A naukri.com initiative
AI News
Hackernoon
1d
253
Image Credit: Hackernoon
OpenAI Makes it Easier to Build Your Own AI Agents With API
- OpenAI's Assistance API provides a solution to make it easier for building AI agents for advanced integrations, document retrieval, executing python code for computations, larger context limit, and more. It addresses limitations of chat completion models such as lack of persistent message history, direct handling of large documents, challenging tasks for coding, limited context windows, and synchronous processing. Through OpenAI Assistance API, users can create sophisticated AI assistants featuring tools like code interpreters, function calling, and thread handling.
- Chat completion models like GPT-4o and GPT-4 are simple as they expect a sequence of messages as input. These models are synchronous and expect a single response after the question is asked where in the Assistance API, you can request multiple requests in parallel and combine the results without careful orchestration.
- The Assistance API dynamically selects which message is to be included as context and thus reduces the distance between the previous conversation and current, making it possible for the API to process longer and larger interaction smoothly.
- With the Assistance API, document retrieval involves dividing the text into small chunks, converting them into embeddings, storing them in a vector database, and retrieving the relevant chunks at query time.
- Code interpretation allows the assistant to use Python code in response to requests like reversing a string, finding today's date. The assistant, in turn, does not solely rely on token predictions.
- The OpenAI Assistance API keeps track of message history, supports document retrieval, executes Python code, manages larger contexts and enables function calling for advanced integrations.
- The threaded messaging feature of the Assistance API allows the previous message content to be stored in threads, allowing assistants to keep the conversation context in multiple turns of conversation. Assistance API current supports GPT-4 (1106 preview) and will support custom fine-tuned models in the future.
- The Assistance API offers solutions that address the core limitations of standard chat completions in real-time computation, document-based Q&A, or dynamic interactions in AI applications.
- By using instructions, threads, tools, and function calling, users can create AI assistants that can easily handle everything from reversing string to advanced integrations.
- OpenAI Assistance API brings new possibilities for building sophisticated AI-driven systems for real-world scenarios and making it easier to build AI agents.
Read Full Article
15 Likes
TechCrunch
1d
130
Image Credit: TechCrunch
Google is using Anthropic’s Claude to improve its Gemini AI
- Contractors working to improve Google's Gemini AI are comparing its answers against outputs produced by Anthropic's competitor model Claude.
- Gemini contractors use the internal Google platform to compare Gemini to other unnamed AI models and have noticed references to Anthropic's Claude.
- Claude's safety settings are stricter than Gemini's and it avoids certain prompts it considers unsafe.
- Google has not disclosed whether it obtained Anthropic's approval to access Claude.
Read Full Article
7 Likes
TechBullion
1d
272
Image Credit: TechBullion
Transforming Creativity with the AI Spicy Story Generator
- The AI Spicy Story Generator offered by My Spicy Vanilla is revolutionizing storytelling by combining artificial intelligence with creativity.
- This tool allows users to generate customized and unique stories based on their preferences of genre, tone, characters, and themes.
- The generator excels at crafting engaging narratives with unexpected plot twists and dynamic character arcs.
- It is not only beneficial for professional writers but also serves as an educational tool and provides entertainment for all users.
Read Full Article
16 Likes
Medium
1d
303
Image Credit: Medium
ZOMATO’s Secret Sauce: Grouping Unique Address Using SBERT
- Zomato used SBERT for text-based clustering of addresses, eliminating the issue of word embedding alone not being able to solve the sequencing of words.
- SBERT allowed for uniformity in processing addresses of varying lengths, resulting in embeddings of a consistent size and meaningful representation of sentences.
- These embeddings of one fixed-length vector can be clustered into one string using DBSCAN, generating one final label for the different addresses of the same location that customers enter.
- SBERT is a type of artificial intelligence model that learns words and their meanings like BERT, but also checks if two sentences mean the same, offering a big idea of what it learned, instead of talking about every single word.
- The Siamese network structure of SBERT, where two identical BERT models share weights, allows for a direct comparison between two input sentences or addresses.
