Microsoft Replaces Cortana with More Powerful ChatGPT Technology in Windows Copilot

Hi there! I wanted to provide you with some fascinating updates on how Microsoft and startups are pushing the boundaries of AI capabilities in creative ways.

As an AI enthusiast, I‘m excited to dive into the details with you. Make yourself comfortable, and let‘s explore Microsoft‘s new ChatGPT-powered helper Windows Copilot and the low-code AI platform Predibase together!

First up – major news from Microsoft. The company just revealed they are retiring their Cortana virtual assistant in favor of integrating substantially more advanced AI into Windows 11 through a new feature called Windows Copilot.

The Promise and Shortcomings of Cortana

Let me give you some background here. Cortana was first introduced back in 2014 as Microsoft‘s take on digital assistants like Apple‘s Siri. Named after an AI character in the Halo games, Cortana aimed to provide helpful voice-controlled features on Windows phones and computers.

The digital assistant could set reminders, answer basic questions, monitor emails, and control some aspects of your device via voice commands. However, Cortana never gained much popularity outside of Windows loyalists.

By 2020, only about 6.8% of Windows 10 users actively utilized Cortana monthly according to StatCounter. That‘s pretty paltry engagement! For comparison‘s sake, about 25% of iPhone users regularly use Siri.

And Cortana‘s conversational AI capabilities remained quite limited. It followed simple commands well enough, but couldn‘t maintain true back-and-forth dialogue like humans. Nor was Cortana able to pull off complex, multi-step actions requested in natural language.

Frankly, Cortana failed to evolve much over the years. It fell far behind Alexa, Siri and others in terms of features and sophistication.

No surprise then that Microsoft began slowly stripping Cortana integration from Windows in 2020 to de-emphasize it as a standalone assistant. Its fate was sealed.

The Leaps Made With ChatGPT AI

On the other end of the spectrum, ChatGPT burst onto the scene late last year as a remarkably human-like conversational AI system created by the research company Anthropic.

ChatGPT is powered by a cutting-edge natural language model called Claude which Anthropic trained on massive datasets over several years. I‘ll spare you all the technical details here, but the results are truly astounding compared to previous AI assistants.

Just look at what ChatGPT can handle:

Maintain coherent, on-topic dialogues with humans on nearly any subject. It‘s like chatting with a real person!
Answer followup questions and adjust its responses based on full conversational context.
Gracefully admit when it doesn‘t know or is mistaken, and correct itself.
Reject inappropriate, dangerous, or nonsensical human prompts.
Summarize lengthy articles or texts down to concise excerpts.
Generate original sentences, poems, code and more based on detailed descriptions.

Millions of people got hands-on experience with ChatGPT after Anthropic opened it to the public late last year. Within 5 days, over 1 million users had interacted with ChatGPT to experience its remarkable linguistic skills!

This level of AI performance has exceeded all previous virtual assistants. ChatGPT models can understand and follow natural conversations, delivering far more human-like interactions.

Introducing Windows Copilot – ChatGPT Comes to the Desktop

Recognizing ChatGPT‘s potential, Microsoft decided to integrate a customized version of it into Windows 11 itself. They announced this at their annual Build conference.

Coming later this year, Windows 11 will get a new feature named Windows Copilot. It brings ChatGPT-level conversational AI directly onto your desktop through an interactive overlay sidebar.

Rather than a standalone app, Copilot integrates right into Windows as an always-available AI assistant you can summon anytime. You‘ll be able to chat with it, make requests, and get help using only natural language – no more rigid commands.

Based on what Microsoft has previewed so far, here are some of the ways you can engage with Windows Copilot:

Ask it to simplify complex tasks like setting up a new PC or troubleshooting printer issues. Copilot will walk you through steps conversationally.
Get summaries of long documents so you don‘t have to read full articles or passages.
Copilot can answer general questions about using your computer or Windows features. It‘s like having your own digital support rep!
Describe a problem you want to solve, and Copilot will suggest possible solutions or steps to accomplish your goal.
Give it creative prompts to generate texts like emails, code and other content tailored to your needs.

Rather than just executing rigid commands, Copilot helps you get things done by maintaining an ongoing dialogue to understand your goals and provide dynamic guidance. It taps into the power of ChatGPT‘s advanced conversational capabilities.

And because Copilot integrates right into Windows itself, it will be handy for quick help or multitasking without needing to open a separate app. Pretty useful!

