Testing OpenAI's o1 Model in Microsoft Copilot: An Overwhelming Experience

2025-02-04

Microsoft’s integration of OpenAI’s o1 model into Copilot has introduced a new feature called “Think Deeper,” which offers enhanced reasoning and memory capabilities. Released last week, this feature provides users access to the o1 model at no extra cost, giving them an opportunity to interact with OpenAI’s advanced reasoning capabilities over extended periods. This article explores the functionality of this new feature, demonstrating how it offers more complex insights while also being slower than typical AI responses. Through several test queries, I tested the Think Deeper feature’s ability to handle complex tasks and its impact on user experience.

Testing the o1 Model with Different Requests

To explore the Think Deeper functionality, I tested it with three distinct requests that ranged from simple to more complex.

Interior Design Request

My first test involved a relatively simple yet nuanced task: redesigning a small bedroom with a $500 budget. The AI not only provided cost breakdowns for furniture and decor, but also recommended paint colors, lighting, and personal items, with thoughtful suggestions for free changes like scents and sounds. The output even included a philosophical reflection on design and a Frank Lloyd Wright anecdote, showing a deep level of contextual awareness.

Travel Planning

For the second experiment, I asked for help planning a 5-day trip to London on a $2,500 budget. The AI’s response was detailed and immersive, presenting an itinerary with day-specific themes and attractions, including transportation and lodging recommendations. At the end, the AI added a philosophical touch, encouraging me to think beyond landmarks and embrace meaningful experiences. It felt like having a conversation with an insightful travel agent who was also a passionate storyteller.

A Simple Query: First Date Suggestions

Lastly, I asked for advice on planning a perfect first date. The AI’s response was notably human-like, offering a personalized approach by incorporating my previous mention of DIY projects into the date ideas. While the suggestions might not be everyone’s idea of the perfect first date, the AI’s ability to maintain context and remember past conversations was impressive.

Copilot with o1: Smarter but Slower

Using Copilot with the o1 model definitely enhanced the experience. Unlike the regular Copilot model, which tends to be quicker but more superficial, the o1 integration provides smarter and more thoughtful responses. The downside, however, is that it’s slower, making it less ideal for users in a rush. This deeper, more reflective approach allows the AI to provide more nuanced and personalized suggestions, but it comes at the cost of speed.

What Undercode Says:

The integration of OpenAI’s o1 model into Microsoft Copilot marks an important step forward in the capabilities of AI assistants. As seen from the tests above, the o1 model excels in providing deeper insights and maintaining contextual memory across interactions. This is a significant advantage when it comes to handling complex or multifaceted requests that benefit from thoughtful, long-form responses.

The key feature that stands out with “Think Deeper” is the model’s ability to remember context throughout the conversation. While standard Copilot models tend to forget details quickly, o1’s improved memory ensures that previous requests and responses are taken into account, leading to more coherent and personalized results. This is particularly beneficial for users who require continuity in their interactions, such as planning a trip or designing a space.

The AI’s philosophical reflections, like the ones seen in both the travel and interior design tests, highlight its ability to elevate mundane tasks into meaningful dialogues. While these philosophical touches might not always be necessary, they do make the interaction feel more human and less transactional. This depth is what separates the o1 model from simpler AI versions that are typically more direct and task-oriented.

However, it’s important to note that this level of depth comes at the expense of speed. Users looking for quick, direct answers will find the o1-powered Copilot less efficient compared to its faster, more superficial counterparts. In environments where time is of the essence, this can be a drawback. The slower response time also suggests that the model is performing more complex processing behind the scenes, which could be a tradeoff for users in fast-paced settings.

One of the more fascinating aspects of the o1 model is its ability to act as a collaborative partner. It doesn’t just respond to requests; it builds on them, offering insights that might not have been immediately considered. This makes it a valuable tool for tasks that benefit from extended thinking, such as detailed planning or creative brainstorming. The model’s tendency to “overthink” can lead to brilliance in some cases, but it may also result in answers that are more convoluted than necessary.

The integration of o1 into Copilot also signifies the growing sophistication of AI in handling not just data but also context, emotion, and intent. The philosophical musings incorporated into the responses are a testament to the advancements in AI’s emotional intelligence, reflecting a deeper understanding of human interaction. While not every user will appreciate this level of reflection, it’s clear that o1 is pushing the boundaries of what we can expect from AI models in terms of empathy and insight.

In conclusion, the integration of the o1 model into Microsoft Copilot offers a fascinating glimpse into the future of AI. It demonstrates that AI can be more than just a tool; it can become a thoughtful, intelligent partner capable of providing nuanced, context-aware advice. While the slower response time may not suit every user, those who are looking for deep insights, personalized recommendations, and more human-like interaction will find the o1-powered Copilot a valuable addition to their workflow. The future of AI collaboration is certainly looking deeper—and more thoughtful.

References:

Reported By: https://www.techradar.com/computing/artificial-intelligence/i-tested-openais-o1-model-in-microsoft-copilot-and-its-a-little-overwhelming
https://www.stackexchange.com
Wikipedia: https://www.wikipedia.org
Undercode AI: https://ai.undercodetesting.com