepSeek is an open source, open weight set of pre-trained models people can use for inference through the DeepSeek web app on chat.deepseek.com by agreeing to their terms of service and privacy policies or training by downloading the models. Similarly to OpenAI, the privacy policy lets users know that data and logic shared is collected and fair game for the company to retain and retrain their models, so be careful what you share. However, since DeepSeek is open source, AI creators can download models locally and set-up to use for local AI development and retraining of weights to efficiently create and tune focused models for optimized functions and tasks. Check out DeepSeek’s open models on Github at github.com/deepseek-ai.Mike Le Galloudec at Oakland shares a great 6 min review and will be sharing how to get R1 running locally on private systems.Are you doing something similar? Share how you’re getting DeepSeek models running locally!

Anbu VExplorer

posted in Artificial Intelligence

Titans: Learning to Memorize at Test Time

A research paper by Ali Behrouz, Peilin Zhong and Vahab Mirrokni from Google Research introduce a groundbreaking neural network module that addresses the limitations of current models in handling long term dependencies. This work is a significant leap forward in the design of memory augmented architectures, offering theoretical and practical advancements in sequence modeling. Below we explore the theoretical underpinnings, technical innovations and experimental results of this advanced neural network module.The Challenge of Long-Term DependenciesModern neural architectures like Transformers and recurrent neural networks (RNNs) have revolutionized sequence modeling. However, they face some challenges:Transformers: While effective at capturing short term dependencies through attention mechanisms, they are computationally expensive for long sequences due to quadratic complexity in context length. Their context window limits their ability to model long term dependencies Recurrent Models

Anbu VExplorer

posted in Artificial Intelligence

Inside AI’s Black Box: Mechanistic Interpretability as a Key to AI Transparency

Image generated using CopilotArtificial intelligence (AI) systems are becoming increasingly sophisticated, capable of generalizing across tasks and domains. However, this complexity makes them harder to understand, raising concerns about their trustworthiness and safety. Mechanistic interpretability offers a promising solution by reverse-engineering neural networks to uncover how they process information internally. Just as neuroscience advanced our understanding of the brain by studying internal cognitive processes, mechanistic interpretability seeks to move beyond black-box methods to provide precise, causal explanations of AI behavior.This article explores the principles, methods, and implications of mechanistic interpretability. By examining its core concepts, such as features, circuits, and hypotheses like superposition and linear representation, we aim to shed light on how this field can contribute to AI safety while addressing the challenges it faces.Mechanistic interpretabilit

jrgosalvezCommunity Manager

asked in Artificial Intelligence

How Private are Your Chats?

With the proliferation of chatbots and agents, users are rushing to learn more about the technology. The speed with which GenAI is growing is fueled by the data with which the AI has access to. Chatbots are curating a combination of synthetic and user based content to collect questions, data, chain of thought, and context.Do you believe sharing your data and logic online with AI means giving up rights to your IP?

Anbu VExplorer

posted in Artificial Intelligence

The rapid evolution of Large Language Models (LLMs) has ushered in a new era of AI-powered chat assistants, leveraging techniques like supervised instruction fine-tuning and reinforcement learning with human feedback (RLHF) to enhance their instruction-following and conversational abilities. These advancements have led to models that are increasingly aligned with human preferences, often outperforming their unaligned predecessors in user satisfaction. However, this progress has brought forth a significant challenge: how to effectively evaluate these sophisticated models. Traditional benchmarks, which typically focus on core capabilities such as knowledge retrieval and problem-solving, fall short in assessing how well these models align with human preferences in real-world applications. We've observed instances where models excel in benchmark tests yet underperform in practical scenarios that demand alignment with human expectations.While human ratings remain the gold standard for eval

alexander.klugHP Employee

published in Artificial Intelligence

📅 GenAI Productionize 2.0 on October 29th

The premier conference for GenAI application development returns on October 29th! Join us at GenAI Productionize 2.0 for an incredible lineup of speakers from HP, Galileo, Databricks, Twilio, Walmart and more.This industry-first summit is all about productionizing generative AI in the enterprise. Gain insights from top AI experts across research labs, startups, and global brands on governance, operational frameworks, and practical evaluation techniques.Don’t miss out—register now: GenAI Productionize 2.0 Registration

janetteguzmanHP Employee

posted in Artificial Intelligence

AI Innovations to Watch for from HP Imagine – Get Ready

With the wrap-up of HP Imagine yesterday, AI and remote work took center stage in every new product announcement. One key highlight was the introduction of Z by HP Boost, a solution that enables teams to share on-demand GPU power from their combined fleet of local workstations. Additionally, HP is expanding AI Studio with the launch of the Z by HP Gen AI Lab, strengthening LLM development with a “built-in layer of trust” by protecting against hallucinations and bias.HP also announced several exciting AI-powered PC, audio, and visual products, including the new HyperX Cloud MIX™ 2 headset, Poly Labs, and the HP OmniBook Ultra Flip 2-in-1 Laptop Next Gen AI PC, to name a few that will elevate both work and play experiences.🎉 Overall, these are exciting strides in AI for HP, coming in November 2024 and the first quarter of 2025.Check out the full press release here: HP Transforms the Future of Work | HP® Official Site

