Leveraging AI

156 | AI impact on jobs, AI Agents, And US AI infrastructure are shaping our future, and many more AI news for the week ending on January 17

• Isar Meitis • Season 1 • Episode 156

Share episode

Are You Ready for the AI Revolution That’s Reshaping Business and Jobs?

In this episode of Leveraging AI, Isar Meitis takes you on a journey through the latest in AI innovation, from job market disruptions to groundbreaking AI infrastructure investments in the U.S. and the explosive potential of AI agents in 2025. Packed with practical insights and real-world examples, this episode is a must-listen for any business leader navigating the rapidly evolving AI landscape.

Here’s what you’ll learn in this episode:

  • AI's Impact on Jobs: How companies like Meta and Microsoft are leveraging AI to reshape their workforce—and what it means for job creation and elimination.
  • The Rise of AI Agents: Why 2025 is poised to be the year of AI agents and how they’ll revolutionize business operations.
  • U.S. AI Infrastructure Investments: The ambitious economic blueprint designed to keep the U.S. at the forefront of AI innovation—and the challenges it faces.
  • Soft Skills vs. Tech Skills: The increasing importance of leadership, empathy, and adaptability in an AI-dominated workplace.
  • Rapid AI Developments: Highlights on cutting-edge releases from OpenAI, Google, and Microsoft shaping the future of generative AI and robotics.

Take Action Now!

  • Enroll in the AI Business Transformation Course starting February 17th for a step-by-step guide to implement AI in your business. Use promo code LEVERAGINGAI for $100 off.
  • Don’t forget to leave a 5-star review and share this episode with fellow leaders who need to hear this!

Join us on Tuesday for a deep dive into a game-changing AI use case you can implement immediately.

About Leveraging AI

If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!

