Instead of our usual guest interview, we're dedicating another episode to answering the questions that you have been sending our way. It's an opportunity for us to dive deeper into topics that you, our listeners, are most curious about.
We'll be exploring a range of subjects related to personal data and artificial intelligence. The brilliant team at Fulcra will be sharing their thoughts on everything from the latest AI breakthroughs to the potential of personalization in our daily lives. We'll also be discussing and questioning some of the challenges and opportunities that lie ahead as we continue to enter the future.
We hope this episode will be a great listen and perhaps spark even more questions as we continue to explore the intersection of technology and personal growth.
Before we begin, you can find the full episode here:
Now, let's get started with your questions.
Ash: We created Fulcra to take all of the data that your life is creating and bring it under your control so you can actually make it useful for yourself. The quantified self movement has been around for years, but it's been largely disappointing to many people. We believe this is because so much of the data people's lives create, whether via apps, devices, or large platforms, is siloed and out of people's control. By bringing it all together into one place and making it accessible to users and data science tools, we think we are going to empower people to live much more deliberate lives.
Michael: I want to be a cyborg, starting now, before I get cybernetic implants. I'm pretty sure there are ways to cybernetically enhance myself just by using the tools around me better, especially in ways that are cognitively enhancing. Superhuman self-awareness would make me happy. For thousands of years, the go-to method for achieving self-awareness has been radical simplification. But I want to pay attention to a lot of things. So isn't there some way where I can use all of the wearables, all of the artificial intelligence around me to pay attention to things on the periphery of my life while I focus on a few things? I want to achieve the self-awareness of a monk while engaging with the world like a bon vivant.
Michael: The most exciting AI news for me this summer is the release from a collaboration between Harvard's Lichtman Lab and Google. They've produced the greatest map of brain tissue ever made, and the whole dataset has been released open source. It's an extraordinary map of the connections within a cubic millimeter of brain tissue, containing about 60,000 cells and 150 million synapses.
What's particularly exciting is that this breakthrough applies cutting-edge AI vision technology, specifically a segmentation model, to understand the brain's structure. The Google team used their segmentation model to pick out every structurally independent element in this section of brain tissue. This allows us to understand how neurons and dendrite structures relate to each other.
It's fascinating to see AI breakthroughs being applied to understanding the human brain. I wouldn't be surprised if understanding the depth and breadth of that network gives us ideas about how evolution has improved deep neural nets, which might lead to lessons that go back into AI development.
Michael: I'm really surprised by the one-shot capabilities of SAM2 (Segment Anything Model 2) from Meta, both with images and video. This capability, which allows the model to recognize and segment novel objects in a single attempt, is something I was explicitly pessimistic about ten years ago. It's a capability that small children have naturally, but I thought we were far from AI systems achieving it.
The practical applications of this are profound. We're seeing this extraordinary democratization of technology. I wonder if we're going to see an explosion in innovation, similar to what happened after the human genome was sequenced. How will computationally complex problems now become tractable because there's an amazing high-performance open-source segmentation model available? I'm really optimistic about that.
Michael: I think the short answer is yes. It turns out that getting our current set of frontier models to come up with good multi-step processes is still really, really hard. We can be impressed with 95% reasoning accuracy when asking for one step of a procedure. But when you need to ask for something that is a 10-step process, where each step is going to be 95% accurate, that 5% defect rate compounds. That means the end result is often unusable.
Ash: I think all of these businesses have the same fundamental problem. They need to continue to exist and get users, but as Michael alluded to, the models just aren't good enough yet. They're not good enough at reaching out and executing tasks or repeatably doing things reliably. But I think we're both confident that they're going to get there. So everybody's kind of skating to where the puck is going to be, making promises today based somewhat on what they think will be possible in the future. The question is, how long can you do that before users get frustrated?
I don't think any of them are going to be great personal assistants until they can become anticipatory. Right now, you can tell several of the AI personal assistants to make a phone call for you or send somebody a message, but it's almost the same amount of effort as doing it yourself. Until the tools start figuring out what you need to do based on the contextual data of your life and reaching out and offering to do things for you or just doing things for you, which is a little scary, they're not going to be truly useful.
Amna: We realized that we were having some pretty interesting conversations amongst ourselves and felt like we wanted to share them with the world. The cool thing about everyone at Fulcra is that we all have such unique, different backgrounds and different ways of approaching life and the world. The common thread is tied to innovation, making sure to make the world a better place, and really focusing on how we can make ourselves more present and understand ourselves better.
We thought it would be very interesting to create this content series with Michael and me as the hosts because Michael has a cybersecurity background, a very technical software background. My background is more in hardware materials and biomedical engineering, and I have a passion for fitness, wellness, and nutrition. Combining that and talking to different hosts could bring out a very unique and interesting conversation and perspective on everything.
Amna: Personalization has great potential. Up until now, we have been doing a lot of work with data sets that are all-encompassing of all people. Now, many things are moving towards personalization, whether it's precision nutrition or personalized workouts. We're realizing that every individual has specific needs beyond just a universal baseline.
There are multiple different industries and ways that personalization can be applied. In our case, as it relates to our interests, personalization is crucial for personal assistants because each of our lifestyles is so different. Two people can be CEOs but live completely different lifestyles if one is single and one has four kids.
Similarly, in medicine, many doctors are learning that they can't just always rely on what their patients tell them. It's really important to have information about how you're sleeping and why, what you're actually putting into your body and eating, how your body reacts to these things, if you're working out, and how the workout is affecting your body. Having that information helps with personalization, whether it's with a doctor, a coach, or a personal trainer.
Ash: I've got three different wearables that will tell me if I didn't sleep well enough last night or maybe suggest what time I should go to bed. But it wasn't until I took all of my personal data and piped it through Fulcra into a computational notebook that I shared with a longevity doctor who could look deeply at my diet, my sleep, and a whole bunch of other stuff on a highly personal level that I was able to actually get the information I needed to change my diet to sleep more. And that's the type of highly personal result that I think is becoming possible.
Amna: I think about the recent OpenAI and Condé Nast partnership. I find it very interesting that this is happening. I wonder if it's going to lead to these models being trained by sensationalized news, which is then going to affect how they produce results for us. I'm very interested to see how that affects us as humanity and our mental health because always knowing about everything going wrong isn't always a great thing.
Ash: What I find so striking about the media deals that are happening with large language model companies is how cheap they are. A lot of these companies have had their data scraped for years and have been very unhappy about it to train large language models. And now they are doing deals to monetize that data and try to capture some revenue related to it. But I don't feel like they're getting paid enough for what they're providing in general.
Ash: I wonder how sustainable that business model is for anybody. The LLMs need the training data. The question is, will the companies that are providing the training data, now that they are either taking legal action or negotiating agreements to monetize their training data that they're providing to LLM companies, will their businesses be sustainable if they're not getting paid enough for the data? And will there be data to train the models in the future?
Amna: It is very interesting to see a lot of media companies have been struggling and they've been looking for ways to increase revenue. But I hope that they're able to survive because a lot of them provide great information.
We hope this episode has given you a deeper understanding of the driving force behind Fulcra and our mission to empower individuals through their personal data. It was a blast diving into your questions and sharing some of our own experiences and insights along the way. You, our listeners, are the reason we do this work. Your curiosity, experiences, and ideas are what propel us forward and shape the future of Fulcra. This platform is for you, and it's shaped by you. So let's keep this conversation going.
Stay in the loop with exclusive content, product updates, and news: