this video is brought to you by 880,000 hours. Hi, welcome to another episode of Cold Fusion. Before I even utter a syllable, I think we should make something clear. When we say AI, our minds go straight to the often trashy consumer generative AI that's been all over the news. But there's a distinction between that and regular neural networks that are quietly being used in novel ways in the background. I've called artificial intelligence and dichotomy multiple times for this reason. And just a last bit of housekeeping, we're not going to split hairs on semantics here because basically, artificial intelligence in the modern Zid K is synonymous with neural Nets. So the terms will be used interchangeably just for Simplicity. Okay, so there's no question that AI certainly has split opinions among the public. We can take America as an example to quantify this. A 2024 yugov poll breaks it down. The four most common emotions towards AI are caution, concern, skepticism, and curiosity. Interestingly, 14% think it's already more intelligent

than people. And another note: the younger people are, the more positive they're likely to view AI. This includes views on AI's effects on society, their own life, or the economy. Whatever the figures are globally, one quick look online will show you a fact: a lot of people love to hate AI online. All you hear about is scummy companies that lie, that overpromise its capabilities, and underdeliver. And then there's the other end: you also hear about sneaky companies replacing workers with AI. For example, this unfortunate guy: I just lost my job, and I lost it to AI. I've been a graphic designer for the past eight years of my life. On Monday this week, I got told that I'm being made redundant. Turns out basically all of the material that I've provided over the past six years is now being fed to AI and templated. A design that would take me 30 minutes now takes AI 30 seconds. In fact, a little while back, I made an episode looking at the corporate deception when it comes to AI. But that's all the

surface stuff that everyone keeps talking about. What about something new? What if we dug deeper and took a comprehensive look at the flip side? That is, how can AI be used for good? What I mean is this: are there any world-benefiting positive applications of AI that can't be achieved with any other technology? So, in this episode, we'll explore how AI can and is making a difference in many fields, like disability support, medicine, designing better chips, and other important sectors. Keep in mind that some of these projects featured are still in their early stages, so a lot could change in the future. But conceptually, I found these very interesting. We also feature a special segment from Tasmania, Australia. I traveled there to see how AI is being used to help replenish a species that has declined by 95%. So, in this episode, let's dive into the real-world impacts of AI. [Music] You are watching ColdFusion [Music] TV. It's no secret by now that AI is predominantly powered by Nvidia GPUs, especially in the training process. So

much so that it's seen Nvidia's value as a company quite literally skyrocket along with the AI hype. These GPUs rely on powerful silicon to deliver their capabilities, but there's a twist. Now AI is improving the very same chips that they run on. It's an interesting closed-loop cycle. Let's see how it works in chip manufacturing. There's a key step known as lithography. This is where detailed circuit designs are etched into semiconductor wafers. It's achieved with something called a photo mask. Nvidia designed a new GPU-based algorithm along with a new library which accelerates computational lithography by 4 times. Basically, they've developed an algorithm which parallelizes their task among many GPUs. This reduces the amount of time needed for computing a photo mask from several weeks to an overnight job. The main problem was that creating photo masks used to be a slow, arduous task that could take weeks, but now Nvidia has a new platform that consists of generative AI algorithms that can speed up the process. What used to take weeks is now

40 times faster and can be done overnight. There are 89 reticles for the Nvidia H100 running on CPUs. A single reticle currently takes 2 weeks to process. Kitho running on GPUs can process a reticle in a single 8-hour shift. Nvidia's new platform is called Q litho, and it chews through nanoscale computational lithography problems at incredible pace and at incredibly small scales. We're talking 2 nanometers in size. That's like comparing the width of a human hair to the size of a football field. This AI-driven approach isn't all just about speed; it's about making what seemed physically impossible possible. TSMC, the world's largest independent semiconductor foundry, is already using this accelerated process. Other big names in the industry like ASML have joined the effort. Together, these companies are setting the foundation for the next generation of semiconductors, aiming for 2 nanometer technology and beyond. So then, if Nvidia has been using AI to create their own