Hello and welcome to a weekend news episode of the Leveraging AI podcast, the podcast that shares practical, ethical ways to leverage AI to improve efficiency, grow your business and advance your career. This is Isar Meitis, your host, and we're going to have three major topics today and then a lot of small rapid fire topics. So three major topics are going to be job losses or the impact on jobs of AI in the next few years. We're also going to talk about the AI infrastructure investments in the US and what's expected in the next few years. And we're going to talk about agents and their impact on the world and what's expected in 2025. And then we'll end with a lot of rapid fire news about the major companies in the field, including some interesting releases. So let's get this started. As I mentioned, we're going to start by talking about jobs and how AI is impacting and is going to impact them. And the first piece of news is that Meta and Microsoft both have announced job cuts that are taking place immediately. Meta is cutting 5 percent of its global workforce. That's 3, 600. Jobs, which is a lot. Again, it's only 5%, but it's still 3, 600 jobs. And Microsoft announced a 1 percent reduction in its global workforce, which is something that is a continuing downsizing of Microsoft. They have had over 10, 000 layoffs in 2023 and more than 2000 in 2024. Now this comes when both companies are on the rise, You usually see job cuts when companies are not doing well, but it's actually both companies are doing extremely well. Strong market positions are growing every month across multiple sectors of the stuff that they're doing. And they've announced these cuts in the very first two weeks of the year. Now let's look at a quote from Mark Zuckerberg in an interview with Joe Rogan just last week. And I'm quoting, I think this year, probably in 2025, we at Meta, as well as other companies that are basically working on this are going to have an AI that can effectively be sort of mid level engineer. That you have at your company that can write code. In the beginning, it will be really expensive to run, but then you can get it to be more efficient. And then over time, we'll get to the point where a lot of the code in our apps and including the AI that we generate is actually going to be built by AI engineers instead of people engineers. So that gives you a hint of what's actually happening. All these companies are developing better and better code generators. And as software companies, a huge amount of their employees are writing code. And if now a lot of the code or some of the code can be generated by people, effectively, you don't need as many people. I just did a demo this week at NASDAQ, which was by the way, really cool to a publicly traded company about AI. One of the things I showed them is my ability to write code, to create dashboards for data they don't have dashboards for, I don't know how to write code. I never learned that ever in my life, but with the use of AI tools, I was able to build a platform for them where they can drag a CSV file is a direct export from their different systems that generates an amazing dashboard for them in seconds. Now, if I can do that, think about what an actual software engineer can do, which means all the same amount of code, or even a growing amount of code, you need less software engineers. So I think that a lot of these cuts, especially when the company's growing comes from the basic fact that they can do more with less software engineers. Now, early this week, the world economic forum has released their future of jobs reports for 2025. This is a huge undertaking. They have interviewed over a thousand companies that represent over 14 million workers worldwide. Now, there's a lot of interesting findings in this report. First of all, they anticipate that AI will generate 170 million new jobs between now and 2030, and that it will eliminate 92 million jobs with now and 2030, which gives us a net positive of 78 million jobs, which sounds like great news. The people surveyed also 86 percent of the companies surveyed expect AI to transform their operations by 2030, making a very significant difference compared to how their operations is right, is working right now. To put some numbers behind this, 40 percent of companies plan workforce reductions due to AI and automation. 66 percent intend to hire talent with specific AI skills. 77 percent will implement retaining programs for AI collaboration. 70 percent plan to hire AI tool design specialists. And the top skills that these people are claiming they're going to hire for is AI and big data expertise, network and cybersecurity And they're going to invest in technological literacy. The jobs they're seeing at the highest risk are service clerks, executive secretaries, payroll staff, graphic designers, legal secretaries, and basically positions that do regular daily routinely. tasks, which AI will be able to do pretty effectively, more or less starting now. In addition to the AI related technical skills that are going to be in demand, there is a clear growing importance for Soft skills that are highlighting the human importance, things like flexibility and agility, leadership, empathy, lifelong learning, and all of these are going to play a bigger role in your success in a world where the day to day tasks are done by AI. Now they have a specific section that talks about barriers to AI adoption. And based again on these thousands of individuals surveyed in over a thousand companies, the number one barrier is lack of skills to support the adoption. About 50 percent of the people mentioned that as the biggest reason. The next one was lack of vision among managers and leaders. And the third is high costs of available AI products and services. Which explains why 87 percent of leaders in these companies mentioned that they're going to invest in re skilling and up skilling of employees to get them to learn AI skills and be able to benefit from this transformation. Now, if you've been listening to this podcast, you know, this is what I do for a living. And