Blogs – Udiwonder Blog

ChatGPT comes to Microsoft’s enterprise-focused, Azure-powered managed service

ChatGPT, OpenAI’s viral, AI-powered chatbot tech, is now available in a more enterprise-friendly package.

Microsoft today announced that ChatGPT is generally available through the Azure OpenAI Service, the company’s fully managed, corporate-focused offering designed to give businesses access to OpenAI’s technologies with added governance and compliance features. Customers, who must already be “Microsoft managed customers and partners,” can apply here for special access.

ChatGPT joins the range of other OpenAI-developed systems already served through the Azure OpenAI Service, including the text-generating GPT-3.5, the code-generating Codex and the image-generating DALL-E 2. Microsoft has a close working relationship with OpenAI, having invested billions into the startup and inking an exclusive agreement to commercialize OpenAI’s AI research.

It’s paying dividends, apparently. Microsoft said today that over 1,000 brands are enrolled in the Azure OpenAI Service.

The Azure OpenAI Service flavor of ChatGPT is priced at $0.002 per 1,000 tokens, or about 750 words, with billing for ChatGPT usage to begin March 13. (ChatGPT consumes raw text represented as a series of tokens; the word “fantastic” would be split into the tokens “fan,” “tas” and “tic,” for example.) That’s the same price as the developer-focused ChatGPT API, which launched on March 1.

Eric Boyd, the CVP of AI platform at Microsoft, noted in a blog post that Azure OpenAI Service customers can configure ChatGPT’s responses to “align with their organization.” That’s an important callout given users have been able to prompt ChatGPT to answer questions in racist and sexist ways, invent facts without disclosing that it’s doing so and bypass certain high-level safeguards.

“Developers can integrate custom [ChatGPT-powered] experiences directly into their own applications, including enhancing existing bots to handle unexpected questions, recapping call center conversations to enable faster customer support resolutions, creating new ad copy with personalized offers, automating claims processing and more,” Boyd wrote.

Despite ChatGPT’s more problematic aspects, uptake has been swift. OpenAI has yet to provide API usage numbers, but as of December, the consumer-facing ChatGPT app had an estimated over 100 million monthly active users. Brands including Snap and Quizlet have integrated — or plan to soon integrate — it into their respective platforms. For instance, Instacart says it’ll use ChatGPT to create Ask Instacart, a tool that’ll allow Instacart customers to get “shoppable” answers informed by product data from the company’s retail partners. And Office Depot is building a ChatGPT-powered chatbot to support several of its internal business units, specifically in HR.

There’s been pushback from some employers, though. Wells Fargo is among several firms that’ve reportedly imposed usage limits on ChatGPT out of productivity and compliance concerns.

Source: TechCrunch

Meta is working on a decentralized social app

If there is a social media phenomenon getting some kind of popularity, Meta will try and jump in. We have seen the company copy different kinds of formats ranging from Stories to short videos after seeing the success of other platforms. Now, the Mark Zuckerberg-led company is working on a decentralized text-based app.

Meta confirmed this development in a statement but didn’t give out details about when it plans to release the app.

“We’re exploring a standalone decentralized social network for sharing text updates. We believe there’s an opportunity for a separate space where creators and public figures can share timely updates about their interests,” a Meta spokesperson said.

This new decentralized app, codenamed P92, is still under development — as first reported by MoneyControl. According to the documents seen by the publication, the app will let users log-in through their Instagram credentials. This could irk people who might not want to share that data with another Meta app.

A report by Platformer said that the project will be overseen by Instagram head Adam Mosseri. The company is already involving the legal department to sniff out early privacy concerns before the app is public, the report added.

Meta’s move is seen as its attempt to build a Twitter alternative or a Mastodon competitor. The latter gained popularity after Elon Musk took over Twitter. The decentralized network is part of the Fediverse — a network of decentralized servers — that supports the ActivityPub protocol. Meta’s new app also plans to support ActivityPub making it easier to connect with other instances like Mastodon, according to MoneyControl.

There are plenty of other tools that have implemented (or planning to implement) ActivityPub support including Tumblr, Flipboard, and Flickr.

However, former Twitter engineer Blaine Cook told TechCrunch last year that the existence of competing protocols is a good thing.

“I think the diversity of protocol is important, as is the diversity of the applications built on top of the protocols. That said, I strongly believe that interoperability between ActivityPub and Bluesky won’t be difficult. The only thing preventing, for example, interoperability between Twitter and Facebook’s timeline has been protectionist policies by those companies,” he noted.