chips. You'd expect to see a huge bump in performance, right? Well, where is it? Well, take a look at this chart comparing Nvidia's GPU performance to Mo's law. It's hard to know exactly how much the AI system is helping, but AI is making a difference here. So, what's the takeaway? Well, what's fast today is about to get a whole lot faster. AI could take the computing power of tomorrow to levels we've only dreamed of. And as a side, I'm sure most of you know about DLSS. It's Nvidia's AI upscaler that manages to provide higher quality visuals without demanding excessive computational power. Better chip design is something that one might probably expect in terms of contributions of AI, but using AI to help regenerate kelp plants in remote parts of Australia? Well, that's new, even for me. I was invited to travel to Tasmania, Australia, to check out how AI was being used to help restore Australia's giant kelp population, thanks to a project by Google, the Commonwealth Scientific and Industrial Research Organization, or CSIRO.

The Institute for Marine and Antarctic Studies, or IMAS; The Nature Conservancy, or TNC; the Great Southern Reef Foundation; and the Kelp Alliance. That's a lot of collaboration, so something must be going on here. And before we dive into it, we got to understand what's at stake here. So, Australia has a massive kelp forest stretching from midway down the West Coast all the way to the island nation's East Coast. In fact, 70% of Australians live within 50 km, or 31 mi, of a kelp forest. Due to rising water temperatures, a large population of kelp has died off over time. In Tasmania, this number is at a staggering 95%. That's right, only 5% of the original population remains. So, who cares about some undersea plants? Well, kelp is actually important in sustaining thousands of types of sea life. For example, some animals feed off kelp, others make kelp their homes, and other animals feed off those animals, and so on. It's a massive community, and

ecosystem with thousands of types of sea life depending on kelp. It's not hard to imagine the runaway domino effect if the kelp forest goes permanently extinct. To get started on fixing a disaster in the making, mapping out the scale of the problem was essential. Now, doing this by hand is virtually impossible, so how does Google's AI help? Well, Google is using their own engine from Google Earth and a cloud platform called Vertex AI to accelerate the discovery of the remaining kelp forest at superhuman speeds. Google is partnering with the NGIS to quote locate and analyze kelp forest in more than 7,000 square km of satellite imagery for the first time. This helps researchers paint a more complete picture of these ecosystems. We sat down in person with Leia Kaplan, sustainability business leader at in Tasmania, to share more. The giant kelp is interesting because the canopy floats on the surface, which means that we can actually detect it with satellite imagery. The reason that is important is because understanding the

entire geographical span of the giant kelp in Australia actually hasn't been done before. To generate a map of where the kelp is remaining, so that we can understand how restoration efforts are progressing, that's really, really important. It's something we can do at scale with satellite imagery, which is just way more affordable than the old way of doing it, which was basically to fly a plane and take photographs. So, satellite imagery can be automated, the processing and the AI detection of CP can be automated. So, we can actually have ongoing remnant maps once the mapping is carried out. Google, the CSIRO, and IMAS can use AI to do what it does best: pattern recognition and prediction. Using the 5% kelp that has survived the higher temperature range, AI analysis can uncover the genetic patterns that enable this kind of kelp to survive. Once that's figured out, this special kind of kelp can be bred and then used to replenish the dying population with a new breed of

heat-resistant kelp. We had a tour of the CSO and IMASS labs to see parts of this in action, to get more insight into the genetic profiling. I had a chat with Andrew Carroll, product manager at Google Research, to give some further details. So, one component of any sort of study is to understand what amount of genetic diversity is present in the species, just to understand where the starting point is. The second set is to understand what are the traits that give resilience to various challenges, and that the key challenge here is thermal tolerance. It's the ability for kelp to withstand an increase in ocean temperatures. This research project wouldn't just make a difference in Australia, but could have a positive impact globally. Google states that they're making the geospatial map accessible to all scientists, and the AI tools used are open source. Perhaps one of the most incredible ways that AI could make a difference is potentially giving mobility back to those who have lost it, introducing AI-powered prosthetics. Yeah, so it's a full robotic arm.

