10th Indian Delegation to Dubai, Gitex & Expand North Star – World’s Largest Startup Investor Connect
Artificial Intelligence

Microsoft brings new design-focused features to Copilot


Copilot, Microsoft’s family of AI-powered chatbots and assistants, is getting a few new upgrades timed with a flashy Superbowl LVIII ad campaign.

In a post on the official Microsoft blog, Yusuf Mehdi, Microsoft’s chief marketing officer, outlined what users can expect.

“Today marks exactly one year since our entry into AI-powered experiences for people with Bing Chat,” he wrote. “In that year, we’ve learned so many new things, and seen the use of our Copilot experiences explode with over 5 billion chats and 5 billion images created to date … Now, with Copilot as our singular experience, for people looking to get more out of AI creation, we’re introducing further … capabilities.”

The Copilot experience on the web, Android and iOS now features an improved AI model, Deucalion, along with a more “streamlined look and feel,” Mehdi said — with a cleaner style for answers and a carousel of suggested prompts to feed Copilot (e.g. “How would you explain AI to a sixth grader?”).

Designer in Copilot, meanwhile — a tool that taps GenAI models like OpenAI’s DALL-E 3 to turn prompts into images — has new editing capabilities.

Microsoft Copilot

The upgraded Copilot experience on the web.

All English-speaking Copilot users in the U.S., U.K., Australia, India and New Zealand can now edit images in-line in the flow of a chat, for example colorizing an object, blurring an image background or changing the style of the image (to pixel art, say). And subscribers to Copilot Pro, Microsoft’s $20-per-month premium Copilot plan, can resize and regenerate images between “square” (i.e. portrait) and landscape.

Coming soon to Copilot is Designer GPT, Mehdi said, which will offer a more “immersive, dedicated canvas” inside Copilot where users can “visualize their ideas.”

Designer caused quite a stir earlier this year when malicious users, mostly from the image board 4chan, used the tool to create pornographic deepfakes of Taylor Swift — and spread them across X (formerly Twitter). Designer had guardrails designed to prevent inappropriate prompts, Microsoft claimed — but users found loopholes like misspelling names and describing images that didn’t explicitly use sexual terms but generated the same result. 

Last month, Microsoft said that it addressed the Designer loopholes by making it impossible to generate celebrity images. But, as with all GenAI tools, it’s likely to be a never-ending cat and mouse game between bad actors and vendors.

“Microsoft’s advancements in AI align with our company mission to empower every person and organization on the planet to achieve more,” Mehdi continued. “With Copilot, we’re democratizing our breakthroughs in AI to help make the promise of AI real for everyone.”

Mehdi didn’t address the performance issues with Copilot Pro — a common complaint among early subscribers.

Copilot Pro is supposed to come with priority access to the underlying OpenAI models powering Copilot even during peak times, but users report dealing with exceptionally long generation times and other, potentially related bugs. Windows Central speculates that the root of the problem is insufficient server capacity, but lacking official comment, it’s impossible to know for certain.



Source link

by Team SNFYI

Facebook is testing a new feature that invites some users—mainly in the US and Canada—to let Meta AI access parts of their phone’s camera roll. This opt-in “cloud processing” option uploads recent photos and videos to Meta’s servers so the AI can offer personalized suggestions, such as creating collages, highlight reels, or themed memories like birthdays and graduations. It can also generate AI-based edits or restyles of those images. Meta says this is optional and assures users that the uploaded media won’t be used for advertising. However, to enable this, people must agree to let Meta analyze faces, objects, and metadata like time and location. Currently, the company claims these photos won’t be used to train its AI models—but they haven’t completely ruled that out for the future. Typically, only the last 30 days of photos get uploaded, though special or older images might stay on Meta’s servers longer for specific features. Users have the option to disable the feature anytime, which prompts Meta to delete the stored media after 30 days. Privacy experts are concerned that this expands Meta’s reach into private, unpublished images and could eventually feed future AI training. Unlike Google Photos, which explicitly states that user photos won’t train its AI, Meta hasn’t made that commitment yet. For now, this is still a test run for a limited group of people, but it highlights the tension between AI-powered personalization and the need to protect personal data.