you know that I've been saying this for a very long time. The two most. impactful aspects that will get you to a successful AI implementation in your company. One is training and education for your team. And that means every one in your company, that means the front employees, but it also means middle management, Apple management, C suite, and the board. Each and every one of them has to be trained different kinds of training. As I mentioned, it just came from Nasdaq. I did training for the leadership team of a publicly traded company and their board. That's a very different kind of training then the one I would perform to frontline employees, but they all need to be trained. And speaking of training of your employees and yourself, we just announced another cohort of the AI business transformation course. We have been teaching this course since April of 2023, thousands of business leaders have taken the course and are transforming their businesses with AI based on what they learn in the course. The course is comprised of four sessions of two hours each. So within a month, you'll be in a completely different place than you are today. If you decide to take the course, the course has everything from introduction to hands on experimentation across multiple aspects of AI and use cases, all the way to how to build strategy around AI in your business, regardless of what industry you're in. If it's something you're interested in, there's going to be a link in the show notes. And using the promo code, LEVERAGINGAI, you can get a hundred dollars off the course. And now back to the report. My biggest problem with this report Is that it ignores the fact that the people that are interviewed are leaders in large companies. the smallest company they interviewed had 500 employees. Many of them had tens of thousands or hundreds of thousands of employees. The problem with that is that most of these people do not have in depth knowledge of what is actually happening in the AR world and how that may impact their businesses. So this report is based on a survey and they're surveying people who are experts on the topic that they're being asked about. So when they're being asked about general jobs and their companies and what they see, The answers are probably really accurate because they've been doing this for a while. When it comes to the impact of AI, I actually think that they do not have enough knowledge to answer the questions properly. As an example, the report talks a lot about white collar jobs and the impact of them. And very little, it talks about blue collar jobs, even though there's mentions of it every now and then in the report. If you are following what's happening in the humanoid robots field right now, you know that by 2030, we're going to have millions or maybe tens of millions of these robots already deployed doing blue collar jobs, which means many of these jobs are at risk as well. And there are many other examples in these answers that hint to the fact that many of these people do not really know how advanced AI already is. And they definitely do not know what the big labs are already planning and working on, which means they cannot really analyze how it's going to impact their businesses. That is if anybody can even analyze what's going to happen in the business world. But since we started talking about robots, openAI is making an interesting move in this field, and they've opened many job listings for robotics specialists, meaning they're planning their own general operations. Purpose robots with custom sensor suites that are going to run on software developed by open AI and open AI had a department like this early in their journey that they shot off several years ago in order to focus on AGI and the development of the AI software, and the job by reviving that division that based on the job posting Are planning for, and I'm quoting high volume, 1, 000, 000 plus production runs, meaning they're really going all in on manufacturing robots, which they might do in partnership with somebody or develop their own in house capabilities, but they're definitely planning on developing robots. They're talking about the development of adaptive, versatile robots, capable of human like intelligence with custom sensor suites and integration with open AI, proprietary AI models, and they're planning multiple robots form factors. So not just humanoid robots like the big craze right now, but probably other types of robots for specific tasks. Staying on this robotic topic, that again is going to impact jobs in the near future. Elon Musk made an announcement on X saying that Tesla plans to manufacture between half a million to a million Optimus robots. Humanoid robots in 2027, following an initial production run of 50, 000 to a hundred thousand units in 2026. Elon before was quoted several times saying that humanoid robots are going to be the biggest product in history. And he envisions that the global ratio of robots to human is going to be five to one, suggesting that in a few years, there's going to be 30 billion robots in this world. Now, Elon is known for making really big predictions that do not happen on the time that he assumes they're going to happen. And B Eventually actually happened. So if you look at all of his predictions so far, he was always exaggerating in how quickly we're going to get to something, but we eventually got there. That is true in space. Tesla. Car production that is true in self driving cars. That is true on many different predictions that is made. And so I will take him seriously. If you look at different analysts and how they're reviewing this process, bank of America expects Tesla to deploy a thousand Optimus robots by the end of 2025. Again, Elon expects. Five to 10 times that number. And then 50 to a hundred times that number in 2026, Deutsche bank projects 10 billion in annual revenue by 2035 to Tesla by selling 200, 000 units at 50, 000 each. Again, I actually think the number is going to be significantly higher than I think the cost is going to be significantly lower, but time will tell who is