Every time Microsoft launches a major AI feature this year, I can’t help but become more skeptical of the company’s new direction. Microsoft, a notoriously conservative and slow-moving giant, reshaped its products around generative AI soon after most people learned of its existence. The last time it made such a dramatic shift, we launched Windows 8, a failed attempt to make its flagship operating system tablet- and touchscreen-friendly.
Now that the company is bringing artificial intelligence into the core of Windows, I can’t help but wonder: Is Microsoft’s involvement in artificial intelligence really meant to make its products better? Or does it just want to claim itself as an AI innovator and hope the technology lives up to the hype? At this point, it’s really hard to say.
As the Zune, WebTV, and Windows Phone have shown, Microsoft’s timing isn’t great. Its products often either launched too early to make an impact (like the sluggish WebTV) or too late to make an impact (like the truly excellent Zune HD). But when the company launched the AI-powered Bing Chat earlier this year, it was well-positioned to ride on the success of ChatGPT, which reportedly hit 100 million users in just two months. UBS analysts said this would make ChatGPT the fastest-growing consumer app in history. What better time to combine the power of generative artificial intelligence with one of its notoriously troubled products? Microsoft has nothing to lose.
After investing a total of $13 billion in ChatGPT maker OpenAI (and acquiring a 49% stake in the process), Microsoft may be eager to show off its shiny new toy before Google and others. The launch of Bing Chat officially kicks off Microsoft’s year of artificial intelligence: Copilot launches on Microsoft 365 products like Edge, Word, and Powerpoint, and eventually makes its way to Windows 11. Even more surprising, the company recently announced that Copilot will be coming to Windows 10—a sign that it wants to see the AI capabilities in front of as many people as possible. (Windows 11 reportedly accounts for 26% of Windows installations, while Windows 10 still accounts for 69%. By targeting both platforms, Copilot could reach as many as 1.4 billion users.)
There’s no doubt that the co-pilot makes a great first impression. Type a few words (or say them out loud) and it’ll come back and answer your questions directly, like a smart assistant. There are no ads to view and you can simply click on other links if you wish. It’s a glimpse into the world beyond search engines, where artificial intelligence can help guide us through an increasingly confusing media landscape.Microsoft’s Copilots can also help in specific applications: In Edge, it can summarize the web page you’re viewing; it can help transcribe and generate action points in team meetings; it can help discover hard-to-find settings in Windows (For example, you can just type "How to turn on night mode?" turn it on).
But Copilot’s confident exterior masks the fact that it often gets it wrong and sometimes completely misunderstands your question. It’s far less responsive than using a typical search engine because there’s a lot of opaque artificial intelligence processing happening in the background. In my testing, it also crashed more often than you might think, requiring a “restart” of your session (but at least it doesn’t flash a blue screen like Windows does).
To temper our expectations, Microsoft added a helpful note at the top of Bing’s AI chat: "Bing is powered by artificial intelligence, so surprises and errors may occur. Please share feedback so we can improve!" Microsoft seems to be showing a bit of humility, admitting its AI chat isn’t perfect, and trying to win some brownie points by showing it’s listening to your feedback. Mostly, though, this warning is a way out for Microsoft. It can advertise Copilot’s ability to write your essays and have vaguely realistic conversations for you, but once it screws up, all the company can say is, "This is just a beta version, haha!"
The big test facing Microsoft Copilots and other generative AI tools comes down to one thing: trust. Can users trust that relevant information will be provided when asking questions? Can we be sure that Copilow correctly understands our query?Aaron Woodman, vice president of Windows marketing at Microsoft, told us that trust ultimately lies with the user "kick tires" See for yourself how Copilot performs. "I think over time we’ll see this organic growth," He told Engadget in an interview at the Windows Copilot launch event in September. "I bet it’s going to be explosive because the value is there, and I think customers will see that very quickly."