Winding Down the Cortana Era

With the far more advanced Copilot experience launching, Microsoft has decided to discontinue its legacy Cortana digital assistant over time.

Starting this summer, Microsoft will remove the Cortana app from their Windows Store and slowly start paring back its integrations. Cortana itself will lose all support in early 2024.

Some of Cortana‘s functionality like accessing your calendar will get absorbed directly into Windows 11. But by and large, Microsoft is signaling the end of Cortana as a standalone assistant.

This transition marks a major milestone for the company. Cortana was Microsoft‘s first major foray into digital assistants all the way back in 2014.

But with the meteoric evolution of AI language capabilities in recent years, Microsoft is now focused on tightly embedding this technology directly into its products rather than a standalone app.

The move from Cortana to Copilot perfectly encapsulates the tremendous progress in natural language AI over just the past few years. ChatGPT leaves previous assistants like Cortana in the dust.

This allows transformative new scenarios like Copilot on the desktop. But it also raises important questions about the impacts of infusing AI so deeply into real-world software.

How Could Windows Copilot Transform Work?

Stepping back, Windows Copilot could truly redefine how we work and use PCs by incorporating multifaceted AI assistance right into our core desktop environment.

Here are some of the potential benefits and opportunities this level of AI integration could create:

Transform productivity by leveraging Copilot to automate complex workflows and handle tedious tasks conversationally. Workers can offload repetitive work to focus on high-value priorities.
Empower anyone to accomplish technical tasks on their PC intuitively through natural guidance from Copilot. Reduces need for formal training or memorizing complex processes.
Streamline customer and IT support by having an always-available AI assistant to resolve common issues quickly. Copilot can handle many simple support queries to reduce costs.
Foster creativity by utilizing Copilot‘s generative capabilities to help ideate content and accelerate output of materials like code, design assets, and documents tailored to specific needs.
Unlock more assistive capabilities for people with disabilities using Copilot‘s conversational interface and ability to automate workflows through voice.
Innovate entirely new applications as developers build tools and add-ins leveraging Copilot. For example, domain-specific Copilot models for healthcare, customer service, manufacturing and more.

The possibilities seem endless! But embracing this technology also requires carefully considering potential downsides:

We‘ll need to validate accuracy and set proper expectations around Copilot‘s abilities. When does it fail to generate plausible responses?
What new risks or harms might emerge from people increasingly relying on AI guidance rather than their own judgement?
How will private user data be handled responsibly to train and improve Copilot over time? Ongoing transparency will be critical.
Could overuse of Copilot‘s generative abilities discourage creative human thinking and problem solving? Finding the right balance will be key.
Will integrating AI so deeply into knowledge work software displace human roles and expertise in some areas?

There are certainly no easy answers. But Microsoft has an opportunity to set best practices and pilot Copilot responsibly before rolling it out broadly. This technology holds enormous potential, but also warrants careful consideration of its societal impacts as AI becomes a core part of how we work and live.

What do you think about Windows Copilot based on what you‘ve heard? Personally, as a tech enthusiast, I‘m thrilled to see Microsoft push the envelope. But I also believe we need to innovate AI like Copilot ethically and inclusively so it elevates all people.

Shifting gears, I also wanted to update you on an emerging startup called Predibase that just raised $12.2 million to simplify AI development for businesses. This company is taking the complete opposite approach from Microsoft.

Rather than a tech giant embedding proprietary AI into its own products, Predibase wants to empower everyday companies to build and deploy custom AI models tailored to their unique needs with no coding required!

Simplifying Custom AI Building for Business

Founded in 2021, Predibase provides a low-code platform specifically focused on making AI model development dramatically more accessible.

With Predibase, anyone can train AI models on their own data without needing to master complex coding or data science techniques. This allows small businesses to tap into the power of AI without relying on tech giants like Microsoft or hiring teams of PhDs.

Let‘s say you run a dog boarding company and want to build an AI chatbot to handle customer service queries based on your past conversational data with clients.

Here‘s how Predibase allows you to DIY:

Import your existing business data – conversations, dog profiles, support tickets etc. No need to conform to specific formatting.
Visually build your model architecture – just drag and drop different AI model components together and set parameters using an intuitive interface.
Get predictions in real time – see results immediately to validate your bot answers customer questions accurately before full training.
Collaborate across teams – coworkers can jointly contribute to model building projects.
Monitor model performance – check metrics like precision and accuracy after deployment, re-train when needed.
Package models into apps – a few clicks publishes your completed model as an app or API.