like there's an AI chip in here, and then this like attaches to like the remainder of my arm. There's uh sensors and stuff in here that reads the signals and nerves in my arm that the surgeons REM. So like me thinking about opening, closing, like I'm using the same nerves and stuff. He evidently looks quite happy, and you can feel his joy. Let me tell you another story. This is Sarah Deard. Two years ago, she lost her arm in a subway accident at age 45. She's gotten a new AI-assisted bionic arm. It trains itself and gets better the more she uses it. But the thing is, it's heavy and has to be charged every day. But still, the important thing is that it's given her the ability to do everyday tasks that were once impossible. Simple things like making a coffee, straightening her hair, and hugging her daughter. The New York Times did a piece on her earlier this year, and they put it all perfectly. Quote, AI is seeping into further fields like healthcare while many researchers have

raised alarms about AI's risks. Other experts said that those concerns must be weighed against the technology's potential to improve lives. End quote. And that is the central theme of this episode: AI can do good. So, with all seen prosthetics before, but how do these AI-powered ones work? Neural nets and machine learning are used in these prosthetics to interpret electric nerve signals from the patient muscles. It allows for more precise and intuitive control. AI basically gives the prosthetic a brain. Imagine a prosthetic leg that knows that you're climbing a stair or a hand that can grab a cup smoothly without spilling. That's neural networks in action. It can use sensors within the limb to gather information from the external environment. Sensors read the muscle signals, then move the limb, and it can be smart and adapt. Several companies are using AI to create smarter prosthetic limbs. Autobook offers a bionic hand which uses AI for improved functionality and control. Rewok is known for their powered walking assistant system. Last year, they revealed their

plan to use AI for autonomous decision-making in the new exoskeleton prototype. How long had it been since you took a step? Almost 20 years. 20 years. When you stand up and you hadn't stood up in years, that was something that um, that was the best part of it. Other examples include Osor, an Icelandic firm that makes bionic prosthetics for lower limb users. Actively powered technology allows powered knee users to expand their activities, including those that are still challenging for users of passive knees. An atom limbs, which combines advanced senses and machine learning for more accurate and realistic movement in prosthetic arms. We basically cover your stump in electrodes, and you train it up with your phantom limb that we can't see, but you can flex your finger, extend your finger, rotate your wrist, rotate it the other way, and our machine learning and AI system takes a huge amount of training and spits out a result so that now you, whenever you think, can just move. So there's nothing in your head, nothing on your head, nothing in your arm, it's all this.

surface electrode and AI that basically powers it. According to an estimate, there's over 550 million amputees worldwide. There's 2 million in the US alone, and that number is expected to double by 2050. So, just take a second to think about how many people's lives could be changed by AI and advanced prosthetics. But there's a catch. Even though there's huge potential, this technology is not quite ready for prime time for people who need it most. One big reason is, as you've probably already guessed, the cost. Building a robotic system that can mimic all the movement is quite challenging. To add to this, recreating sensation is probably still decades away at this point, if it's even possible at all. But even with what's available right now, it's no doubt changing lives for those who have access to it. Over time, as the technology improves and manufacturing becomes cheaper, access will increase, and that is a great thing. Prosthetics are just the start, but in the field of healthcare, neural networks are showing potential. It's

augmenting the way we diagnose, treat, and manage health on a global scale. This graph extensively lists the areas in which healthcare is already having an impact. Sectors such as medical equipment prep, medical assistance, and a variety of technicians are the ones being most affected. The implications for healthcare are enormous, and most doctors right now are sleepwalking. They do not understand how artificial intelligence is going to completely change the practice of medicine. I'm a doctor myself. Once more, this is not an attack against any one doctor, but I believe the medical profession has completely sold out, and they totally deserve what is about to come to them over the next few years. And if any doctor out there thinks they can shield themselves from these changes, well, good luck to you because you are absolutely not going to be able to do that. Big change ahead. Better brace yourself for that. This next section is bypassing that tide conversation of AI replacing doctors and other healthcare workers. That's not what I'm going to be talking about. This is more about AI

