After months of speculation, OpenAI have finally launched their next model, called “o1”. It’s an AI chatbot, like their previous flagship “ChatGPT-4o” model, but it’s a bit different in how it works. It’s a new tool designed to handle reasoning and problem-solving tasks better than previous models.
Over the past few days, I’ve been experimenting with it, I’ve also spoken to friends in the AI field and read insights from other AI experts.
This guide to help you figure out when o1 is the right choice and when it’s better to stick with other models like GPT-4o, Claude 3.5 Sonnet, or Gemini 1.5 Pro.
The o1 model stands out because of its “chain-of-thought” reasoning. It breaks complex problems into smaller steps, working through them like a human expert.
This makes o1 especially good for solving detailed, multi-step problems. But, it’s not always the best tool – sometimes, you’ll want the faster (and cheaper) performance of other models.
In this post, I’ll explain when to use o1 and when other models might be better suited to your needs. I’ll also cover the current usage limits and how to plan around them.
Table of Contents
Example: Getting Detailed Feedback on my Book Manuscript
One test I wanted to try with o1 was to see how it dealt with giving feedback on a large amount of text. I gave it effectively 2 full chapters of my upcoming book “The AI-Powered Thought Leader”, and asked it for detailed line by line feedback. You can see here how it thought for 62 seconds before replying. Note that this is an exceptionally long time for o1, in other tests it came back with a response within 5 to 10 seconds typically.
I’ve expanded the thinking section and included it below to show you what it’s telling me:
Here’s the introduction and first chapter of my book “The AI-Powered Thought Leader”. I want you to read through it carefully, analyse it, and from the context of a business book editor, I want you to give me criticism and specific actionable suggestions for improvement. Be very specific about location and text of changes suggested. Any other suggestions or feedback you have would be useful.
{I included the entire text of introduction and chapter 1 from the manuscript – omitted here for space}
Thought for 62 seconds
Reading and analyzing
I’m gathering insight from the user’s request, noting a need for careful reading and analysis of the book’s introduction and first chapter, with critical feedback and actionable suggestions for improvement.
Reviewing policy adherence
I’m mapping out the user’s request to ensure it aligns with OpenAI’s policies. The text appears acceptable, focusing on providing useful and professional advice.
Setting the scene
I’m mapping out the introduction and first chapter of “The AI-Powered Thought Leader,” noting the engaging anecdotes and conversational tone. I’m also identifying areas for enhancing clarity, structure, and conciseness to better guide the reader.
Crafting a coherent introduction
I combined the original sentences to create a more cohesive introduction, clearly identifying the subject as a Star Trek fan.
Enhancing emphasis
I’m tweaking the text by using italics instead of all caps for “FICTION” to avoid distractions.
Enhancing clarity
I revised the sentence to ensure it remains engaging by making clear that AI’s benefits and potential impact will unfold as it continues to evolve, fostering topics for further exploration.
Refining the message
I noticed redundancy and a grammatical issue. After tweaking the content to reduce repetition and ensure clarity, I crafted a revised message that communicates the transformative role of AI for thought leaders.
Navigating AI’s impact
I’m piecing together the challenge of integrating AI into thought leadership, emphasizing the need for authenticity and impact amidst rapid technological change.
Synthesizing insights
I’m pulling together 17 years of tech expertise and insights from top AI, leadership, and business experts featured on my podcast to create a comprehensive and actionable book.
Shaping perspectives
I’ve explored future of work, AI ethics, human-AI collaboration, and authority-building with top business minds. This journey highlights the multifaceted challenges and rich insights we’ve tackled together.
Combining insights
I’m piecing together the enlightening and challenging insights from various esteemed figures, merging their unique perspectives to reshape my own thinking thoroughly.
Refining the message
Simplifying the text to clearly convey how my work as an AI business consultant, podcast host, and author aids thought leaders in leveraging AI for authority and impact. Progress is being made in streamlining the verbiage for clarity.
Enhancing the introduction
First, the statement is revised to give it a stronger impact. Next, the bulleted points are improved for clarity and consistency. Finally, the introduction wraps up with a clear reflection on the significance of AI.