by Team SNFYI

News Update Bymridul     |    March 14, 2024 Meesho, an online shopping platform based in Bengaluru, has announced its largest Employee Stock Ownership Plan (ESOP) buyback pool to date, totaling Rs 200 crore. This buyback initiative extends to both current and former employees, providing wealth creation opportunities for approximately 1,700 individuals. Ashish Kumar Singh, Meesho’s Chief Human Resources Officer, emphasized the company’s commitment to rewarding its teams, stating, “At Meesho, our employees are the driving force behind our success.” Singh further highlighted the company’s dedication to providing opportunities for wealth creation despite prevailing macroeconomic conditions. This marks the fourth wealth generation opportunity at Meesho, with the size of the buyback program increasing each year. In previous years, Meesho conducted buybacks worth over Rs 8.2 crore in February 2020, Rs 41.4 crore in November 2020, and Rs 45.5 crore in October 2021. Meesho’s profitability journey began in July 2023, making it the first horizontal Indian e-commerce company to achieve profitability. Despite turning profitable, Meesho continues to maintain positive cash flow and focuses on enhancing efficiencies across various cost items. The company’s revenue from operations for FY 2022-23 witnessed a remarkable growth of 77% over the previous year, amounting to Rs 5,735 crore. This growth can be attributed to Meesho’s leadership position as the most downloaded shopping app in India in both 2022 and 2023, increased transaction frequency among existing customers, and a diversified category mix. Additionally, Meesho’s focus on improving monetization through value-added seller services contributed to its revenue growth. Meesho also disclosed its audited performance for the first half of FY 2023-24, reporting consolidated revenues from operations of Rs 3,521 crore, marking a 37% year-over-year increase. The company achieved profitability in Q2 FY24, with a significant reduction in losses compared to the previous year. Furthermore, Meesho recorded impressive app download numbers, reaching 145 million downloads in India in 2023 and surpassing 500 million downloads in H1 FY 2023-24. Follow Startup Story Source link

by Team SNFYI

You might’ve heard of Grok, X’s answer to OpenAI’s ChatGPT. It’s a chatbot, and, in that sense, behaves as as you’d expect — answering questions about current events, pop culture and so on. But unlike other chatbots, Grok has “a bit of wit,” as X owner Elon Musk puts it, and “a rebellious streak.” Long story short, Grok is willing to speak to topics that are usually off limits to other chatbots, like polarizing political theories and conspiracies. And it’ll use less-than-polite language while doing so — for example, responding to the question “When is it appropriate to listen to Christmas music?” with “Whenever the hell you want.” But Grok’s ostensible biggest selling point is its ability to access real-time X data — an ability no other chatbots have, thanks to X’s decision to gatekeep that data. Ask it “What’s happening in AI today?” and Grok will piece together a response from very recent headlines, while ChatGPT, by contrast, will provide only vague answers that reflect the limits of its training data (and filters on its web access). Earlier this week, Musk pledged that he would open source Grok, without revealing precisely what that meant. So, you’re probably wondering: How does Grok work? What can it do? And how can I access it? You’ve come to the right place. We’ve put together this handy guide to help explain all things Grok. We’ll keep it up to date as Grok changes and evolves. How does Grok work? Grok is the invention of xAI, Elon Musk’s AI startup — a startup reportedly in the process of raising billions in venture capital. (Developing AI’s expensive.) Underpinning Grok is a generative AI model called Grok-1, developed over the course of months on a cluster of “tens of thousands” of GPUs (according to an xAI blog post). To train it, xAI sourced data both from the web (dated up to Q3 2023) and feedback from human assistants that xAI refers to as “AI tutors.” On popular benchmarks, Grok-1 is about as capable as Meta’s open source Llama 2 chatbot model and surpasses OpenAI’s GPT-3.5, xAI claims. Image Credits: xAI Human-guided feedback, or reinforcement learning from human feedback (RLHF), is the way most AI-powered chatbots are fine-tuned these days. RLHF involves training a generative model, then gathering additional information to train a “reward” model and fine-tuning the generative model with the reward model via reinforcement learning. RLHF is quite good at “teaching” models to follow instructions — but not perfect. Like other models, Grok is prone to hallucinating, sometimes offering misinformation and false timelines when asked about news. And these can be severe — like wrongly claiming that the Israel–Palestine conflict reached a ceasefire when it hadn’t. For questions that stretch beyond its knowledge base, Grok leverages “real-time access” to info on X (and from Tesla, according to Bloomberg). And, similar to ChatGPT, the model has internet browsing capabilities, enabling it to search the web for up-to-date information about topics. Musk has promised improvements with the …