helping with the heavy lifting in many ways. Google's Med LM aims to do just that. Med LM is a set of AI tools created to do things like quickly write up patient notes and help find important markers and diseases. The idea is to make healthcare tasks easier and more efficient. It's not just Google, though. Nvidia has rolled out some AI tools to make surgeries and medical imaging smarter and faster. Titans like Johnson and Johnson and GE Healthcare are on board with Nvidia's AI tools. Looking to the field of medicine, our Ural from EY Americas has seen a positive shift in the bio-pharma industry. They note that AI has gone from a meaningless buzzword to a real game changer in just a year and a half. What do they mean by this? Typically, finding new drugs is a lengthy journey filled with research, clinical trials, and testing. This process can cost pharmaceutical companies billions of dollars and all with a high risk of failure. Now, with the help of AI, they can cut down on time and cost, and that's a

big deal. We can point to a few biotech companies that are already using AI and deep learning to boost their research. For example, Recursion Pharmaceuticals is using NVIDIA's AI supercomputer BioH 2 to find new medicines faster. Benevolent AI checks drugs that are almost ready for testing. They want to treat diseases that don't have good treatments yet. Atomwise, on the other hand, utilizes its AtomNet deep learning neuron network for binding affinity prediction. This is a crucial step in the drug discovery process. This helps them find effective drugs quickly and more often. Neural networks are speeding things up and making the whole operation more efficient and smarter. Mr. Armchair Skeptic out there might shout, I call BS, but this is not just theoretical. McKenzie published a research study where they found that AI does indeed positively impact productivity and boost success rates in bio-pharma research. This is a hidden movement going on under the noses and away from all the general AI hype. Already, 270 AI-driven drug discovery companies have been

founded, combined with rapid advancements. If corporate greed stays out of it, we could see new medicines with less side effects for a cheaper price. It's a tall ask, but a man can dream. Others in the space, like Delire Diagnostics, have developed a test that uses AI to identify indicators of lung cancer. Similarly, researchers at the Children's National Hospital in Washington have created an AI tool designed to diagnose rheumatic heart disease in children. And in the past year, there's been plenty of these medical solutions coming online. In a previous episode not long ago, we already saw how cancer detection is being made easier through the use of AI. All you need is a simple sample of dried blood. Detection rates can reach up to 85%, and best of all, it's quick and cheap. And it's not just one research paper or one hospital that's used this solution; it's almost prolific. So in science, independent verification is always great, so this is really exciting. In robotic surgery, AI is helping doctors and surgeons with precision. We've covered

the robot surgeon Da Vinci all the way back in 2018. But now it can utilize machine learning, image recognition, and more for advanced image analysis, surgical planning, and real-time adjustments tailored to the patient's unique anatomy. So, in all of this, it's important to note that the training data is critical for healthcare-related solutions. The effectiveness of the data can be cut down due to variables like demographics, age, gender, and environmental factors. So, even though I think these early signs are very promising, we have to be realistic. It will probably take some time before we see a robust, meaningful impact for the average person, but this is still a great start. What if artificial intelligence could help create better batteries? Lithium-ion types of batteries are crucial for everything from mobile devices to electric vehicles, but lithium is very expensive and has a negative impact on its intermediate environment. Replacing lithium is a major challenge that could take years, and it could involve testing millions of alternative materials. AI can be extremely useful in this context, and it's already being used to great effect.