- Encoders in Transformers are used to analyze and encode input sequences into a rich, contextual representation, while decoders generate the output sequence step by step using information from the encoder.
- Transformers process all input tokens simultaneously, making training faster and more efficient than traditional recurrent neural networks.
- Zomato's use of SBERT enabled it to group unique addresses, reducing discrepancies in cost calculations for delivery and the time and resources wasted by delivering to the same address but to different groups of people, which can be harmful for any last-mile delivery aggregator.
- SBERT's fixed-length embeddings offer a meaningful representation of entire sentences, enabling the clustering of addresses of varying lengths with a consistent size.
- Large Language Models, such as Chat-GPT and BERT, are trained on vast amounts of data and used for a wide range of language-related tasks, including understanding and generating human-like text.
Read Full Article
18 Likes
Medium
1d
49
Image Credit: Medium
Scaling Smarter: An Overview of Large Language Models (LLMs) and Their Compression Techniques Part…
- Part 1 provides an overview of LLMs, discussing their advantages, disadvantages, and use cases.
- Some important LLM models/frameworks/tools with pros, cons, and use cases are listed below. The ones given are GPT-3.5, GPT-4, GPT-2, LLaMA 2, Alpaca, DistilBERT, MiniLM, TinyBERT, BERT, Sentence-BERT, RoBERTa, Faiss (Facebook AI Similarity Search), ONNX Runtime, TensorRT, Hugging Face Transformers, Transformers.js, and ggml.
- Type: Large Transformer-based LLM
- Type: Medium-sized LLM
- Type: LLM
- Type: Fine-tuned LLaMA
- Type: Transformer-based LLM
- Type: Transformer-based LLM
- Type: Transformer-based LLM
- Type: Sentence Embedding Model
Read Full Article
3 Likes
Medium
1d
219
Image Credit: Medium
The Beauty of ChatGPT Poetry: Strikingly Elegant Yet Missing Two Vital Elements
- AI-generated poetry appears flawless and captivating on the surface but lacks authenticity and personal experience.
- ChatGPT's poetry lacks the ability to feel emotions or appreciate the beauty of life.
- Its creations, though beautiful, lack the passion and personal truths found in human poetry.
- AI-generated poetry can mimic structure, but it falls short in delivering the emotional depth and shared human experience.
Read Full Article
13 Likes
Tech Story
1d
372
Intel Announces CES 2025 Keynote: Set to Compete with AMD and NVIDIA
- Intel announces CES 2025 keynote, scheduled for January 6, 2025.
- Intel expected to showcase 14th-Gen Meteor Lake processors, AI-powered solutions, expansion of Arc GPUs, and next-gen data center innovations.
- AMD and NVIDIA also set to deliver keynotes on January 6, showcasing new processors, GPUs, and AI advancements.
- CES 2025 provides Intel an opportunity to battle market challenges, showcase AI innovations, and reinforce brand leadership.
Read Full Article
22 Likes
Dev
1d
242
Image Credit: Dev
Async Pipeline Haystack Streaming over FastAPI Endpoint
- This tutorial explains how to use Server-Sent Events (SSE) in a Python-based pipeline and how to serve the processed query results over an endpoint using FastAPI with an asynchronous, non-blocking solution.
- The post describes a workaraound approach to create a pipeline task and set "sync" streaming callbacks on the event loop for chunk collection and yield the chunks in a server-sent event.
- The pipeline is designed synchronously, and components to the pipeline can be added dynamically. The API KEY is passed through the end-point, and the openai generator is used to create a pipeline component, which is used to generate responses for user input.
- The AsyncPipeline is defined to run the pipeline, and the server-sent event is used to stream the generated answers in SSE format.
- The ChunkCollector is defined to handle and queue the generated answers and yield them in SSE formatting in an end-point.
- The end-point can be served using fetch-event-source as a frontend to display the streams of generated answers.
- The post concludes by suggesting that the use of sockets would be useful considering performance issues while handling a large volume of data.