Refining the title
I’m noting the need for a consistent heading format for “Why Thought Leadership?” and suggesting italicizing “Are you ready?” for emphasis. Progress is being made!
Revamping the tone
I’ve updated the introduction to address potential mild skepticism about the term “thought leader,” refining it to avoid confusion for readers.
Enhancing thought leadership
I’m refining the thought leadership definition by clarifying its impact and expertise. Progressing through knowledge dissemination, it now emphasizes industry advancement and problem-solving.
Highlighting the impact
Thought leadership is powerful, fostering problem-solving, industry education, innovation, clarity in complex issues, and shaping key discussions. This approach establishes authority, creates opportunities, and maintains relevance.
Mapping a thought leadership journey
I’m working through the steps to become a thought leader: harness unique knowledge, engage with others, learn continuously, and build credibility with quality content.
Transforming thought leadership
I’m outlining the crucial role of AI in redefining thought leadership. It’s reshaping our approach, tools, and strategies, ensuring thought leaders stay relevant in the evolving landscape.
Injecting urgency
I’m pulling together a sense of urgency and reality, emphasizing the immediacy and inevitability of technological advancements affecting thought leaders’ relevance.
Contemplating AI’s impact
I’m piecing together how AI is revolutionizing industries like consulting, manufacturing, and marketing. It’s unlocking immense potential for those who adapt, while stirring fears about job displacement and obsolescence.
Investigating challenges
I’m digging into how AI’s rapid content generation might devalue human thought leadership by making it hard to distinguish between human and AI work.
Emphasizing the impact
Failing to embrace AI and digital transformation will undermine expertise and influence, making acceptance and adaptation crucial.
Refining content
I’m pulling together key ideas and ensuring consistency in capitalization, aiming to present a clear and accurate perspective on AI’s integration into daily life. Progress is steady and thoughtful.
Highlighting AI’s acceleration
I’m working through the exponential rise in AI’s capabilities. What was recent progress is now outdated, showcasing breakthroughs once deemed impossible. This makes me think that the pace of change is relentless.
Examining AI’s impact
Currently, healthcare diagnoses use AI for accuracy and drug discovery. In manufacturing, AI enhances quality control and streamlines supply chains.
Enhancing AI’s impact
I’m thinking through AI’s role in marketing. It tailors customer experiences, boosts ad engagement, and crafts captivating content. This approach effectively personalizes and targets users, amplifying marketing strategies.
Highlighting authenticity
I discuss how it’s crucial to emphasize authenticity in producing unique content. Generic AI-generated content can’t replace your distinct experiences and insights, which are key to building trust and thought leadership.
Crafting the narrative
I’m curious about ensuring a clean and engaging storytelling flow, with a consistent tone, clear grammar, and a well-structured business case for AI.
Overall Feedback
The introduction and first chapter of “The AI-Powered Thought Leader” set a promising foundation for your book. Your personal anecdotes and conversational tone make the content relatable and engaging. However, there are areas where clarity, structure, and conciseness can be improved to enhance the reader’s experience. Below are specific suggestions for improvement, organized by sections and specific locations within the text.
{Alastair: specific detailed, useful advice followed – I’ve omitted for space}
Model: ChatGPT o1-preview
The feedback it provided was incredibly detailed, and was useful. I implemented several of the suggestions.
When to Use the o1 Model
1. Complex Problem-Solving
From what I’ve seen in my tests, o1 is excellent at breaking down tough challenges that require step-by-step thinking. If you’re dealing with complex technical issues, o1 can really help. It’s especially useful for tasks like:
- Coding: Debugging or solving multi-step coding issues.
- Large sequential problems: Like solving physics problems or analysing large sets of health data.
My friends in academia, as well as AI experts like Ethan Mollick from Wharton, have been talking about how o1’s reasoning capabilities are like having a human expert work through tough problems. It might be slower than other models, but it’s more accurate when you need a careful, step-by-step approach.