right. But what is very obvious and combine that with many other companies that are now developing multiple variants of different form factor of robots, that robots are coming and they're coming fast and they're coming 2030 to multiple jobs that are now blue collar. That is not in that report and it's going to have a profound impact on the global job market. Now, the third topic that we're going to talk about, as I mentioned earlier, as I mentioned in the opening, is Agents. And I said that multiple times, 2025 is going to be the year of agents. We're going to see AI agents everywhere. First of all, everything's going to be called agents, whether it's real agents or not, well, according to Anthropx chief scientist, Jared Kaplan, AI agents are poised to significant advancements in 2025. And he mentioned four specific categories. One is enhanced tool utilization. the ability to go beyond simple tasks to more complex tasks, because they'll be able to access and use multiple tools that we use. If you think about Anthropx computer usage platform that they released a few months ago, it was. a toy and it's not perfect and it's not working, but that's the direction that it's going. It's the ability to control our computer. If it can control our computer, well, it can do everything that we can do online in a digital environment. We will obviously have to figure out how much do we trust them? these solutions will have to have some kind of way for us to limit what these agents can and cannot do, but that's the direction that it's going. So the first one is enhanced tool utilization. The second one is improved contextual understanding. So the ability To learn and understand what our company does, what the organization does, what are our goals and so on, just by looking at the data in the organization. So that can be, documents. This could be emails. This could be Slack channels. This could be down to the user level or to a group level or to a department level or to the entire company and be able to provide better results because these agents will understand the context. The third thing is advanced coding assistant, going from writing simple, short pieces of code to an evolution that will write sophisticated software, including the ability to debug it. analyze and deploy it without human intervention or with very little human intervention. I mentioned before, I started doing this and I have no clue what I'm doing. And I'm able to create simple software that does things that are very useful for businesses right now. This is what I'm doing. Again, expected to evolve dramatically in 2025 and robust safety measures. there needs to be something controlling these agents, but the biggest fear, by the way, that the leading labs have both open AI and Anthropic and probably Google as well is preventing prompt injection attacks That can take control of an agent from doing what it's supposed to do to doing something else. This could happen a lot easier when it's an agent that gets access to multiple things, because the injection can come from a malicious website that pretends to be something that it's not. And then taking over and changing the direction of the agent to do things that it's not meant to do. And that's one of the reasons why they haven't released these agenda capabilities to the broad public yet. Staying on the topic of agents, Cohere just launched their enterprise platform called north, which claims to enable the development of AI agents for enterprises. And they're claiming that it can reduce the completion time by over 500%. So five X less work to complete the same tasks. They're introducing autonomous AI agents that can execute complex, Business tasks independently across the entire company. Now, those of you don't don't know, go here. We talked about them several times in the past, go here, started with the big labs early on, but they don't have the same bandwidth or the same amount of finances. So they niche down focusing specifically. On enterprises and how they can use AI in a safe way with their data internally. And this is just an evolution of that. So they now just announced what all the other big players have announced is the ability to develop and deploy agents within the organization. And again, because Cohere runs within the company's closed environment, not sending the data outside, it can handle sensitive organizational data that. You may not want to share with OpenAI through their API. And the third big topic for today is infrastructure. So OpenAI just unveiled the economic blueprint for U. S. AI leadership and development. It's a document that is aimed directly to the new U. S. government in order to push them to invest in AI infrastructure, to keep the U S in the lead. And Sam Altman is scheduled to launch the initiative in DC on January 30th, the key. Objectives of this thing is to attract 175 billion in global AI investment funds to the U S counter potential Chinese influence in global AI development, create nationwide standards on how to handle AI instead of state by state regulation and focus on four critical areas, which we discussed many times in the past on this show, Chips, data, energy, and talent, right? So the combination of these things is what's going to drive success in AI. And right now there are many bottlenecks on all of these aspects. They're also advocating for balanced regulation that will promote innovation on one hand, and will ensure safety on the other hand. And they're suggesting the core principles for these regulations should be Pre market competition to drive innovation, developer and freedom to provide freedom to develop different systems still with clear safety standards, prevention of government AI misuse, which is definitely a big scare from my perspective, democratic values based AI development. So staying ahead to make sure that the democratic countries of the world have a lead versus countries like China and Russia and equitable distribution of AI benefits. This is definitely a very big problem, especially if we combine it with everything that we said before, where agents