Рубрика: AI

UiPath, a leading enterprise automation software company, is developing foundational models, according to Daniel Dines, co-founder and chief innovation officer at UiPath. While speaking at the UiPath DevCon 2024, held in Bengaluru, Dines also pointed out that the company revealed its vision model in DevCon the previous year.

At the event, UiPath also announced a host of new features in its platform designed to enable developers to build, test, and accelerate implementation of automation.

The company announced UiPath Autopilot for Test Suite, now in public preview. Autopilot for Test Suite provides a collection of advanced AI capabilities designed to boost the productivity of testers throughout the entire testing lifecycle. Autopilot for Test Suite includes:

• AI-Powered Quality Checks: acts as an expert reviewer at a tester’s fingertips analysing requirements, from popular lifecycle tools such as Jira or Azure DevOps or documents such as a PDD, process design document.
• AI-Powered Test Design: generates manual tests with step-by-step instructions from developers’ requirements and supporting docs, such as workflows or design documents.
• AI-Powered Test Automation: converts any text, including manual test steps, into automated tests in UiPath Studio Desktop with little effort; and
• AI-Powered Test Insights: provides real-time, actionable insights into the test case portfolio, empowering rapid issue resolution and continuous improvement almost instantaneously.

“With the UiPath developer community in excess of 2 million, and developers and learners across 43 countries, the innovations we provide this community are absolutely essential in the AI and automation era. To build modern automations, it is not only about driving the UI. Developers also need API integrations, data, and AI solutions. We are making it possible for developers to access all these constructs in one integrated platform,” Dines said in a press release.

“We are infusing Generative AI across all our products. We believe when AI is combined with automation, magic happens. It’s like combining an intelligent brain with highly capable body to carry out powerful tasks,” said Munil Shah, chief technology officer, UiPath Automation Cloud.

The post UiPath is Building Foundational Models: CEO appeared first on Analytics India Magazine.

Image by Author

Exploratory Data Analysis (or EDA) stands as a core phase within the Data Analysis Process, emphasizing a thorough investigation into a dataset's inner details and characteristics.

Its primary aim is to uncover underlying patterns, grasp the dataset's structure, and identify any potential anomalies or relationships between variables.

By performing EDA, data professionals check the quality of the data. Therefore, it ensures that further analysis is based on accurate and insightful information, thereby reducing the likelihood of errors in subsequent stages.

So let’s try to understand together what are the basic steps to perform a good EDA for our next Data Science project.

The Significance of EDA

I am pretty sure you have already heard the phrase:

Garbage in, Garbage out

Input data quality is always the most important factor for any successful data project.

Unfortunately, most data is dirt at first. Through the process of Exploratory Data Analysis, a dataset that is nearly usable can be transformed into one that is fully usable.

It's important to clarify that it is not a magic solution for purifying any dataset. Nonetheless, numerous EDA strategies are effective at addressing several typical issues encountered within datasets.

So… let’s learn the most basic steps according to Ayodele Oluleye in his book Exploratory Data Analysis with Python Cookbook.

Step 1: Data Collection

The initial step in any data project is having the data itself. This first step is where data is gathered from various sources for subsequent analysis.

2. Summary Statistics

In data analysis, handling tabular data is quite common. During the analysis of such data, it's often necessary to gain rapid insights into the data's patterns and distribution.

These initial insights serve as a base for further exploration and in-depth analysis and are known as summary statistics.

They offer a concise overview of the dataset's distribution and patterns, encapsulated through metrics such as mean, median, mode, variance, standard deviation, range, percentiles, and quartiles.

Image by Author

3. Preparing Data for EDA

Before starting our exploration, data usually needs to be prepared for further analysis. Data preparation involves transforming, aggregating, or cleaning data using Python's pandas library to suit the needs of your analysis.

This step is tailored to the data’s structure and can include grouping, appending, merging, sorting, categorizing, and dealing with duplicates.

In Python, accomplishing this task is facilitated by the pandas library through its various modules.

