AI Daily News March 10 2025: 🚀Manus AI vs OpenAI's GPT operator 🗺️Ex-OpenAI scientist’s new path to ASI 👀Microsoft looking to move beyond OpenAI 💊Stanford AI’s obesity treatment breakthrough
16:03 Share

Welcome to AI Unraveled, the show that brings you the latest and greatest from the world of AI. I'm your host. And if you're enjoying these deep dives, we'd really appreciate it if you could hit that like button and subscribe to the podcast on Apple. And I'm the AI expert here to help guide us through today's landscape of AI news. We're taking our usual approach, sifting through a mountain of reports all from a single day. It's like a time capsule of AI innovation.

Today happens to be March 10th, 2025. You got it. It's all about giving you, the listener, a clear and concise understanding of what's making waves in AI right now. So you can stay ahead of the curve. We're talking about cutting through the noise, right? Extracting the most significant bits, those hidden gems of insight that you might otherwise miss. What do you say we jump right into it? Sounds good to me. Let's dive in. What caught your eye in the world of AI agents?

Well, there's an interesting development brewing between Manus AI, a company from China, and OpenAI's GPT operator. We're witnessing two very different approaches to how they are building AI that can actually get things done, you know, like truly autonomous agents. Right. And it looks like Manus AI is all about that autonomous action. I mean, we're seeing reports of it generating comprehensive reports, think like full-fledged reports, and even deploying entire websites.

all on its own from start to finish. That's a pretty impressive level of independence, isn't it? For sure. And it doesn't stop there. It can browse the live web, pull information, create downloadable files based on that information, and then host those files online. I mean, that's a whole other level of capability. Definitely shows their focus is on creating AI that can handle those complex multi-step processes. Exactly. And it's not just text.

They are talking about what they call multimodal capabilities. So it can work with text, images, even computer code. It's really interesting to see how it integrates all those things. So it's not limited in terms of the types of data it can process and utilize. Now, in contrast, what about OpenAI's GPT operator? How does that compare? Well, with GPT operator, it seems more like a really advanced assistant.

It can definitely offer suggestions and it can certainly help with tasks, but it's not quite at that level of autonomous execution that we see with Manus AI, at least not from what we're seeing in the reports. I see. So it's more about guiding and supporting the user rather than taking complete control of a task from beginning to end. Right. And there are reports that it's a bit slower when it comes to browsing the web. Plus, it seems to have had some issues accessing certain platforms, which kind of makes sense given that difference in design philosophy.

So it's about choosing the right tool for the job. If you need something that can take full ownership of a complex task, Manus AI looks promising. But if you're more focused on assistance and guidance for simpler tasks, especially online, then the GPT operator might be a better fit. Absolutely.

But taking a step back, what's the bigger takeaway here? Why is this comparison so important in the grand scheme of AI? I think it really highlights the growing competition, especially between the West and China, in developing these advanced AI agents. This whole push for autonomy where AI can initiate and complete complex tasks without any human intervention, it signals a potential shift in how we use AI in the future.

And that raises some pretty important questions, too. Questions about efficiency, of course, but also security and ultimately, you know, who's going to lead the way in AI development. For sure. Those are some big questions to consider. Now, let's switch gears for a moment. There's some exciting news coming out of Stanford AI regarding a potential breakthrough in obesity treatment. This is really interesting stuff. So Stanford's developed this AI driven approach centered around a new treatment they're calling BRP.

What's really interesting is that it targets specific areas in the brain rather than having a broad effect on the entire body. That's fascinating. What's the advantage of that targeted approach? Well, they believe that by being so precise, BRP could potentially avoid many of the side effects that come with current obesity medications, like nausea and muscle loss.

You know, things that people often struggle with when taking medications like Ozempic. It would be a game changer for people looking for a more comfortable and effective treatment option. Exactly. And the results from their animal studies are pretty impressive. They found that a single dose of BRP cut food intake in half for both mice and mini pigs. Wow, that's a significant reduction. Yeah.

And they also found that obese mice lost a substantial amount of fat after only two weeks of treatment. Those are some pretty remarkable findings. Now, how exactly did AI play a role in developing this BRP treatment? They use an AI system they call, get this, peptide predictor.

It was designed to analyze, get this, 20,000 human genes. Can you believe there? 20,000 human genes. That's a lot of data. Right. And the goal was to find a natural molecule that could be effective in treating obesity. So the AI was basically sifting through all these possibilities trying to find the needle in the haystack. Acting as a super-powered research assistant, essentially. Exactly. And they're moving pretty quickly. They've already formed a company to get this into clinical trials.