are going to be out there. Robots are going to be out there. AI is going to improve, and it's going to take more and more jobs. And how does that impact the economy? How does that impact society? That's Maybe my biggest fear in the next few years, and I don't think anybody has a good solution right now, or even a direction on how to solve this, but we will have to figure this out in the next few years. And I'm glad that this is being raised to the government early in 2025. Now, we all know that president elect Trump is very much pro the growth of the U S. So I think this will align beautifully with the vision that he has for America. And to start that with the right foot forward, he just announced a major foreign investment deal worth 20 billion with UAE billionaire, Hussein Sajwani and his funds. And that investment is going to go to construct data centers across the U S the initial eight States are Texas, Arizona, Oklahoma, Louisiana, Ohio, Illinois, Michigan, and Indiana. But there's a potential doubling of that amount with another 20 billion commitment, depending on how the first stage goes. And it's part of a broader investment push by the Trump administration to bring foreign investments into this topic. This follows December's announcement of SoftBank to invest a hundred billion dollars in the U. S. in data centers. So it's definitely moving in that direction. The fact that the money is coming in is great, but there are a lot of other questions that are still open, such as Who's going to build them? Where are the chips going to come from? Can we have enough chips? Can they be made in the U S or are we dependent on Taiwan and other places to manufacture them? Do we have the power capacity in order to power all these data centers? What does that do To the cost of electricity for everybody else. What does that do to global warming? if we continue using fossil fuels and emissions to drive these data centers, what does that do to water? Because there needs to be a lot of cooling. There's a lot of questions that are unanswered. Yet, but I think it's a step in the right direction. And I really hope that at least some of that money will go to solve some of those other problems and not just to buy chips and build data centers. Now, the interesting part of this going back to our job conversation is that this will create substantial job opportunities. In the States where these things are going to be built, because each of these data centers will require construction and will require maintenance and will require power and will require a lot of other things around it. And so I definitely anticipate that these will be centers that will attract a lot of people in a wide variety of jobs that will be required for that. And if this is something that is planned for the next decade, we have a decade of new jobs being created alongside these projects across multiple States. Continuing on the same line, Microsoft has unveiled a plan to invest a record of 80 billion in in AI cloud infrastructure in the 2025 fiscal year. Now, about 50 percent of that is supposed to go to the U S and the other 50 percent is supposed to go to additional 13 countries around the world with a network expansion that will reach 40 countries around the world in total. And they're obviously not the only company that is doing this. So, Amazon AWS and Google cloud are investing a total of 180 billion in data centers in 2024. So a huge amount of money is going into infrastructure to support AI. This investment that we talked about so far is just for the data centers themselves. There's the entire ecosystem around them that needs to be there in order to enable it to run, such as cooling and power and other infrastructure and talent and so on to run these things, which will drive significantly more investments, which is not necessarily a bad thing. And hopefully some of this investment will go to helping us find a way to counter the negative environmental impact of this development. Now, an interesting side conversation to this is that Microsoft are temporarily suspending the construction of its 3. 3 billion data center in Wisconsin. that was supposed to support open AI supercomputer development. That is obviously interesting because it comes a few days after they announced this 80 billion investment in data centers, and they cite that they need to evaluate the scope of recent changes in technology, this. May hint to some more troubling waters with their relationship with open AI, or maybe not. They didn't exactly say that, but the first phase of that development is still on track to be deployed in 2026. and then the following phases are now being questioned and reviewed. The interesting thing about this whole thing and the timelines that they're talking about, about developing a data center in two years and the big investments S. is, if you remember, we talked about this, late last year, that Colossus, the supercomputer that X has built, Went from an empty building without any of the necessary power generators, transformers, or anything else to a working AI supercomputer in 122 days, that is magical. Nobody believed they can do it, but they did meaning it's doable. Meaning with the right infrastructure and the right process and the right money, some magic can be done in the U S In order to generate data centers much faster than most of the plans look at today. So what does these three things tell us? The infrastructure investment, the agents, growth, and their capabilities and the job implication. It tells us that we're going to see a dramatic shift in almost every industry because of this. AI is going to grow. It's going to grow faster than we know today. Agents will be able to do things that are way beyond anything we've seen AI do today because they'll be able to understand context and work within our environments and be part of the employees of a company. So we're going to have human employees, we're going to have digital employees, and we're going to work jointly in order to achieve goals. And the ratio between the work done by digital employees to the