The preparation process for tabular data doesn't adhere to a universal method; instead, it's shaped by the specific characteristics of our data, including its rows, columns, data types, and the values it contains.

4. Visualizing Data

Visualization is a core component of EDA, making complex relationships and trends within the dataset easily comprehensible.

Using the right charts can help us identify trends within a big dataset and find hidden patterns or outliers. Python offers different libraries for data visualization, including Matplotlib or Seaborn among others.

Image by Author

5. Performing Variable Analysis:

Variable analysis can be either univariate, bivariate, or multivariate. Each of them provides insights into the distribution and correlations between the dataset's variables. Techniques vary depending on the number of variables analyzed:

Univariate

The main focus in univariate analysis is on examining each variable within our dataset on its own. During this analysis, we can uncover insights such as the median, mode, maximum, range, and outliers.

This type of analysis is applicable to both categorical and numerical variables.

Bivariate

Bivariate analysis aims to reveal insights between two chosen variables and focuses on understanding the distribution and relationship between these two variables.

As we analyze two variables at the same time, this type of analysis can be trickier. It can encompass three different pairs of variables: numerical-numerical, numerical-categorical, and categorical-categorical.

Multivariate

A frequent challenge with large datasets is the simultaneous analysis of multiple variables. Even though univariate and bivariate analysis methods offer valuable insights, this is usually not enough for analyzing datasets containing multiple variables (usually more than five).

This issue of managing high-dimensional data, usually referred to as the curse of dimensionality, is well-documented. Having a large number of variables can be advantageous as it allows the extraction of more insights. At the same time, this advantage can be against us due to the limited number of techniques available for analyzing or visualizing multiple variables concurrently.

6. Analyzing Time Series Data

This step focuses on the examination of data points collected over regular time intervals. Time series data applies to data that change over time. This basically means our dataset is composed of a group of data points that are recorded over regular time intervals.

When we analyze time series data, we can typically uncover patterns or trends that repeat over time and present a temporal seasonality. Key components of time series data include trends, seasonal variations, cyclical variations, and irregular variations or noise.

7. Dealing with Outliers and Missing Values

Outliers and missing values can skew analysis results if not properly addressed. This is why we should always consider a single phase to deal with them.

Identifying, removing, or replacing these data points is crucial for maintaining the integrity of the dataset's analysis. Therefore, it is highly important to address them before start analyzing our data.

• Outliers are data points that present a significant deviation from the rest. They usually present unusually high or low values.
• Missing values are the absence of data points corresponding to a specific variable or observation.

A critical initial step in dealing with missing values and outliers is to understand why they are present in the dataset. This understanding often guides the selection of the most suitable method for addressing them. Additional factors to consider are the characteristics of the data and the specific analysis that will be conducted.

Conclusions

EDA not only enhances the dataset's clarity but also enables data professionals to navigate the curse of dimensionality by providing strategies for managing datasets with numerous variables.

Through these meticulous steps, EDA with Python equips analysts with the tools necessary to extract meaningful insights from data, laying a solid foundation for all subsequent data analysis endeavors.

Josep Ferrer is an analytics engineer from Barcelona. He graduated in physics engineering and is currently working in the Data Science field applied to human mobility. He is a part-time content creator focused on data science and technology. You can contact him on LinkedIn, Twitter or Medium.

More On This Topic

• Exploratory Data Analysis Techniques for Unstructured Data
• A Data Scientist’s Essential Guide to Exploratory Data Analysis
• A Lightning Fast Look at Single Line Exploratory Data Analysis
• How Visualization is Transforming Exploratory Data Analysis
• Mastering the Data Universe: Key Steps to a Thriving Data Science Career
• KDnuggets™ News 22:n05, Feb 2: 7 Steps to Mastering Machine…

Soon after DALL-E gave rise to text-to-image AI, companies took a step ahead and started creating text-to-video models. Over two years, the landscape has evolved from being noisy to producing hyper-realistic results using text prompts.

While the results may still be imperfect, several models today display a high degree of controllability and the ability to generate footage in various artistic styles.