ptr.nStarter

published in Artificial Intelligence

OpenAI Release o1: Its First Model with Advanced Reasoning Capabilities

OpenAI has announced the release of its latest AI model, codenamed Strawberry, now officially named OpenAI o1. This new model is designed to improve how AI handles complex tasks in math, science, and coding by evaluating its steps before providing a response.Key Updates: Advanced Reasoning: Unlike previous models, OpenAI o1 takes more time to think through problems, emulating a more human-like reasoning process. This approach helps the model better handle difficult questions and adhere more closely to safety guidelines. New Naming and Features: OpenAI o1 introduces a new naming system and will be available alongside existing models like GPT-4o. There’s also a lightweight variant, o1-mini, focused on code generation. Rolling Out Access: The new model is being introduced in phases. Starting today, ChatGPT Plus and Team users will get limited access to both o1-preview and o1-mini, with educational and enterprise users gaining access next week. Current Limitations: While promising, O

miracFenceZ by HP Data Science Ambassador

posted in Artificial Intelligence

Hello everyone!I just created a blog post about exploring current Medical LLMs. If you are interested about AI in Healthcare and GenAI, and would like to learn how to create demos with Python and Gradio, I invite you to read it! https://medium.com/@miracfence/exploring-fine-tuned-medical-llms-cbb42dd4911f

SothanCommunity Manager

posted in Artificial Intelligence

Gartner Forecasts 40% of Generative AI Solutions Will Embrace Multimodal Capabilities by 2027

According to Gartner, by 2027, 40% of generative AI (GenAI) solutions will feature multimodal capabilities—encompassing text, images, audio, and video—up from just 1% in 2023. This transition from single-modal to multimodal models promises to improve human-AI interactions and create unique opportunities for differentiated GenAI-powered products and services.The Rise of Multimodal GenAIMultimodal Generative AI is set to revolutionize enterprise applications by introducing innovative features and capabilities previously unattainable. Its influence spans across various industries and applications, enhancing interactions at every point where AI meets human engagement. Currently, many multimodal models incorporate only two or three modalities, but this is expected to expand significantly in the coming years.Open-Source LLMsOpen-source large language models (LLMs) are transforming enterprise value by democratizing access to advanced generative AI. They enable businesses to tailor models for

miracFenceZ by HP Data Science Ambassador

posted in Artificial Intelligence

AI in the Film Industry

As it is already well-known, for the last couple of years companies have been using AI in films in order to optimize multiple processes, from computer graphics, to animation, etc. Today, however, we can see it being hugely involved in the process, even in the whole automation of creating a film. For some artists, this may be a huge concern, for others it MAY be an opportunity to explore their creativity. In posts and programs like this https://runwayml.com/news/runway-partners-with-tribeca-festival we can see the potential in GenAI applications in the film industry, as long as the latest research https://runwayml.com/research/introducing-gen-3-alpha What do you think that are the actual limits of this? Do you think it will eventually ceil and just become one more tool for humans to explore their own creativity? Or, do you think it will become something like that one Black Mirror’s episode “Joan is Awful” where content is being automatically generated without any restrictions, just for

SothanCommunity Manager

posted in Artificial Intelligence

Google Unveils 'GameNGen': AI-Powered Game Engine Recreates Doom in Real Time

Google researchers have unveiled GameNGen, a groundbreaking game engine entirely driven by AI that can simulate intricate video games in real-time. This innovative technology was showcased by recreating the iconic first-person shooter, Doom.GameNGen employs a neural network to generate game frames at more than 20 frames per second, creating visuals that closely mirror the original game. The AI model underwent a two-phase training process: initially, an AI agent was taught to play Doom, followed by training a separate model to predict the next frame based on previous actions and frames.“We introduce GameNGen, the first game engine entirely powered by a neural network, capable of real-time interaction within a complex environment over extended periods and at high quality,” the researchers detailed in their paper.This technology holds the potential to revolutionize game development by enabling the creation and modification of games through text descriptions or example images, moving away

SothanCommunity Manager

published in Artificial Intelligence

NVIDIA's New AI Model Techniques Slash Costs and Enhance Accuracy with Pruning and Distillation

NVIDIA's research team has introduced an innovative approach to developing smaller, yet highly accurate language models by leveraging structured weight pruning and knowledge distillation. These techniques are designed to streamline the model without compromising its performance, offering several key benefits for developers and organizations focused on AI development.Key Benefits and Innovations: Performance Improvement: The refined models achieve a 16% improvement in MMLU (Massively Multilingual Language Understanding) scores. MMLU is a benchmark used to evaluate a model's understanding and processing of multilingual content, making this improvement significant for applications that require high linguistic accuracy across different languages. Training Efficiency: The method reduces the number of tokens required to train new models by a factor of 40. Tokens are the building blocks of language models, and reducing their number drastically cuts down on computational resources and time n

jrgosalvezCommunity Manager

posted in Artificial Intelligence

Veronica LLaVA-Llama3 Multi-modal LLM Senses and Communicates Surroundings

HP Data Science Ambassador Akash James uses a zero-shot approach to create Veronica to be aware of her spatial surroundings through visual, audio, and text interactions to describe what she sees. Check out his LinkedIn post to see a demo of how it works…

SothanCommunity Manager

posted in Artificial Intelligence

OpenAI brings fine-tuning to GPT-4o with one million training tokens through September 23rd

OpenAI announced the launch of fine-tuning for GPT-4o, addressing one of the most frequently requested features from developers. This new capability allows developers to fine-tune the GPT-4o model with custom datasets, enhancing performance while reducing costs for specific applications.To support the rollout, OpenAI is offering 1 million free training tokens per day for every organization through September 23. This incentive is designed to help developers explore fine-tuning without immediate cost concerns.Developers can begin fine-tuning by visiting the fine-tuning dashboard. The process involves:Accessing the dashboard. Clicking "Create" to start a new project. Selecting gpt-4o-2024-08-06 from the base model options.GPT-4o Mini Also AvailableOpenAI also offers GPT-4o Mini for fine-tuning. This smaller model is available with 2 million free training tokens per day until September 23. To use it, developers should select gpt-4o-mini-2024-07-18 on the fine-tuning dashboard.Data Privacy

SothanCommunity Manager

published in Artificial Intelligence

🎥 Building Trustworthy Gen AI Solutions with Sensitive Data

At the recent AI Summit conference in London, Alex Klug, Product Sr. Manager of Data Science & AI Solutions at HP, and Yash Sheth, COO & Co-founder of Galileo, discussed the challenges of creating trustworthy generative AI systems, particularly when handling sensitive data. They highlighted the widespread use of generative AI (Gen AI) and the significant issue of "hallucinations," where AI systems produce inaccurate or misleading information. This problem is increasingly concerning for businesses due to potential liability and the need for accuracy in AI applications.The discussion emphasized the need to rethink the entire development lifecycle to address these challenges, from model selection and procurement to experimentation, deployment, and monitoring. Solutions must ensure that sensitive data is managed securely and reliably while addressing problems such as hallucinations and model drift.HP’s AI Studio offers secure, local environments for the development and deployment o

Sarita NavuluruExplorer

published in Artificial Intelligence

Galileo's Hallucination Index: Unveiling the Top Performers in Language Models for 2024

2024Galileo's Hallucination Index report is a critical resource for AI developers, offering detailed insights into the performance and cost-effectiveness of various large language models (LLMs). This enables developers to make informed decisions to enhance the reliability and accuracy of GenAI applications. The report highlights the current landscape of LLMs and sets the stage for future advancements.Key FindingsModels Evaluated: The report covers 22 models, including 10 closed-source and 12 open-source models from providers like OpenAI, Anthropic, Meta, Google, and Mistral. Models were assessed based on context length capabilities and source type.Major TrendsOpen Source Improvement: Open-source models are rapidly closing the performance gap with closed-source models. Model Size: Smaller models sometimes outperform larger ones, challenging the belief that bigger is always better. Context Length Performance: Many models maintain high performance even with extended context lengths. Anthr

SothanCommunity Manager

published in Artificial Intelligence

Preliminary Results: Exploring AI Creation and Model Hub Preferences

We’re excited to bring together Data Scientists and IT Decision Makers (ITDMs) to explore the landscape of model hubs and compute systems. Our focus is on understanding the preferred model hubs that Data Scientists use for accessing pretrained models, as well as the compute systems they rely on for AI development and creation. As our research progresses, we’re eager to share some preliminary insights with you. Here’s an update on our early findings:Model Hubs (DS only)Data scientists most commonly use platforms such as Microsoft AI GitHub, Google Model Garden, AWS Model Zoo, and PyTorch Hub. Hugging Face 25% Tensorflow Hub 19% Pytorch Hub 31% Microsoft AI GitHub 38% Model Zoo by OpenAI 6% ONNX Model Zoo 6% NVIDIA NGC 6% Google Model Garden 38% AWS Model Zoo 31% TorchVision Model Zoo 13% Qualcom AI Hub 6% Model Zoo for Int

1
2

Page 1 / 2