Katrin Svensson, one of the researchers involved, even said that BRP could revolutionize how we treat weight loss in the future. That's incredible. It sounds like they are really pushing the boundaries of what's possible in medicine. They are. And it's not just about this one molecule. They are also using AI to personalize weight loss strategies.

So instead of a one-size-fits-all approach, they are using AI to analyze individual metabolic data and even behavioral patterns to create customized treatment plans. So it's really about tailoring the treatment to each individual's needs. That's the idea. And when you think about it, this really highlights how AI is transforming healthcare. We're talking about the potential for treatments that are more precise, more personalized, and ultimately more effective for patients. It's

It's a really exciting time to be following these developments. Now let's dive into something a bit more future focused. There's news about an ex-OpenAI scientist who set his sights on achieving artificial superintelligence or ASI. Right. Ilya Setskiver, a former key figure at OpenAI, has started his own AI initiative with a bold goal to achieve artificial superintelligence, but with a focus on safety and alignment.

you know, making sure that superintelligence is beneficial for humanity. That's a pretty ambitious undertaking. Is he suggesting a different approach to what's currently being done? It seems that way. He reportedly told investors that he's found a completely new direction for AI development.

You even describe it as climbing a different mountain, emphasizing how different this path is. It definitely piques my curiosity. And what about the financial side of things? His new company, SSI, is reportedly seeking funding with a valuation of a whopping $30 billion. Wow, $30 billion. Yep. And that's even though they haven't generated any revenue or released any product yet. That's a testament to the level of confidence investors have in his vision. What's their plan for reaching ASI?

Well, their strategy seems to be very focused. They're not planning on releasing any commercial products until they've actually achieved superintelligence. And they're operating with a surprisingly small team of only 20 employees. That's a pretty lean operation for such a grand ambition. And for those who might not remember what led to Sutzkever's departure from OpenAI. He left after the whole Sam Altman situation. Remember the one where Altman was ousted and then reinstated back in November 2023?

Sutzkever later said he regretted the role he played in that whole event. So this could be seen as a new beginning, perhaps incorporating lessons learned from his time at OpenAI. Now, for our listeners, what are the potential implications of actually achieving artificial superintelligence? It's a big question, right? I mean, if we're talking about AI that could potentially surpass human intelligence in every way imaginable,

The implications are huge. We're talking about fundamental questions of control, safety and the very future of humanity's relationship with such advanced AI. It's definitely something to think about. It's a topic that deserves serious consideration and one that will only become more important as AI continues to advance.

Speaking of big players, it seems Microsoft might be looking to diversify its AI strategy. You're right. There are reports that Microsoft is exploring other AI partnerships and also ramping up its own internal AI development. They might be moving away from their exclusive focus on OpenAI. Interesting. What's driving this shift? Well, it seems they've been developing their own AI models. They call them MAI. And these models are apparently performing as well as the top models from OpenAI and Anthropic.

Microsoft plans to offer these MAI models through their Azure Cloud platform. So they're building their own internal capabilities. Are they looking at options beyond their own development as well? It seems so. They're reportedly testing their MAI models as replacements for OpenAI technology in their co-pilot AI Assistant.

And they're even experimenting with models from XAI, Meta and DeepSeq. It sounds like they are hedging their bets and making sure they stay at the forefront of AI development. Was there a specific event that prompted this potential shift away from OpenAI? There were reports that Mustafa Suleiman, the current CEO of Microsoft AI, was frustrated last year when OpenAI refused to share details about how their O1 reasoning model works.

That's understandable. In a field that's evolving as rapidly as AI, transparency and access to knowledge are crucial. It definitely adds another layer to the story. And to add to that, OpenAI renegotiated their deal with Microsoft back in January, giving them the freedom to work with server providers other than Microsoft Azure. So things are definitely changing between these two tech giants. And what does this all mean for the broader AI landscape? I think it shows how competitive and dynamic the field is right now.

You have all these major players racing to develop the most advanced AI models. This could lead to even more innovation and a wider variety of approaches in the industry. Definitely an exciting time to be following AI. Now let's explore some more unusual and maybe even surprising ways AI is being used. This should be interesting for our listeners. For sure. Get this. A church in Finland recently held a service where almost everything was generated by AI. Prayers. Hymns.

Even the sermon. Wow. That's definitely pushing the boundaries of AI applications. What kind of questions does something like that raise? Well, it makes you think about authenticity, the meaning of spirituality in a digital age, and the relationship between technology and faith. It's a really fascinating intersection. It definitely challenges our traditional notions of religious practices. What other surprising applications have you come across? Well, how about AI wingmen for the dating world?