Here are six latest text-to-video AI models you could try out.

Sora

ChatGPT creator OpenAI just showcased Sora, their new text-to-video model. Everyone’s excited since the model has “a deep understanding of language” and can generate “compelling characters that express vibrant emotions”. People on social media are flipping out over how realistic the videos look, calling it a total game-changer.

But, before releasing it to the public, the AI startup is taking measures to be careful about safety. They also admit that Sora has some hiccups, like struggling with keeping things smooth and telling left from right. [Sam Altman Brings CRED Founder Kunal Shah’s Wild Imagination to Life with Sora]

Click here to know more.

Lumiere

Google’s got this video generation AI called Lumiere, powered by a new diffusion model known as Space-Time-U-Net, or STUNet for short. According to Ars Technica, Lumiere doesn’t mess around with stitching together still frames; instead, it figures out where things are in a video (that’s the space part) and tracks how they move and change at the same time (that’s the time part).

It’s like one smooth process, no need for puzzle pieces.

Lumiere has yet to be ready for common folks to recreate with. But it hints at Google’s knack for crafting an AI video powerhouse that might outshine the generally available models like Runway and Pika. Google has made a tech leap in AI video games within two years.

Click here to know more.

VideoPoet

VideoPoet, is a large language model schooled on a colossal dataset of videos, images, audio, and text. This model can pull off various video generation tasks, from turning text or images into videos to jazzing up videos with style, video inpainting and outpainting, and video-to-audio.

The model is built on a straightforward idea: convert any autoregressive language model into a video-generating system. Autoregressive language models can crank out text and code like nobody’s business. But they hit a roadblock when it comes to video.

To tackle that, VideoPoet rolls with multiple tokenisers that can turn video, image, and audio clips into a language it understands.

Click here to know more.

Emu Video

Meta’s AI model involves two steps. First, it makes a picture from text. Then, it uses that text and image to create a top-notch video. The researchers achieved this by optimising noise schedules for diffusion and multi-stage training.

Human evaluators claimed they preferred it 81% more than Google’s Imagen Video, 90% picked it over NVIDIA’s PYOCO, and 96% said it was better than Meta’s own Make-A-Video. Not just that, it’s even beating commercial options like RunwayML’s Gen2 and Pika Labs.

Notably, their factorising approach is well-suited for animating images based on user text prompts, surpassing prior works by 96%.

Click here to know more.

Phenaki

The team behind Phenaki Video used Mask GIT to produce text-guided videos in PyTorch. The model can generate videos guided by text and go up to 2-min long.

The paper suggested that instead of just trusting the predicted probabilities, they’re suggesting a tweak – bringing in an extra critic to decide what to mask during sampling iteratively. This helps determine what parts to focus on during the video-making process. It’s like having a second opinion.

The model is versatile and open for researchers to train on text-to-image and text-to-video. They can start with images and then fine-tune on video for unconditional training.

Click here to know more.

CogVideo

A group of researchers from the University of Tsinghua in Beijing developed CogVideo, a large-scale pretrained text-to-video generative model. They built the model using a pre-trained text-to-image model called CogView2 to exploit the knowledge it learned from pre-training.

Now, this computer artist named Glenn Marshall tried it out. He was so impressed initially that he said directors might lose their jobs to this thing. The short film he made with CogVideo, ‘The Crow’, performed well and even got a shot at the BAFTA Awards.

Click here to know more.

The post 6 Text-to-Video Generative AI Models appeared first on Analytics India Magazine.

Just when we thought that Google made OpenAI and Microsoft dance with the Gemini 1.5 release, OpenAI chief Sam Altman stole the spotlight with Sora – a super cool, text-to-video tool that generates life-like, hyper-realistic video footage, something the world has never seen before.

“The difference that is getting more and more obvious between OpenAI and Google every day is one company talks, and the other company shows,” wrote Nikunj Kothari, a partner at Khosla Ventures.