2. Creative Problem-Solving
o1 also shines when you’re exploring different solutions to creative problems. In product development or strategic planning, o1’s ability to work iteratively is a big advantage. You can use it to test various options before settling on the best one. It’s helpful in situations like:
- Innovation: Trying new approaches to tackle unique problems.
- Strategy: Planning long-term projects that require detailed thought before deciding on a course of action.
Here’s an example of how I’ve used o1 in product development:
- Analyse market trends and spot gaps
- Generate several product ideas
- Evaluate each idea
- Refine the best ones
- Build a step-by-step plan to develop the product
3. Education and Tutoring
One of o1’s best features is its ability to teach by walking you through problems. It doesn’t just give the answer, but shows the reasoning behind it. This is especially valuable for education, where understanding the process is as important as getting the right result.
For example, if you’re solving a complex math problem, o1 might:
- Break the problem into smaller parts
- Explain each step clearly
- Show you other ways to solve the same problem
- Use real-world examples to explain abstract concepts
Right now it seems to be showing us some of its reasoning, but not all – I think that’s a protective mechanism from OpenAI to prevent giving away proprietary details of how it works.
4. High-Stakes Fields
If you work in industries where mistakes are costly, like healthcare or aviation, o1 is ideal. Its step-by-step approach is perfect for tasks where accuracy matters. In these cases, the extra time it takes is worth it. o1 can help with tasks like diagnosing a condition or reviewing safety data.
When to Stick with Other Models
While o1 is great for complex tasks, you’ll find that other models, like GPT-4o or Claude 3.5 Sonnet, work better for simpler, time-sensitive tasks. Here’s when to use those instead:
1. Quick, General Responses
If you need fast answers, models like GPT-4o are a better choice. They’re ideal for:
- Customer Support: Answering FAQs or simple problems quickly.
- General Q&A: Getting basic information fast without needing detailed reasoning.
2. Content Creation and Marketing
If you’re creating blog posts, marketing copy, or social media content, stick with faster models. They can generate creative, flowing text quickly and work well when speed matters more than in-depth thinking.
3. Conversational AI
For chatbots and virtual assistants, other models handle real-time conversations better. They provide smoother, more engaging dialogue without the delay you get from deeper reasoning models like o1.
4. Summarisation and Light Analysis
If you need a quick summary of a long document or basic data analysis, the speed of models like GPT-4o will serve you better. These models are good for:
- Summarising long documents: Pulling out key points quickly.
- Basic data analysis: Offering high-level insights without needing too much depth.
5. Time-Sensitive Tasks
When time is of the essence, models like GPT-4o or Claude 3.5 are your best bet. They excel at things like customer service chats where you need to provide fast answers in real-time.
How to Choose Between o1 and Other Models
Here’s how to decide which model to use:
- Is the task complex and multi-step? Go with o1. If you need careful, step-by-step reasoning, it’s the right choice. Mollick explains that o1 is great for tasks that an expert could solve but not quickly.
- Do you need speed? Use another model if you need fast results. Mollick points out that most people won’t use o1 for everyday tasks because it’s slower than other options.
- Are the stakes high? For high-stakes tasks, o1’s slow and careful approach is worth the wait. It might take longer, but the accuracy it delivers is essential.
- Do you need creativity or conversation? If you’re focused on content creation or need smooth, fast dialogue, stick with GPT-4o or Claude 3.5 Sonnet.
Important: Usage Limits for o1-preview
As of September 2024, OpenAI has set strict usage limits on o1-preview. If you have a ChatGPT Plus or Team account, you can send 30 messages a week with o1-preview and 50 messages with o1-mini. Once you hit the limit, you won’t be able to select o1-preview until the next week. We expect these limits to ease over time, but for now, plan your tasks carefully to avoid wasting valuable messages.
Conclusion
OpenAI’s o1 model is excellent for tasks that need careful thinking or multi-step solutions. It’s perfect for complex, high-stakes problems. But for simpler tasks that need fast responses, models like GPT-4o or Claude 3.5 Sonnet are the better choice, especially due to the usage limits.
If you need help using AI in your business, schedule a call with me – let’s discuss how o1 and other AI tools can make a difference for you and your team.