work done by human employees is going to shift over time to more and more digital workers and less and less human employees. At least, that's what I think. I know the report from the world economic forum thinks otherwise. And again, I have no doubt that there are a lot of really smart people over there. I think the people they surveyed may not really understand in depth what I can do today and definitely where it's going in the next two to five years. Only time will tell where this is going. I really, really hope that they are right. And I am wrong. Either way, the key factor is training, train yourself, train your people, train your team, train your company on how to use these tools, how to approach them, and how to be very agile in understanding how these tools work and how to implement them. Because that is definitely going to be a key differentiator between companies in the next few years to come. With that. Let's switch to our rapid fire topics. We will start with open AI. So open AI just introduced tasks. It's a really cool feature that now is built into chat GPT. so if you go to the dropdown menu, where you can select different types of models, one of the options is chatty, with tasks that allows you to schedule. Tasks with Chachi pt, and this could be anything from research to reminders to, summarizing specific data from specific sources in whatever cadence you want. So this could be helping you with your personal morning routine or giving you industry news once a week Well summarized based on whatever information sources you're gonna ask you to do, in whatever format you want the output to be. It will give you notifications on the browser when it's completing the task. It actually works pretty well. I've tried it on several different things so far, and I'm very happy with the results. I'm using it right now to keep myself updated with AI News. the output that you're getting every single week is going to benefit from me being able to do this a little better. faster and more efficient, but you can do this for anything that you want. And like I said, it's available right now to any plus member teams and pro users. Another interesting piece of news from open AI is that they're enabling people to register, to use Chachapiti over the phone. So without actually going online and signing up, I must admit personally, I don't understand that, but I guess there are places around the world where this makes sense. So people will be able to sign up for Chachapiti and use it just with a phone call without actually giving access to their email address. they'll obviously be limited with what they can do with Chachapiti without the email, such as not being able to upgrade to the paid plans. Maybe that's a way from OpenAI to open the doors to a bigger audience. Another interesting piece of news from OpenAI that is again, just important to know, something to think about is that, People started seeing that OpenAI 01 exhibits unexpected behavior where it's spontaneously switches to Chinese, Persians, and other languages in the middle of its reasoning process, while the input and the output still happens in English. Now, nobody still can exactly explain why this is happening. There's been A few theories to try to explain that Google DeepMind researcher, Ted Shio suggests that it may be due to the influence of Chinese data labeling services that is used as part of the training that does not explain some of the other languages that it has shown to be used in its thinking process, a professor at the University of Alberta, Argues that the model simply doesn't distinguish between these languages. It just doesn't know its switching languages. it just finds the most effective tokens in different steps of the process. And so it doesn't really know it's switching languages. It's just doing the thing that it's trying to do in the most effective way a hugging face engineer suggests that it could be similar to how bilingual humans switch between languages in their head as they think about different problems. I know that as a bilingual speaker, that there's things I think easier in English and things that I think easier in Hebrew, and it just depends on the task. I don't actually think about it. It just happens, but I guess that's the suggestion, here as well. But the bottom line is. That we don't fully know how these models work and how they're going to behave and how the training process is going to impact how they're going to execute things. And it's important to know that because everything we talked about before on the impact of the world, we don't a hundred percent control on how it's going to be because this is not traditional software. And so I like this example because it will hopefully get you to think about the broader implications that this may have on our ability to control where this extremely powerful and extremely fast revolution is going. Switching from open AI to their closest partner, Microsoft. Microsoft just announced a significant organizational restructure by creating a new. Department called core AI platforms and tools. And it is going to be led by Jay Perrick, a former Meta VP and recently the CEO of Lacework, which is an AI safety and security startup. He's going to report directly to Satya Nadella, and the goal is to combine several different departments that has to do with ai, including development and AI platform and product teams into one unified group within Microsoft. Now, the goals that they have said is creating an AI first app stack in Microsoft and transforming Azure into the primary infrastructure for AI in the world, integrating Azure AI Foundry, GitHub, and Visual Studio Code into one environment, developing agents to revolutionize SaaS application categories and implementing service as a software approach to Multiple aspects across the Microsoft infrastructure really unifying everything AI within Microsoft to make Microsoft the obvious global leader in AI, combining a lot of things that are doing today in silos within the organization Still in Microsoft, but now switching to the development of models, or actually training of models, Microsoft just announced that they've developed a new way to train reasoning models, and they're calling it RSTAR Math. In this particular case, it's geared towards math only, and it allows to train small language models to outperform OpenAI 01 Preview in complex mathematical problems. And the way they've achieved this Is by fine tuning existing models like Quen 2. 