Partially, Google did make OpenAI dance when Altman personally sought prompts from users on X to create a buzz around Sora. “We’d like to show you what Sora can do, please reply with captions for videos you’d like to see and we’ll start making some!” wrote Altman, asking users to not hold back on the details or difficulty. The rest was history.

“Have to step out, more videos coming in about 45 mins,” cried Altman, as the wave of requests continued to overflow. And upon his return, CRED’s Kunal Shah’s was one of the cherry-picked requests that came to life.

Altman didn’t stop at that. He even gave much-deserved credit to the team behind Sora, saying, “OpenAI is the most talented and nicest group of people I have ever seen in one place.” Further, he said that they were working on the toughest, most interesting, and most important problems with all the resources in place, focused on building AGI. “You should perhaps consider joining us,” he added.

The team behind Sora is led by Tim Brooks, a research scientist at OpenAI, William Peebles, also a research scientist at OpenAI, and Aditya Ramesh, the creator of DALL·E and the head of Videogen.

All this feels like a stop-gap mechanism by OpenAI as it prepares to release GPT-5 in the coming weeks. “It will be more intelligent, multimodal, and faster,” said Altman at the World Government Summit in Dubai. In a recent episode of Unconfuse Me with Bill Gates as well, he highlighted the importance of multimodality as the key aspect of GPT-5 that enables it to process video input and generate new videos.

Sora might be just what GPT-5 needs.

A ChatGPT Moment in Video Generation?

OpenAI’s all new super-cool, text-to-video tool can create videos of up to 60 seconds featuring highly detailed scenes, complex camera motion, and multiple characters with vibrant emotions. Many are also calling this a ChatGPT moment in video generation.

Sora is not just a video generation model for OpenAI, it’s a stepping stone to AGI. OpenAI is teaching Sora to understand and simulate the physical world in motion, with the goal of training models that help people solve problems requiring real-world interaction.

How Sora Works

Sora is just like an LLM using Transformer architecture. While LLMs utilise text tokens, Sora employs visual patches. Patches serve as a highly scalable and effective representation for training generative models on various types of videos and images. Similar to how LLMs predict words, Sora is trained to predict the original ‘clean’ patches, given input noisy patches.

Sora has multiple features like animating DALL·E images, extending generated videos, video-to-video editing and connecting videos. However, apart from video generation, the possibilities with Sora are endless. It can simulate some aspects of people, animals and environments from the physical world.

Moreover, it can also simulate digital worlds, which includes popular games like Minecraft. It can simultaneously control the player in Minecraft with a basic policy while also rendering the world and its dynamics in high fidelity.

The introduction of Sora comes on the heels of Google’s recent launch of Lumiere, a text-to-video diffusion model crafted to synthesise videos, producing realistic, diverse, and coherent motion. Sora translates to ‘sky’ in Japanese. Sora is surely a sky full of stars instead of a mere constellation. It also a wake-up call for other text-to-video generators like RunwayML, Pika, and Stable Video.

The post OpenAI Steals the Spotlight with Sora ✨ appeared first on Analytics India Magazine.

According to a recent survey report, generative AI has improved the productivity of software engineers by 70%. Developers have witnessed the reduction in task completion time for existing code updates, especially impactful in leveraging existing codebase functions and reducing development cycle time

The study, conducted by New York-based management consulting firm Zinnov and data engineering and lifecycle company Ness Digital Engineering is based on based on data collected using Ness’s Matrix platform from over 100 software engineers, across various use cases and development settings, offering an analysis of their practical experiences in live engineering scenarios. This was over 14 sprints where the team evaluated sprints before and after generative AI deployment.

“We identified multiple gains from this. We saw productivity enhancements in line with the complexity of tasks and based on the experience of the engineer. The generative AI integration into engineering workflow tasks has led to a significant 38% decrease in the time required to complete tasks, with a notable 48% reduction in task completion time observed among senior engineers,” Sidhant Rastogi, managing partner of Zinnov, told AIM, emphasising that as the engineers get more conversant with generative AI, efficiency will increase further.

