From 4o to o1
Formerly know as Strawberry
Upgrade your customer service with ShopRep's AI-powered calling system. Automate calls, enhance productivity, and engage customers effortlessly. Try it now!
Friends,
I remember when computers were mostly glorified calculators. They could crunch numbers at incredible speeds but couldn't hold a conversation or reason through a problem. They did what you told them, nothing more. But this week, OpenAI unveiled something different: an AI model called o1. Unlike its predecessors, o1 doesn't just spit out answers—it takes the time to think. Previously called “Strawberry”, this feels like a new or adjacent phase in the AI story.
Traditional AI models are like race cars. They zoom through computations, delivering quick results but often missing the nuances. o1 is more like a thoughtful student, the kind who sits in the back of the class, pondering every angle before raising a hand. It doesn't just process inputs; it engages with the problem, breaking it down methodically.
One of the most intriguing aspects of o1 is its ability to handle complex, multi-step tasks. It uses reinforcement learning, rewarding itself for correct answers and learning from mistakes. I gave it a strategy problem that had been swirling in my mind—a tangled web of customer needs, product configurations, and ambiguous variables of addressing poverty and education in the third world. I didn't simplify it. I just laid it out and said, "Figure out what to do." In less than a minute, o1 came back with a clear, organized recommendation. It was the kind of analysis that would have taken me hours.
What's remarkable is that I didn't have to guide it with prompts like "Think step-by-step" or "Be precise." as I usually do when prompting. It just got it. It felt like interacting with someone who not only understands the complexity but knows how to navigate it. It was the kind of depth one would get in a conversation with another human.
But this depth comes with a cost. Literally. Using o1 is expensive compared to earlier models. It's also slower. Sometimes it takes over ten seconds to provide an answer, which feels like an eternity in our instant-gratification world. Yet, maybe that's a fair trade-off for thoughtful responses.
This makes me think about the future of AI. We're moving toward models that are specialists rather than generalists. o1 isn't the tool you'd use for simple tasks like translating a sentence or fetching a basic fact. It's the one you'd turn to for the tough questions, the ones that require deep reasoning.
We're at a point where it's up to us, the users, to figure out what we want and which models to use. Not all questions are the same, and different tools will serve different needs. We're not getting a HAL-like machine that knows everything; we're getting a toolbox filled with specialized instruments.
This shift also brings new challenges. As AI becomes more integrated into our workflows, we'll need to rethink how we interact with these systems. Do we need to "train" them like new employees? How do we ensure oversight and accountability?
Imagine AI agents handling tasks like recruiting—scanning profiles, scheduling interviews, even conducting initial assessments—all without human intervention. Or AI systems that create educational content, gather expert input, and monitor engagement, acting almost like team members rather than tools.
o1 gives us a glimpse into this future. It's a step toward AI that doesn't just compute—it thinks. While it's not perfect and still makes mistakes, the potential benefits are significant.
We're at the beginning of a new phase in AI development. Models like o1 are pushing the boundaries of what's possible, moving us closer to machines that can truly understand and engage with complex tasks. It's an exciting time, but also one that requires us to carefully consider how we integrate these powerful tools into our lives.
Stay Curious - and don’t forget to be amazing,
PS. In addition to this newsletter, I recommend some other great ones. All free. Check them out here.
PPS. Thanks for the ongoing messages and good vibes as I return to health. It’s a bit of a yoyo at present, but recovery will continue to be slow.
Here are my recommendations for this week:
One of the best tools to provide excellent reading and articles for your week is Refind. It’s a great tool for keeping ahead with “brain food” relevant to you and providing serendipity for some excellent articles that you may have missed. You can dip in and sign up for weekly, daily or something in between -what’s guaranteed is that algorithm sends you only get the best articles in your area of choosing. It’s also free. Highly recommended Sign up.
Now
The Best 20 Resources on AI in Education. If you want to see the impact of AI on one industry, education shows exactly what is happening. And these are the reports that show what’s happening, what’s ahead and considerations.
Breaking Down OnlyFans’ Stunning Economics: OnlyFan’s revenues, profits, scale, defensibility, reach, and impact make it probably the most successful UK company founded since DeepMind in 2010, and the most significant media platform founded since TikTok (via Musical.ly in 2014), and dedicated creator economy platform… ever Semi - related: The Palace Coup at the Magic Kingdom
The Core ‘Friend Group’ Is a Myth—and It’s Making Us Feel Bad About Ourselves
A tight crew that does everything together isn’t that common, and it’s unrealistic for many of us.
The Psychopaths of TikTok - Content creators are turning to social media to destigmatize—and make money off of—their personality disorders.
Here's How to Nap Like a Pro Without Destroying Your Sleep Quality at Night
Napping without disturbing the quality of sleep you get at night can be a delicate balance. Try these six tried-and-true tips. Health Related: Are We Losing Our Appetites for Veganism?
Next
Anthropic’s Mike Krieger wants to build AI products that are worth the hype
Anthropic’s new chief product officer on the promise and limits of chatbots like Claude and what’s next for generative AI.
Mark Zuckerberg’s 20-year mistake: Meta’s CEO says he’s done apologizing. Should we worry?
Inside Google’s 7-Year Mission to Give AI a Robot Body - As the head of Alphabet’s AI-powered robotics moonshot, I came to believe many things. For one, robots can’t come soon enough. For another, they shouldn’t look like us.
The Button Problem of AI - What’s the real reason AI hasn’t yet delivered on its hype?
Google DeepMind teaches a robot to autonomously tie its shoes and fix fellow robots - In a new paper, Google DeepMind researchers showcase a method for teaching robots to perform a range of dexterous tasks, including tying a shoe, hanging a shirt, and even fixing fellow robots.
Free your newsletters from the inbox: Meco is a distraction-free space for reading newsletters outside the inbox. The app has features designed to supercharge your learnings from your favourite writers. Become a more productive reader and cut out the noise with Meco - try the app today
If you enjoyed this edition of Box of Amazing, please share and help me grow this group. If you share it on LinkedIn, please tag me so I can thank you.
If you are interested in advertising in this newsletter, please see this link