5, microsoft PHY 3 Mini, and they've achieved this by developing a new training technique for open source models like Microsoft's PHY 3 and Alibaba's Quen models in different sizes. And what they're doing is it allowing it to develop step by step approach by integrating multiple new processes. That then enables to work more accurately on math problems and achieve better results than significantly bigger models on math problem solving, like O1 mini. We talked about this many times on this show that there are two approaches to developing these models. One says bigger is better. We can keep on developing better and better models that will know everything and be able to do everything more. The AGI approach, but there's the. opposite approach that's saying let's develop small, specific models that can do specific tasks very, very well. And we've talked about this in the agent world, it's probably going to be a combination of the two. We're going to have a thinking more like an orchestrating project manager, agent that will delegate tasks to smaller agents that will be very good at specific tasks. And this is just one example on how these smaller agents on this particular case models can work specifically in this case in math. Let's switch from Microsoft to Google. So Sundar Phai, the CEO, just made an announcement that they are planning major releases in the first few months of 2025, following a series of successful launches in 2024. So in 2024, they released Gemini two and Willow and VO2 maybe not to the broad public, but you can at least sign up for the waitlist like I did. And I suggest you do the same. those of you who don't know VO2 is their video generation model. And it's just a waitlist right now that you can sign up for. And it seems incredible with the results that it's generating. But the. Upcoming releases are supposed to be Gemini 2 flash, general availability for developers in January. So in the next few days, 2. 0 experimental advance for paid subscribers, Notebook LM plus for Google one subscribers, Gemini 2 integration with more Google products. So the Gemini that runs within the different applications and AI overviews within more aspects of search. New specific cool features include daily listen in search labs, offering personalized podcast content that happens daily that you can listen to on whatever topics you want. So think about notebook LM that runs as a task on a specific topic that you wanted to research. So you have your own personalized podcast on the topics that you're interested in. I think that it's really cool. Time will tell if it's actually worth something, but I think over time it probably will. And enhanced AI capabilities across all the Google product ecosystem. Another interesting announcement from Google is that AI for workspace is going to be included in the Workspace cost, which is going to go up by 2 a month. So everybody who has a Google workspace account is going to pay 2 more a month or using it, but that will include the major features, including the Gmail email summaries and the other Gmail functionality of Gemini automated spreadsheet and video design tools. Meeting, no taking automation within Google meet notebook, LM research assistant, full access to Gemini chatbot, gross application writing tools, and so on and so forth. Basically everything we know from the Google workspace universe will be included without paying for it separately. This follows a similar move by Microsoft in November of 2024, where a co pilot features became standard for Microsoft 365 subscribers. So not a surprising move by Google. I'm really expecting the point that these two companies will finally deploy something that is worthwhile workspace, meaning allow the various aspects of these workspaces to actually work together. So give me one user interface that will allow me to understand everything that's happening in my organization, based on all the data that I have in all these different sources versus having to work with limited functionality in each and every one of the tools. I have no doubt that's where they're all going. It's probably more complicated than I think it is, but I really can't wait for it to finally happen because I think that will unlock significant efficiencies and usability for anybody who's using these tools. By the way, what it means for you and your organization is you should probably commit to one of these environments. So many organizations that I meet have some people working in Microsoft, some people using Dropbox, some people using Google, and that is not going to be a healthy solution because. Eventually there will come the day, hopefully within 2025, where they will have a unified solution that will allow you to query everything that's happening in your organization and maybe agents that will do it on its own in order to make sense of what they need to do. And if you have your data across multiple solutions, then you're going to lose that capability. Another interesting survey, not coming from a company, but actually coming from MIT's, MIT's technology review for that reveals the top five AI trends that will shape 2025. So one of them, which we talked about a couple of weeks ago on the show, Generative virtual worlds that will become more and more mainstream. So they're saying, following the evolution from generating images in 2023 to generating video in 20 24, 20 25, will see the rise of interactive virtual environments. So basically the ability to create a 3D fully immersive universe generated by ai. I mentioned several companies who are already doing this. So Google deep minds, Jenny too, and companies like world labs are already developing these kinds of platforms. And MIT