There are now AI-powered bots that help people write their dating profiles and even engage in flirtation on dating apps. Now, that's an interesting twist. What are the implications of that? It raises questions about authenticity and deception in online dating. If AI is helping craft your profile and even your conversations, are you really presenting your true self?

And are you really connecting with the other person or just an AI generated persona? That's a valid concern. It definitely blurs the lines and online relationships. What else caught your attention? On a more global scale, the World Health Organization just announced a new center focusing on AI governance in health. That sounds crucial. It is. As AI becomes more integrated into health care,

We need clear guidelines and regulations to ensure that these tools are used safely, ethically, and effectively. It's essential to be proactive in addressing the challenges that come with integrating AI into such a critical field. Absolutely. And speaking of ethical concerns, Scale AI, a company that labels data for AI training,

is under investigation by the U.S. Department of Labor. What are the concerns surrounding scale AI? It's about their labor practices and how they treat their workers. You know, it brings up bigger questions about working conditions in the AI industry, things like fair wages and ethical treatment of the people who do the often unseen but very important work.

of labeling the data that powers these AI systems. - That's a critical reminder that the advancement of AI has real world implications for the people involved in its development. - For sure. And finally, in a more environmentally focused application, Google just released SpeciesNet. - SpeciesNet, what's that?

It's an AI model specifically for identifying wildlife from images and videos. That sounds like a valuable tool for conservation. It is. It can be used to track endangered species, monitor their movements, and detect changes in their habitats. It's all about helping us understand and protect the natural world.

That's fantastic. It's really amazing to see how AI is being applied in so many diverse fields. It is. And that's just a small sampling of what's happening out there. It really is a fascinating time to be following this field. But before we jump into more news, we wanted to take a quick moment to ask for your support.

If you value these deep dives and want us to keep providing this content for free, please consider making a donation. You'll find donation links in the show notes. Every bit helps us continue to bring you this analysis of the ever-evolving world of AI. Now back to our AI news roundup. We've got a few more exciting developments to cover, all from today, March 10th, 2021.

2025. There's just so much happening. It's incredible. For instance, some former DeepMind researchers just launched a new company called Reflection AI with a cool $130 million in funding. Wow, that's a serious investment. What's their focus? They're aiming to develop autonomous coding systems as a step towards...

You guessed it. Artificial superintelligence. The pursuit of ASI continues. Absolutely. And then there's X, the social media platform. They've added a new feature where you can ask questions to their AI, Grok, just by tagging at Grok in your post. So similar to Perplexity, they are making AI powered search and information readily available right on their platform. Exactly. And Alibaba researchers have published a paper on a new AI model called Start.

It's a tool-integrated reasoning model that's supposed to significantly enhance the capabilities of large language models. Interesting. How does it enhance those capabilities? By allowing them to execute code and even do self-checking to improve their accuracy. Makes sense. That ability to verify their own work and utilize external tools is a big step forward. And the advancements in video generation just keep coming. Hedra just...

announced Character 3. It's an omnimodal model that can reason across different types of data, like images, text, and audio, to create high-quality video. Wow, that sounds impressive. It is, and LumaLabs has released Ray 2 Flash, a faster and cheaper version of their video generation model. They say it's three times faster and costs less. They're making these powerful tools more accessible to a wider audience. Right, and for writers, SudaWrite has a new AI model called Muse.

It's specifically designed for fiction writing with advanced storytelling abilities and an improved attention span so it can handle longer and more complex stories. That could be a game changer for authors. It could.

And finally, Sam Altman's World Network launched WorldChat, an encrypted mini app for chat, connections and even verified human money transfers. So they're integrating AI into communication and even venturing into areas like financial transactions. It really is a whirlwind of new developments all happening in just one day. This deep dive has provided a great overview of the incredible pace and breadth of innovation happening right now in AI. It's remarkable.

Absolutely. It makes you wonder what new breakthroughs, what completely unexpected things we'll be talking about just a few months from now.

AI is moving so fast. It's exciting and a little bit daunting to think about where it's all headed. It's definitely a question worth pondering. And on that note, if you're looking to connect with a large audience of tech-savvy professionals, consider advertising on our podcast. It's a great way to spread the word about your business or service to thousands of engaged listeners. You can find details about advertising opportunities in the show notes. Until our next deep dive, keep exploring, keep learning, and keep questioning. Thanks for joining us.