Even though generative has shown promising results in reducing time and bossing production, concerns arise about the potential negative impact on junior developers’ learning opportunities by limiting exposure to these tasks.

“Generative AI accelerates coding updates, allowing engineers to focus on business context and logic. This shift elevates the workload’s nature and empowers senior engineers to guide junior developers, expediting learning and encouraging innovative contributions to projects,” commented Rastogi.

Key Findings

Besides improving productivity, the report shows that senior engineers experienced a 48% reduction in task completion time, enabling better planning and support for junior engineers.

Generative AI contributed to approximately a 10% reduction in high code complexity tasks, helping engineers in navigating complex coding scenarios efficiently. Additionally, there was a 70% improvement in engagement due to simplified tasks and a more collaborative work environment, leading to a positive professional experience.

“Generative AI significantly impacts routine sustenance tasks in software development, such as legacy code maintenance and updates. Automated tasks, like generating test cases from existing code, ensure continuous alignment with evolving software architecture, reducing regression bugs and ensuring quality assurance. It also facilitates global collaboration, breaks down knowledge barriers, and creates a cohesive product development environment, enhancing decision-making and employee engagement,” Vikas Basra, global head, IE practice, Ness Digital Engineering, told AIM about the report.

The post Generative AI Boosts Work Productivity by 70% appeared first on Analytics India Magazine.

The global insurance industry is undergoing a profound transformation powered by a cutting-edge technology: generative AI. Gone are the days of experimental tests; Fractal Analytics, a leader in the field, reveals how this AI revolution is scaling up, reshaping internal operations, and paving the way for a brighter future.

“Generative AI is no longer the future, it’s the present,” declares Amarava Roy, Principal Consultant at Fractal. This shift isn’t just hype; it’s driven by the immense potential AI holds to streamline processes and unlock deeper insights. Insurance companies, traditionally seen as tech adoption laggards given the risk management nature of the industry and the regulatory side of it, are embracing this change with open arms.

“At the heart of this revolution lies internal efficiency,” Rashid Khan, Engagement Manager at Fractal, explains how document-heavy workflows are being streamlined with AI-powered knowledge management. Underwriters get instant access to critical guidelines, and claims adjusters leverage past case insights, empowering them to make faster, more informed decisions.

But the magic extends beyond simple automation. Code generation AI, capable of boosting programmer productivity by 80%, thus empowering coding professionals to advance their skills in an industry where regulatory changes demand frequent updates to pricing models for example. Generative AI aids in code maintenance and updates, ensuring compliance and accuracy. And self-serve analytics solutions built on generative AI are putting powerful insights at the fingertips of everyone, not just data scientists.

Khan says that one of the standout areas where generative AI is making a significant impact is in streamlining internal operations. “Traditionally document-heavy and reliant on manual processes, the insurance industry is leveraging generative AI to enhance productivity and efficiency,” he added.

In the realm of knowledge management, underwriters now have rapid access to guidelines and policy documents, facilitating quicker decision-making during the critical underwriting process. Similarly, claims adjusters benefit from knowledge bases built on adjuster notes, empowering less experienced professionals by providing insights from past cases.

“Insurance industry is very excited about Generative AI”

The roles of underwriters and claims adjusters is constantly and rapidly evolving with generative AI. Roy and Khan highlighted the importance of setting clear expectations and reassuring users that generative AI is meant to augment human capabilities, not replace them. Addressing fears and curiosities among employees, especially in a technology-enabled ecosystem, is crucial for successful adoption.

One of the things that Khan pointed out was that Generation Z is increasingly not looking for jobs as underwriters and in the insurance industry as a whole. “Generative AI can attract talent from Generation Z, offering an edge to early adopters. However, it is essential to clarify that while it enhances efficiency, it does not replace the core functions and expertise of underwriters and claims adjusters,” Khan added.