Woodman also believes users will understand that Copilot won’t always be perfect, especially early on. "I strangely think that we may be able to understand people’s growth and development better than technology," He said. "I think the best thing we can do is honestly admit it and be transparent about it.To some extent, every conversation we’re having, we’re trying to fit in [that] This is a process of growth. We want to make sure you understand the reference.I think people will understand that we’re trying to accelerate [new] Technology to them."
I’ve been using Microsoft’s AI solution since Bing Chat launched earlier this year, and while it’s helpful for simple tasks like creating a spec sheet comparing two products, it hasn’t completely changed the way I work. Microsoft also had to severely limit Bing Chat’s functionality early on, after it started arguing with users and issuing troubling responses. In Windows 11, Copilot can sometimes help me find settings like dark mode, but it doesn’t always bring up the controls in the Copilot pane, and sometimes it just redirects me if it can’t figure out what I want to do. Send to general settings menu I request.
I recently had a disappointing conversation with Bing when I asked whether now was a good time to buy a Nintendo Switch (which, after some prodding, brought up rumors of a possible Switch follow-up next year). The ability to answer questions about images is still not as useful as Google’s image search.
When I took a picture of my child’s baby monitor and asked, “What is this?” Bing knew what it did, but it got the actual model and manufacturer wrong. The query also took five seconds to complete. A Google image search took half a second to correctly identify it as the Eufy Space Monitor. One point for traditional search (yes, I know it’s also powered by its own set of computer vision models).
We can think of Microsoft Github Copilot, launched in November 2021, as a way for users to learn to use AI. It mainly acts as a partner for experienced programmers: it will look for potential problems and even write some simple code.
According to developer Aidan Tilgner, Github Copilot can be really useful for programmers, as long as you keep your expectations in check.in paper "GitHub Copilot AI Pair Programming: Asset or Liability?" Authors Arghavan Moradi Dakhel, Vahid Majdinasab, Amin Nikanjam, Foutse Khomh, Michel C. Desmarais, and Zhun Ming Jiang found Github Copilot equally useful, but please note "It can also become a burden if used by novices, i.e. those who may not be familiar with the problem context and proper coding methods."
"Copilot’s suggested solution may be problematic and difficult to understand, but newbies may see it as the correct solution," the author added. "Adding such flawed and complex code to a software project can seriously affect its quality."
By relying so heavily on Copilots for the future, Microsoft may also be tying itself too closely to OpenAI, a young company that recently endured one of the most tumultuous weekends in Silicon Valley history. OpenAI’s board fired CEO Sam Altman, but after intense internal pressure (and some cajoling from Microsoft CEO Satya Nadella), it finally ended within days He was later rehired. If OpenAI experiences another turmoil, it won’t just be Microsoft’s $13 billion investment that’s at risk: it’ll be the company’s future plans for nearly all of its products.
according to Windows CenterMicrosoft’s next major Windows update, “Hudson Valley,” will likely be released next year and will come with a host of artificial intelligence enhancements. These include the ability to analyze content displayed in video chats, improved Copilot that remembers everything you do on your PC, and better system-wide searches. Certain features may also require a CPU with an NPU, such as AMD’s last batch of chips and Intel’s new Core Ultra hardware. This is similar to Windows Studio Effects features such as background blur and automatic framing, which also require an NPU.
One of the constants surrounding artificial intelligence today is that everything is changing rapidly. Since I started writing this article, Microsoft announced that Copilot will be upgraded to more powerful GPT-4 Turbo and Dall-E 3 models, which will make them even more powerful. Maybe Microsoft and OpenAI can finally solve all the problems I’ve encountered with Copilot so far and finally deliver a transformative AI tool that everyone can easily use. But when it came to the company’s dual-screen Duo and Neo plans, I also hoped for the best, and all I got was disappointment.
This article was originally published on Engadget: https://www.engadget.com/microsoft-bet-big-on-ai-in-2023-but-its-ai-future-is-still-unclear-143055721.html?src= rss
Source link
