EP 399: AI News That Matters - November 11th, 2024

This is the Everyday AI show, the Everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business and everyday life.

OpenAI has expanded partnerships with the US government. Salesforce has hired 1000 humans to sell AI. What will new President elect Donald Trump's impact on AIB? And is Microsoft giving up on Copilot Pro? We're going to be answering those questions and a lot more today on Everyday AI. What's going on y'? All? My name is Jordan Wilson and I'm the host of Everyday AI and this thing is for you. It is your daily livestream, podcast and free daily newsletter helping us all learn and leverage generative AI to grow our companies and our careers. So if that sounds like you, you are in the right place. If you're listening on the podcast, please make sure as always to check out your show notes for more information. Most importantly, a link to our website. Please if you haven't already, go to your everydayai.com and sign up for that free daily newsletter. Each and every day we recap the day's episode as well as give you all of the information that you need Happening in the world of AI because there is a ton of speaking of a ton happening in the world of AI. We do the AI News that Matters show almost every single Monday like today. Even though it's technically as I record this, I'm in the future, I'm wrapping up a nice little 10 day trip for celebrating my 10 year anniversary with my wife. So super excited for all of you that continue to, you know, tune in and left me some nice comments. I very much appreciate that. But I will be back in Chicago in like 36 hours. And speaking of Chicago, still getting this piece figured out but should be at the Microsoft conference. So if you're going to be at the Microsoft conference next week, I believe it kicks off November 19th here in Chicago. Do let me know because yeah, I believe we'll, I'll be attending at least may even be broadcasting live, so make sure to pay attention for updates on that. All right, enough chit chat y'. All. Let's get into the AI news that matters for the week of November 11th. And again, apologies y', all, I'm in a hotel room, I don't have my normal sound equipment, camera, lighting, all that stuff and my I, I got completely burned. Right? So if you're joining the live stream, you're like why, why is, why is your face so red? It's because I have been enjoying some sunshine all right, let's get into it. So the federal government has expanded its use of OpenAI's Chat GPT technology. So according to reports from Fed Scoop, the federal government is increasingly investing in generative AI technology, particularly through partnerships with OpenAI. So a lot of federal organizations, according to this report, are now starting to use the enterprise version of ChatGPT. That includes NASA. So NASA has committed to using OpenAI's tools, having acquired an annual license for, for ChatGPT Enterprise after initial tests last year. The IRS, the Internal Revenue Service, just bought one hundred and fifty, according to reports, 150 Chat GPT Enterprise licenses for the Department of Treasury, although details on that remains unclear. Also, Los Alamos National Laboratory is also utilizing is it Laboratory? Yeah. Laboratory is utilizing ChatGPT Enterprise as part of its research collaboration with OpenAI. So OpenAI's Vice President of Global affairs revealed that the US Agency for International Development was the first federal customer for CHAT GPT Enterprise as the company aims to make its technology more accessible to federal agencies. So OpenAI is also partnering with the Air Force Research Laboratory. Laboratory. There we go. There's the word. I told you all I was on vacation. So to explore how Gen AI can reduce administrative burn burdens and enhance efficiency in military operations. So this partnership marks OpenAI's first collaboration within the Defense Department, though it has worked with other components previously. So the spokesperson from OpenAI assured that the collaboration right now between the US federal government and OpenAI is only involving unclassified systems adhering to usage policies that prohibit whole harmful applications of their technology. But y', all, let me just say this for all those companies out there, and maybe this is your company, maybe it's not. But for all companies are like, oh, our, our, our data is, is too sensitive for Chat GPT Enterprise. Like, no, it's not, right? You, you probably use a cloud service, you use Microsoft Cloud, Google Cloud, etc, it is the same technology. So yeah, companies hire us, you know, to teach them Chat GPT Enterprise. And the security is just like you would see, like I said, out of cloud storage. So if your company lives on the cloud, which all, almost all companies do, right? Yeah, you might have some things on prem, some things off prem, but for the most part, I mean, even the federal government is using ChatGPT Enterprise. So yeah, like why aren't you, why isn't your company. Let me know in the comments. Actually, I'm always curious. All right, our next piece of AI news that matters. Salesforce, according to reports, is hiring over 1,000 human workers to sell its new AI products. Interesting, right? So Salesforce, according to Salesforce, according to Bloomberg, is planning to hire more than 1,000 employees to support the launch of its new AI agent product, Agent Force. So Salesforce CEO Mark Benoff highlighted that Agent Force, which became available only two weeks ago, is already receiving incredible feedback from users. So Agent Force is designed to automate tasks such as customer support and sales development without the need for human intervention, marking a strategic shift in Salesforce's AI strategy. Right? Yeah. Isn't that a little wild? Right, Like Salesforce has always been about humans selling things and now they have an AI Agent Force product that is, you know, kind of AI taking over sales jobs, but they're hiring a thousand humans to sell it. So I don't know, call me crazy, I think this would have been a great use case for Salesforce to use Agent Force to sell Agent Force. Right. So I don't know, it seems like, at least to me. Let me know if this seems a little backwards to you. I mean, first of all, great in the age of AI first and you know, everyone's worried about, you know, AI taking jobs and it will take a ton of jobs. So I guess kudos to Salesforce for planning to hire a thousand more people, but for how long? Right. If its Agent Force product is as revolutionary as it claims that it is, it can perform all of these tasks that would normally a human would normally do inside of Salesforce. But yeah, if it is that revolutionary, why are we hiring a thousand humans to sell it? Shouldn't Agent Force just sell itself through Agent Force? Well, I don't know. Maybe, maybe not. So Salesforce has been focused on controlling sales expenses in recent years, which has included job costs, job cuts, and encouraging customers to explore self service options. So yeah, a lot going on in the AI agent space. We had a couple of shows in the last two weeks on this. So other tech giants, including ServiceNow, Anthropic