- The packages required for the tutorial include fastapi, uvicorn, haystack-ai, haystak-experimental, pydantic, and python, above version 3.10 and below 3.13.
- The complete code, snippets, and a full explanation for each function are provided above.
- The tutorial is meant for experts who are familiar with FastAPI and python programming languages, as it does not provide a guide for the FastAPI process.
Read Full Article
14 Likes
Medium
1d
226
What Goes Beyond the Prompt?
- Tokenization provides structure for the AI to process the input.
- Transformers use self-attention to handle different cases.
- Transformers process words in parallel, enabling faster computations and improved context analysis.
- The Transformer architecture consists of two main parts.
Read Full Article
13 Likes
Medium
1d
188
Image Credit: Medium
Top 5 OpenAI o1 and o3 Alternatives — Best Reasoning Models
- GenPixy Chat Reasoning Lite is a free and no-sign-up required alternative to OpenAI o1, offering impressive reasoning capabilities for up to 15 minutes in a single request.
- DeepSeek R2 Lite is a research preview model that provides advanced reasoning capabilities, giving users a glimpse of the future of AI reasoning.
- Qwen QwQ 32B Preview is an open-source model available through Hugging Face's chat interface, boasting 32 billion parameters for powerful reasoning.
- GenPixy Chat Reasoning Full is a more mature version with premium features, offering a balanced mix of accessibility and power for more demanding applications.
- Gemini Flash 2.0 Thinking is a streamlined AI development model accessible through Google AI Studio, providing a user-friendly experience for integrating powerful AI reasoning into projects.
Read Full Article
11 Likes
Hackernoon
1d
242
Image Credit: Hackernoon
What Capitalists Got Wrong About the 'Future of Education,' And What They Should Have Done Instead
- The article argues that the vision of education prioritizing tech skills is flawed and alienates people who cannot afford expensive EdTech memberships or fit into the 'one-size-fits-all' pipeline. Those who cannot keep up with the relentless march of innovation are left behind, and society prioritizes productivity over humanity. The article suggests that skills such as problem-solving, innovation, and tech fluency are more like slippery buzzwords, while storytelling, caretaking, and emotional labor have been sidelined. The future of education should not only be about skills but should seek a diverse range of perspectives and skills to build a society that works for humans.
- The article criticizes the consensus around the future of education, arguing that it is optimized for productivity rather than humanity. Society's obsession with progress often comes with a price tag that not everyone can afford.
- The article asserts that the current blueprint for education lacks diversity, only catering to students who fit the narrow mold of tech-savvy whizzes who can code effortlessly and manage digital classrooms with ease.
- The article argues that the tech-driven future of education risks building a marathon where only a few have running shoes while the rest are left on the sidelines, unable to catch up. The global consensus around the future of education runs the risk of exclusionary rather than aspirational.
- The article argues that care taking, emotional labor, and storytelling have been sidelined, and society is losing out on the soul of society, building a future where people are valued based on how well they fit into the machine rather than their humanity.
- The article argues that the assumption that the only skills that matter are those tied to technology and productivity is flawed. The focus should not be on 'future-proofing' skills but on designing a future that caters to humans' diverse needs, cultures, and traditions.
- The article concludes that the future of education is about making sure everyone is onboard, and it should not be a relentless race with no finish line in sight. The question should not be about how to prepare workers for the future, but how to design a future that works for humans.
Read Full Article
14 Likes
Dev
1d
30
Image Credit: Dev
Run LLMs Locally with Ollama & Semantic Kernel in .NET: A Quick Start
- In this post, you’ll learn how to set up Ollama on your machine, pull and serve a local model like llama3.2. and integrate it with Semantic Kernel in a .NET 9 project.
- Ollama is a self-hosted platform for running language models locally. It eliminates the need for external cloud services and offers data privacy, lower costs, and ease of setup benefits.
- Semantic Kernel is an open-source SDK from Microsoft that enables developers to seamlessly integrate AI capabilities into .NET applications.
- The article lists the prerequisites to run Ollama and Semantic Kernel locally in a .NET 9 project.