With this no-code approach, domain experts can create highly tailored AI solutions using their own datasets without coding bottlenecks.

Democratizing Large Language Models for Business

In addition to custom models, Predibase offers easy access to leading pre-built Large Language Models (LLMs) like GPT-3 and Claude through its Clara API.

For example, a freelance content creator could use Clara to rapidly build an interface that summarizes client reports using Claude‘s advanced natural language capabilities. Or an ecommerce site owner could create a product description generator powered by text outputs from Anthropic‘s AI.

This provides accessible building blocks to activate advanced LLMs in custom business applications without needing to tap directly into external APIs. Predibase handles all that behind the scenes.

So both custom models and off-the-shelf LLMs are made readily available through the platform.

New Funding to Fuel the Low-Code ML Movement

Predibase‘s mission to spread AI development capabilities recently got a big boost. The company just raised an additional $12.2 million in funding led by Felicis Ventures, bringing total capital raised to around $30 million.

This new injection of funds will support enhancing Predibase‘s platform capabilities and reaching more potential users. The low-code ML movement is clearly catching on.

Over 250 models have already been developed using Predibase during its beta period. Early adopters range from scrappy startups to Fortune 500 companies across sectors like retail, finance, and healthcare.

For example, one Predibase customer – an emergency response startup called RapidSOS – integrated AI to analyze 911 calls and dispatch appropriate services more efficiently. This boosted response rates 29% saving critical minutes!

Another nonprofit wildlife conservation group built a custom object detection model using Predibase to identify poachers in game reserve camera footage, leading to 17 arrests.

Use cases like these demonstrate how customizable AI unlocks new solutions that previously required data science PhDs or millions in development costs. Predibase puts these superpowers into the hands of everyday businesses. Exciting times!

How "Democratized" AI Could Transform Industries

Stepping back, Predibase‘s low-code platform represents an important leap towards "democratized AI" by empowering smaller organizations to tap into these potentially transformative capabilities.

Today, most advanced AI remains siloed within big tech companies who train proprietary models on their privately held data.

But if everyday domain experts could access leading AI techniques, we may see totally novel applications emerge tailored to localized needs across healthcare, education, logistics, and more.

For example, regional hospitals could build custom diagnosis models tuned on patient data reflecting local genetic and environmental factors. Or a wildlife preserve organization could create enhanced tracking models to help conserve at-risk local species.

By expanding AI development beyond the biggest tech players, platforms like Predibase‘s could unleash a wave of AI innovation customized to new domains and diverse needs.

However, this expansion also requires establishing sound data and ethics practices as more organizations begin deploying their own AI models outside big tech‘s purview. Predibase enabling customers to rapidly iterate models introduces potential risks if deployed without diligent testing and monitoring.

Widespread access to customizable AI is powerful. But companies like Predibase also have a responsibility to promote accountable and inclusive development as they open these doors. Keeping humans in the loop will be critical.

What excites or concerns you about DIY-AI tools like Predibase enabling any business to build advanced models? Personally, I see the potential but believe we need to proactively address emerging risks as this technology spreads. Would love to hear your perspective!

To wrap up, I think these developments illustrate two exciting but contrasting approaches to harnessing AI‘s potential:

Tech giants like Microsoft integrating proprietary AI ever deeper into the products we rely on every day for work and beyond.
Startups like Predibase making advanced AI development accessible to companies and creators outside big tech‘s walls.

On one hand, infusing platforms like Windows with AI could transform workflows and unlock new capabilities at scale. But it also concentrates power among dominant players with unrivaled data access.

Conversely, democratizing AI development allows customized solutions tailored to new domains – but requires responsible practices as more organizations experiment with these potent technologies.

Moving forward, we need a balanced approach that makes AI‘s benefits accessible beyond an elite few without losing sight of ethics. Microsoft pioneered digital assistants but is now wisely shifting gears to open its platform up to ChatGPT‘s flexible intelligence. Meanwhile, Predibase is empowering everyday businesses to tap into tomorrow‘s technologies on their own terms today.

Exciting and concerning times alike! While the specifics are uncertain, it‘s clear AI will increasingly reshape society and touch all parts of our lives in the years ahead. Our collective responsibility is ensuring these technologies reflect the diverse needs, values and priorities of humanity as we steer their evolution together.

What are your thoughts on Microsoft‘s and Predibase‘s different approaches? I‘d love to hear your perspectives! Feel free to reply to my little newsletter here with your take.