CybAI news - - Page 3173

23 May

To optimize guide-dog robots, first listen to the visually impaired

stp2y0 CommentsGenAIEye Care; Disability; Workplace Health; Robotics Research; Engineering; Vehicles; Robotics; Artificial Intelligence; Neural Interfaces

What features does a robotic guide dog need? Ask the blind, say the authors of an award-winning paper. Led by researchers at the University of Massachusetts Amherst, a study identifying how to develop robot guide dogs with insights from guide dog users and trainers won a Best Paper Award at CHI 2024: Conference on Human Factors in Computing Systems (CHI). Guide dogs enable remarkable autonomy and mobility for their handlers. However, only a fraction of people with visual impairments have one of these companions. The barriers include the scarcity of trained dogs, cost (which is $40,000 for training alone), allergies…

23 May

AI, Platforms, And Big Promises: What We Saw At Red Hat Summit

stp2y0 CommentsNewsai, Ansible, Linux, Red Hat

IT platform teams’ responsibilities in 2024 are expanding radically. They’re being asked to provide cost-effective alternative solutions for the VMware virtualization technology that typically dominates enterprise IT. While they’re at it, could they please also stand up state-of-the-art AI platforms that can support generative AI in the data center and cloud that work as well as big cloud-managed AI services but for less money? Those were the big questions hanging over some 6,000 attendees of Red Hat Summit in Denver May 6–9. Red Hat responded with a series of announcements of offerings — some immediate, some long-term — that are…

23 May

Asymmetric Certified Robustness via Feature-Convex Neural Networks

stp2y0 CommentsAIasymmetric robustness, certified robustness, input-convex neural networks

Asymmetric Certified Robustness via Feature-Convex Neural Networks TLDR: We propose the asymmetric certified robustness problem, which requires certified robustness for only one class and reflects real-world adversarial scenarios. This focused setting allows us to introduce feature-convex classifiers, which produce closed-form and deterministic certified radii on the order of milliseconds. Figure 1. Illustration of feature-convex classifiers and their certification for sensitive-class inputs. This architecture composes a Lipschitz-continuous feature map $varphi$ with a learned convex function $g$. Since $g$ is convex, it is globally underapproximated by its tangent plane at $varphi(x)$, yielding certified norm balls in the feature space. Lipschitzness of $varphi$…

23 May

Ethical, trust, and skill barriers slow generative AI progress in EMEA

stp2y0 CommentsChat-GPT

76% of consumers in EMEA think AI will significantly impact the next five years, yet 47% question the value that AI will bring and 41% are worried about its applications. This is according to research from enterprise analytics AI firm Alteryx. Since the release of ChatGPT by OpenAI in November 2022, there has been significant buzz about the transformative potential of generative AI, with many considering it one of the most revolutionary technologies of our time. With a significant 79% of organisations reporting that generative AI contributes positively to business, it is evident that a gap needs to be addressed to demonstrate AI’s value to consumers both in…

23 May

Pocket-Sized AI Models Could Unlock a New Era of Computing

stp2y0 CommentsGenAIapple, artificial intelligence, deep learning, fast forward, google, machine learning, microsoft

When ChatGPT was released in November 2023, it could only be accessed through the cloud because the model behind it was downright enormous.Today I am running a similarly capable AI program on a Macbook Air, and it isn’t even warm. The shrinkage shows how rapidly researchers are refining AI models to make them leaner and more efficient. It also shows how going to ever larger scales isn’t the only way to make machines significantly smarter.The model now infusing my laptop with ChatGPT-like wit and wisdom is called Phi-3-mini. It’s part of a family of smaller AI models recently released by…

23 May

Most US TikTok Creators Don’t Think a Ban Will Happen

stp2y0 CommentsRAG modelsadvertising, algorithms, china, instagram, meta, privacy, social media, tiktok

A majority of US TikTok creators don’t believe the platform will be banned within a year, and most haven’t seen brands they work for shift their marketing budgets away from the app, according to a new survey of people who earn money from posting content on TikTok shared exclusively with WIRED.The findings suggest that TikTok’s influencer economy largely isn’t experiencing existential dread after Congress passed a law last month that put the future of the app’s US operations in jeopardy. The bill demands that TikTok separate from its Chinese parent company within a year or face a nationwide ban; TikTok…

23 May

Instruction Tuning OPT-125M

stp2y0 CommentsViral News

Large language models are pretrained on terabytes of language datasets. However, the pretraining dataset and strategy teach the model to generate the next token or word. In a real world sense, this is not much useful. Because in the end, we want to accomplish a task using the LLM, either through chat or instruction. We can do so through fine-tuning an LLM. Generally, we call this instruction tuning of the language model. To this end, in this article, we will use the OPT-125M model for instruction tuning. Figure 1. Output sample after instruction tuning OPT-125M on the Open Assistant Guanaco…

23 May

Data Moats in Generative AI

stp2y0 CommentsNews

The deep learning wave of the early 2010s led to a surge of data-hungry products. These products needed so much data that gathering it requires significant investment. So, the business community started honing the idea of data as a strategic asset and a business moat. As the Economist put it in a 2017 issue, “The world’s most valuable resource is no longer oil, but data.” This essay discusses data moats in today’s context of generative AI, which is driven by models that are exponentially more data-hungry. But first, what is a data moat? what is even an “AI product”?A data…

23 May

Accelerate Mixtral 8x7B pre-training with expert parallelism on Amazon SageMaker | Amazon Web Services

stp2y0 CommentsAI

Mixture of Experts (MoE) architectures for large language models (LLMs) have recently gained popularity due to their ability to increase model capacity and computational efficiency compared to fully dense models. By utilizing sparse expert subnetworks that process different subsets of tokens, MoE models can effectively increase the number of parameters while requiring less computation per token during training and inference. This enables more cost-effective training of larger models within fixed compute budgets compared to dense architectures. Despite their computational benefits, training and fine-tuning large MoE models efficiently presents some challenges. MoE models can struggle with load balancing if the tokens…

23 May

Study Finds That 52 Percent of ChatGPT Answers to Programming Questions Are Wrong

stp2y0 CommentsChat-GPT

Ah yes. And yet...Not So SmartIn recent years, computer programmers have flocked to chatbots like OpenAI's ChatGPT to help them code, dealing a blow to places like Stack Overflow, which had to lay off nearly 30 percent of its staff last year.The only problem? A team of researchers from Purdue University presented research this month at the Computer-Human Interaction conference that shows that 52 percent of programming answers generated by ChatGPT are incorrect.That's a staggeringly large proportion for a program that people are relying on to be accurate and precise, underlining what other end users like writers and teachers are experiencing:…