It’s important to remember that Meta has tried making new apps and experiences that haven’t always taken off. In the past few years, it has killed experiments like the anonymous teen app tbh, Cameo-like app Super, Nextdoor clone Neighborhoods, couples app Tuned, student-focused social network Campus, video speed dating service Sparked, and TikTok clone Lasso just to name a few. So it won’t be surprising if the new decentralized experience shuts down in a couple of years after the launch.

Source: TechCrunch

Microsoft’s Bing hits 100 million active users thanks to AI chat, Edge browser

Microsoft’s Bing has never been in any danger of overtaking Google as the Internet’s most popular search engine. But the headline-grabbing AI-powered features from the “new Bing” preview that the company launched last month seem to be helping—Microsoft said today that Bing had passed the 100 million daily active users mark.

“We are fully aware we remain a small, low, single-digit share player,” writes Microsoft’s Yusuf Mehdi, driving home just how small Microsoft’s share of the search market is compared to Google’s. “That said, it feels good to be at the dance!”

Google doesn’t provide daily active user numbers for its search engine, but StatCounter data suggests that its market share typically hovers just under 90 percent in the US, compared to 6 or 7 percent for Bing.

Microsoft says that there are “millions of active users” trying out the AI-powered Bing preview and that roughly a third of those are people who weren’t using the service before. The company also credits the slow-but-steady growth of Microsoft Edge, which uses Bing as its default search engine and will repeatedly prompt you to switch back to Bing as your default search engine if you switch to something else.

Microsoft launched the Bing chatbot a month ago in a “limited preview,” using a large language model (LLM) from OpenAI. The company has made several changes to the bot’s behavior in response to its sometimes weird and threatening conversations. One change limited the number of responses the chatbot could give in a single instance, since it was easier for the bot to go off the rails during extended sessions; more recently, Microsoft introduced some “personalities” for the bot to make answers either more straightforward or more entertaining.

Since last month, Microsoft has added “new Bing” features to its Edge browser, Skype, the Windows 11 taskbar, and some of its developer tools. The company is also testing “multimodal AI” technology that can process multiple forms of input, including images, text, audio, and video.

Source: Arstechnica

Discord Introducing ‘AI’ Stuff Nobody Asked For, Or Needs

In a blog post published earlier today, Discord—a communications platform, and nothing more—announced plans to introduce what they call “AI” (but which should more accurately be called “machine learning”) to the service’s moderation, support, and chat.

The post, called DISCORD IS YOUR PLACE FOR AI WITH FRIENDS, is full of positive imagery and promises about how “AI” will make everything easier for anyone using Discord, whether they’re chatting with friends or trying to moderate a group.

Why is Discord becoming the home for AI? Simple: on Discord you can enjoy AI with friends. Rather than just going solo with an app, you and your friends can see what sorts of exciting, wild and sometimes silly results come from prompts like “robo-hamster caught in cardboard box, renaissance painting.”

No thank you! AI-generated imagery is, as we’ve discussed frequently, built on the bones of uncredited and uncompensated human art, and plans to implement a sort of ChatGPT functionality—which is rife with factual severe errors—to Discord conversations seems premature at best.

The contrast between Discord’s optimistic presentation and reaction to the news itself couldn’t be starker. Replies under the announcement Tweet are almost universally negative, with users quickly realizing that handing the keys to so many features over to machine learning is a threat to the privacy, accuracy, and legitimacy of communications on the platform.

“Will there be an opt-out button for server owners who don’t wish to have their server become training material for machine learning?”, asks one user, while most others are simply as many variations of a “nah” image meme as you can find on the internet in 2023.

“Ah yes, ‘sharing AI experiences,’ precisely what I’m on Discord for, my mistake for thinking I was there to spend time with my friends, network, and meet new people”, says another.

They even manage to fuck up the one good announcement among it all, with news of a “shared whiteboard” feature—something people have wanted forever—spoiled by the fact it will come saddled with “an AI-powered text-to-image generator you can iterate and experiment with together”.

Companies, I promise you, you don’t need to do this. I know you are compelled to through the irresistible forces of capitalist inertia, the need to make everything grow all the time, but like, this is a chat program. It doesn’t need any of this crap in it. We’re literally only using Discord for one thing: to talk to people.

Source: Kotaku