In a notable study, AI was instrumental in sifting through millions of possibilities to pinpoint 23 materials that hold promise for Next Generation batteries. This collaborative effort between Microsoft and the Pacific Northwest National Laboratory has been documented in a recent publication in related research in early 2024. Microsoft's Nathan Baker and his colleagues used AI to sift through 23.6 million candidate materials. It tweaked existing electrolyte designs by replacing some lithium atoms with other elements. The result was a battery that required 70% less lithium than other Current Designs. The Breakthrough material that the AI found was named n216. Experts from the Department of energy later ensured the Practical viability of the material. n216 is essentially a novel electrolyte where half of the lithium atoms were replaced with sodium. This unconventional recipe opens up exciting possibilities for battery physics. Using this material, they were able to whip up a rough low conducting prototype that could turn a light bulb on. But the thing is, from the idea to a working prototype only took 9 months.

massively accelerated time scale, especially for a brand new chemical in battery research. But of course, there are some catches. So, not so fast. There are still some questions over the training data, as the available data in battery engineering is limited, and there could be some unforeseen complications in bringing future AI designs to life in these batteries. But this is no doubt some cool progress. I think it's safe to say that by now, we've been inundated with consumer-facing AI applications, with every single company adding AI as some quote novel perk. Honestly, it's been overdone, and frankly, it's annoying at this point. But as a person who's been covering neural networks on this channel since 2015, before all of this hype, it's good to get the full picture. The stories I just covered are ones that I personally found fascinating. It helps to see where we stand in the grand scheme of things and not to get too tied up in the noise. Sure, I guess some people might find it interesting to generate AI stickers in their messaging app for whatever reason, but to have AI

potentially solve some of the broader problems of humanity and to ultimately help us dream of a better world. As clichéd as it sounds, that's what most of us have always imagined AI to be. And now that it's showing some promise in some of these areas, it's good to see that there's some positive movement in that direction. And that's despite the complete mess on the other side of the coin. So, what do you guys think? Feel free to share your thoughts in the comment section below. Today, we've seen AI helping humanity, but what if you wanted a career to do the same? Actually, have you ever thought about how long a career is on average? It's about 80,000 hours of your life, so it's worth spending some time to plan it. And this is where the nonprofit 80,000 Hours can help. After 10 years of conducting research with academics at Oxford University, they've come up with some of the most fulfilling career paths—careers to help solve some of the world's most pressing problems. 80,000

hours has a job board that's constantly updated with hundreds of active job openings that they think might help you have a high impact. You can filter the job by location, role type, job requirements, and what problem area they work on. I want to highlight that everything that they provide is free. They're a nonprofit. Their only aim is to help you find a fulfilling, high-impact career. For example, I think the career path of information security in high-impact areas is critically important, and there'll be more demand as time passes. And they've got an entire career review on it. Go to 8000hour.org/coldfusion to be sent a free copy of their in-depth career guide, which helps you learn about what makes for a high-impact career, get new ideas for impactful paths, and make a plan based on what you've learned and put it into action. So, that's 80,000 hour.org/coldfusion. So, thanks for watching, and I hope you learned a thing or two. Hopefully, this has opened your eyes a little bit just to see that AI isn't all

completely bad. Anyway, that's enough for me. My name is Toogo, and you've been watching Cold Fusion. I'll catch you again soon for the next episode. Cheers, guys. Have a good one. [Music] True s for [Music] [Music]

Are we all wrong about AI?

Hi, my name is Dagogo Altraide and I create and narrate all the videos on here.
This channel aims to let you experience the cutting edge of the world around us in a relaxed atmosphere.
Learn the captivating stories about how our world came to be and also learn what's happening in the cutting edge today.

--- New Thinking Book written by Dagogo Altraide ---
Learn the stories of those who invented the things we use everyday. This book is rated the 9th best technology history book of all time by Book Authority!
Audio Book: https://goo.gl/p8auh3
Get the book on Amazon US: http://bit.ly/NewThinkingbook
Amazon India Link: https://goo.gl/qJCC1T
Amazon UK link: https://goo.gl/RBxisf
Ebook on Google Play: http://bit.ly/NewThinkingGooglePlay

MERCH:

INTERNATIONAL: https://store.coldfusioncollective.com/
AUSTRALIA: https://shop.coldfusioncollective.com/

Contact | *Business ONLY* dagogo@coldfusioncollective.com

Join the fun on instagram: 

@ColdFusiontv

Facebook | www.facebook.com/coldfusiontv

Are we all wrong about AI?

Are we all wrong about AI?

스크립트