- The article shares a step-by-step integration guide for running locally hosted AI models.
- The article also offers sample code and output to help understand how local, generative Artificial intelligence works with Ollama and Semantic Kernel.
- Running AI models locally offers a few use cases including prototyping without incurring cloud costs, internal knowledge bases, and edge or offline applications.
- Combining Ollama and Semantic Kernel lays a foundation for building self-contained, high-performance .NET applications that help maintain complete control over environment and reducing both complexity and recurring costs.
- Further experimentation with Ollama and Semantic Kernel is encouraged, along with experimentation with different models in Ollama.
- Upcoming posts will tackle Retrieval Augmented Generation(RAG) to give LLMs context-aware responses sourced from your own data-all running entirely on local infrastructure.
Read Full Article
1 Like
Unite
1d
138
Image Credit: Unite
Understanding Shadow AI and Its Impact on Your Business
- The rapid adoption of AI by businesses presents a hidden challenge, the emergence of Shadow AI.
- Shadow AI refers to the use of AI technologies and platforms that haven't been approved or vetted by the organization's IT or security teams.
- Over 60% of employees admit using unauthorized AI tools for work-related tasks.
- There are several risks associated with Shadow AI, such as data privacy violations, regulatory noncompliance, and operational and reputational risks.
- Shadow AI is becoming more common due to limited organizational resources, misaligned incentives, and the use of free tools.
- Shadow AI appears in multiple forms, such as chatbots, data analysis models, marketing automation tools, and generative AI applications.
- Managing the risks of Shadow AI requires a focused strategy emphasizing visibility, risk management, and informed decision-making.
- Organizations should establish clear policies, classify data and use cases, acknowledge benefits, educate and train employees, monitor and control AI usage, and collaborate with IT and business units.
- The future of AI will rely on strategies that align organizational goals with ethical and transparent technology use.
- To learn more about managing AI ethically, stay tuned to Unite.ai for the latest insights and tips.
Read Full Article
8 Likes
Nvidia
1d
265
Image Credit: Nvidia
From Generative to Agentic AI, Wrapping the Year’s AI Advancements
- The article summarizes the latest AI advancements and how the technology has changed the way people write, game, learn and connect with each other online, using tools like ChatGPT.
- NVIDIA offers foundation models for enthusiasts and developers, including Gemma 2, Mistral and Llama-3, which can run locally on NVIDIA GeForce and RTX GPUs.
- The introduction of RTX-Accelerated partner applications has made it easier than ever to run models locally in popular applications without relying on the cloud.
- Brave Browser’s Leo AI, powered by NVIDIA RTX GPUs and the open-source Ollama platform, offers fast, responsive AI performance while keeping user data private.
- Agentic AI is the next frontier of AI, capable of using sophisticated reasoning and iterative planning to autonomously solve complex, multi-step problems.
- The AnythingLLM partner application showcases how AI is going beyond simple question-answering to improve productivity and creativity.
- AnythingLLM lets users interact with documents through intuitive interfaces, automate complex tasks with AI agents and run advanced LLMs locally.
- Over 600 Windows apps and games today are already running AI locally on more than 100 million GeForce RTX AI PCs and workstations worldwide.
- Tune into the CES keynote delivered by NVIDIA founder and CEO Jensen Huang on Jan. 6. to discover how the latest in AI is supercharging gaming, content creation and development.
- Generative AI is transforming gaming, videoconferencing and interactive experiences of all kinds. Make sense of what’s new and what’s next by subscribing to the AI Decoded newsletter.
Read Full Article
15 Likes
TechBullion
1d
3
Image Credit: TechBullion
How to Use Sora AI: A Guide to OpenAI’s Revolutionary Video Generator
- Sora by OpenAI is an AI-powered video generator that transforms text inputs into vivid video scenes.
- It utilizes advanced technology, including a diffusion model and a transformer architecture.
- Sora is currently available to OpenAI's Red Team testers and select creative artists.
- Users can input detailed text prompts to generate videos and customize dimensions.
Read Full Article
Like
For uninterrupted reading, download the app