“Generative AI is not here to replace humans, it’s here to empower them,” emphasises Khan. Fractal’s “Crux,” an in-house AI copilot, exemplifies this philosophy. Crux doesn’t just answer questions; it explains the reasoning behind its answers, fostering understanding and collaboration between humans and AI. Positioned as a bridge between generative AI and business intelligence, Crux provides a conversational interface for users to interact with data. This tool not only provides answers but also explains the reasoning behind them, enhancing the user’s understanding of the insights generated.

Yet, with such power comes responsibility! Roy acknowledges the ethical concerns surrounding “hallucination,” where AI models generate inaccurate information. Fractal tackles this head-on with RAG, an approach that verifies information against trusted sources, ensuring data integrity and user trust. This focus on responsible AI isn’t just lip service. Fractal embraces a robust multi-cloud approach, ensuring flexibility and data privacy. They champion open-source language models, allowing for customization and control over algorithms. And their commitment to governance guarantees ethical implementation, mitigating risks, and building trust.

This journey doesn’t end with internal optimisations. As Roy points out, “the focus is on driving customer value ethically and securely.” From personalised risk assessments to faster claims processing, the future of insurance promises a seamless customer experience powered by responsible AI.

All the tech

Fractal has a cloud agnostic approach and does not depend on any specific cloud provider for the same. Khan said that when ChatGPT was released, it was obvious for them to rely on Microsoft Azure for OpenAI services. But with the increasing number of models and companies, they have experimented with AWS and Google Cloud as well.

Furthermore, Khan said organisations are adopting a multi-cloud strategy, choosing cloud providers based on existing infrastructure, preferences, and strategic partnerships. The use of open-source language models adds flexibility and transparency to Generative AI implementations, allowing for customization and control over algorithms.

He also focused on the fact that open-source language models have been crucial for ensuring that their goals in the insurance industry work perfectly for ensuring data privacy and flexibility with the cloud providers.

Data privacy and security are also paramount concerns, along with potential biases in the data used for training generative AI models. Adherence to ethical standards is crucial for building trust with customers and upholding the industry’s reputation.

To solve the hallucination problem, Fractal has been implementing RAG for fetching information from specific sources, instead of relying on it. “We have been using RAG and it has been working pretty well for us,” added Khan. For this, Khan said that there is no specific solution provider they focus on and decide the provider as per specific use cases. The company is also working to figure out how hallucinations can be reduced even further by employing agents in their frameworks.

Looking Ahead: The Future of Generative AI in Insurance

As the insurance industry continues its journey with Generative AI, the focus remains on security, ethics, and internal productivity. The technology is not just a novel addition but a transformative force shaping day-to-day operations. The transition from experimentation to scaling indicates a future where AI seamlessly integrates into the insurance workflows, contributing to enhanced efficiency, informed decision-making, and a more empowered workforce.

“It is important to emphasise that across industries, organisations are experimenting with generative AI, each striving to drive customer value ethically and securely. The focus on responsible AI, keeping humans at the centre, and setting clear expectations emerged as critical considerations for successful adoption,” Roy added.

This is not just a tech revolution; it’s a cultural shift. Fractal’s roadmap includes building a robust “Generative AI Center of Excellence” to fuel experimentation and rapid scaling. By empowering Insurance workforces with human-centric AI solutions, they’re paving the way for a future where insurance thrives on efficiency, informed decisions, and a deeply human touch.

Governance is also identified as a crucial layer to ensure responsible and ethical AI implementation. The need for a comprehensive governance structure involving technical, legal, and leadership perspectives was emphasised to address technical, regulatory, reputational, and revenue risks.

In conclusion, Fractal provides a comprehensive view of the transformative journey of Generative AI in the insurance sector. From internal empowerment and ethical considerations to technology choices and the roadmap for adoption, the insights shared highlight the industry’s commitment to leveraging AI responsibly.

As generative AI continues to evolve, the insurance sector stands on the cusp of a technological revolution that promises enhanced efficiency, informed decision-making, and a workforce empowered by human-centric AI solutions.

The post How Fractal is Leveraging Generative AI for Insurance appeared first on Analytics India Magazine.