are claiming that that's going to be functional and going to be a big deal in 2025 advanced AI reasoning capabilities. We talked about this many times before. So models like oh one and all three and all the different companies that are now chasing them are going to become a big part of our lives, which are going to make an even bigger impact on the things that these tools can do scientific discovery acceleration. So alpha fold has won a Nobel prize recognition In its work in science, but they're claiming that 2025 will see a lot more significant research breakthroughs using AI meta and hugging face are leading initiatives in material science through their open source models. And so this by itself will drive more discoveries using AI, which I'm personally very excited about military tech integration surge. So more and more. AI companies are winning and bidding defense contracts to do different things with AI. I find this inevitable and yet really scary because the implications are really profound that some of the decision making process in some military applications will not be done by humans, which I don't think is a good thing. But that will be an arms race that, as I mentioned, is inevitable because we don't want them to have it and we not to have it. So we will keep on developing these tools and they will do exactly the same. And that's going to create this AI arms race within the military world as well. And the last thing that they found that I actually found really interesting is that they're claiming that Nvidia's market dominance is going to face some significant challenges. Because of the investments from other players like Amazon AMD, and even small startups like Grok, that we talked about many times, that are developing alternatives and different architectures to the one Nvidia has. That being said, I talked about Nvidia on the show last week. They are going after Everything, software, hardware, infrastructure, network, development, of different tools and everything you can imagine. So I will definitely do not underestimate NVIDIA despite the growing competition. And I will say something else, that While their share in the pie may shrink, I think the pie is growing at a very, very high rate, meaning Nvidia will still be able to make more money, most likely, but that's something that MIT is looking at as some of the things that can happen in 2025. I want to talk about two interesting releases of new model. The first one is which is Berkeley's University research team has released Nova Sky, sky, T 1 32 B preview. That's the very long mouthful. Name of this new model is an fully open source AI model that matches open AI oh one preview in math and coding capabilities. And it costs them only 450 and 19 hours to train using 8 NVIDIA H100 GPUs. That is not even a minuscule fraction of what it costs OpenAI to train all one. If you remember just last week, I told you about the Chinese company that did something similar. And so it's very interesting to me that it's possible to get these technological breakthroughs by using different processes and different algorithms and potentially different techniques to train these models to achieve specific results on specific things. So it's not going to be better than 01 across the board, but just the fact that with practically zero money, they were able to train a model that can compete with 01 on anything is incredible. No, that because this is open source, everybody's going to look into this to try to figure this out, which tells us that if we do figure this out, this whole gigantic infrastructure investment that is coming down the pipeline that we talked about earlier might be able to yield significantly better results, and maybe not all of it will be required. Meaning some of these funds can go to take care of other stuff. Luma labs has also made. a interesting release this week. They just released Ray 2, which is their new video generation models. It has the capability to generate up to 10 second high resolution video with advanced physics understanding and object interaction. So significantly better than what they had before. And some would argue significantly better than everything else we have today, other than maybe VO2 that is available to selected few that got access to it. From the demos that I've seen online, it looks really impressive. And as I mentioned in previous shows as well, I think these models now are in many cases, indistinguishable from real videos, unless you're really looking for something that will give you a hint that this is not AI generated. And I do think that we're going to start seeing people generate content that people will subscribe to. So mini series and stuff like that, that will be a hundred percent AI generated most likely in 2025. Final reminder, the AI business transformation course starts on February 17th. If you do not have specific detailed plans on how to train yourself or your team on AI, this is an amazing opportunity because we do these courses open to the public and not privately to specific companies only once a quarter. So the next time, of course, like this would open, we'll probably be around April or May. So don't miss this opportunity, sign up. As always, if you're enjoying this podcast, if you're finding it valuable, please give it a five star review on your favorite podcasting platform. Write a review in your words. Tell us what we need to improve. Tell us what you like so I can learn from it as well and give you even better value and share the podcast with a few people that you know can benefit from it. So while you have your app open, Click on the share button and send it to four or five people that you know, that can benefit from this as well. This is your contribution to helping AI literacy for everybody around the world. And we are all going to benefit from that. We will be back with a fascinating episode on Tuesday that will show you how to do something unique with AI, a specific use case in your business. And until then have an amazing rest of your weekend.

People on this episode

Head Shot

Isar Meitis

Host