Microsoft, are also developing similar autonomous agents, intensifying competition in that AI space. So Benoff has openly criticized Microsoft's AI efforts, particularly its Copilot product, suggesting that Salesforce aims to differentiate itself from through superior offerings. And let me just put this out there. I see what he's doing. I see what the CEO Mark Ben off is doing by criticizing Copilot. But I don't think most people are going to fall for this ploy if I'm being honest. I think what he's doing is, is smart. I would be doing the exact same thing. Right. If you want to compete with the best you try to pick a fight with the leader in the space. And I do think, at least in early AI autonomous agents, it is going to be Microsoft who is winning. They have their co pilot studio with autonomous AI agents rolling out this fall. I'm sure we'll see some announcements at the Ignite conference here in Chicago from Microsoft next week. So it makes sense that Ben off is, is. Is kind of, you know, pitting Salesforce against Microsoft. It makes perfect sense. But I think he's just doing it really just to draw attention to Agent Force. And I don't think it's, you know, I don't think necessarily one is better than the other. I think they obviously serve two very different purposes. All right, excuse me while I take a sip of water and, you know, might be coughing here. All right, fantastic. We're back. All right, so our next piece of AI news. A big win for open AI in court as a court dismissed a copyright lawsuit against open AI. So the Southern District of New York has dismissed a lawsuit filed by Raw Story Media Incorporated in Alternate media, Inc. Again, OpenAI. So this ruling is noteworthy as it addresses the legality of using scraped content from the web for training AI models. So the lawsuit claimed that OpenAI violated. 1202 of the DC, the DMCA, which is the D Alternate, without preserving Copyright Management information, or cmi. But the judge dismissed the case, stated that the plaintiffs failed to demonstrate concrete harm caused by OpenAI's action, which is a crucial requirement for standing in a lawsuit. So the judge highlighted that generative AI models like ChatGPT just synthesize information rather than reproduce it verbatim, making claims of direct infringement just speculative. So this dismissal aligns with other cases where courts have struggled to apply traditional copyright laws to generative AI technologies, emphasizing the challenges of proving copyright violations in this evolving landscape. So the ruling suggests that without clear evidence of harm or exact reproduction of copyrighted works, that plaintiffs may face difficulties in pursuing similar claims in the future. As courts navigate these complex issues, content creators may need to consider licensing agreements to protect their work and ensure proper compensation for use in AI training data sets. Yeah, here's. Let me just cut this one very straight. Large language models scrape copyrighted information from the Internet, period, point blank. Right. Don't have time to even explain into great detail, but a good majority of the information that gets into large language models is scraped content from the web that is copyrighted. But so far, there hasn't really been a marquee lawsuit going against these big AI companies for copyright infringement. And I personally don't see that happening anytime soon, at least not for a text based model. When we get into images, I think, I think we might see, I think we might see that just because you're literally seeing it sometimes spit out direct replicas, you know, artists, signatures, things like that. But I don't see it happening right now for text to text, large language models. All right, our next piece of AI news. Trump's return to the White House could reshape AI policy. So as Donald Trump prepares for his potential return to the presidency after winning the 2024 presidential election, the landscape of AI is really changing and up in the balance on what's going to happen with a new administration in charge. So Trump's first major action on AI is expected to be the repeal of President Biden's executive order on AI, which aimed to address civil rights and national security threats while promoting innovation. Trump's views on AIs have shifted between admiration and concern, particularly regarding competition with China, which he identifies as a major threat for the race for advanced AI development. Also, there's division within Trump's own circle. So this is according to reports from Time in a lot of other organizations, kind of, you know, as an example, Elon Musk maybe taking a visible position in Elon or sorry, in President Trump's cabinet. And he's expressing existential concerns about AI, while others, like Trump's VP, J.D. vance, are advocates for fewer regulations and open models. So, yeah, it should be interesting to see how this pans out. The other thing which we talked about last week, the future of the USAI Safety Institute, which was established under Biden to lead AI safety efforts, the future of that department remains uncertain, with some Republicans viewing it as an obstacle to innovation and their funding is running out. So that kind of the US has been a little different. So we had the Biden executive order, we had the US AI Safety Institute, but we don't technically have any real threat federal laws or aside from the US AI Safety Institute, we don't have a federal governing body. Right. So it is really a tumultuous time right now for people in the AI and tech industries on what's going to happen now in January with Trump as the president again and what some of his new policies may entail. So as an example, there's going to be reportedly an increased tariff on imports, which is a central part of Trump's campaign, but that could inadvertently raise costs for AI related hardware, impacting just about every single big tech company and every single company in Silicon Valley. So A lot of reports are saying, hey, your next laptop or smartphone could cost a lot more. So as an example, the Consumer Technology association warns that prices could skyrocket. Laptops could nearly double in price. Game consoles could go up 40% and smartphones could go up by 26%. So these tariffs on imports could really affect those in the AI industry as well. Reportedly anywhere from 10 to 20% tariffs or up to 60% tariffs for goods coming in from China specifically. So that should be very interesting to see how this impacts huge companies such as Nvidia, Microsoft, Alphabet, especially those who have hardware partnerships coming in from overseas. Because what that kind of, what's that, what's being reported is that costs for AI will have to go up because the hardware costs associated with producing these state of the art AI models are going to go up and that cost has to get passed on to someone. So should be pretty interesting to see how this pans out. Also, with people like Elon Musk potentially having a huge say for AI policy could be interesting because as polarizing as Elon Musk is, he has his own interest, right? He has his GROK AI model. So, you know, we'll see what actually happens here, but we'll obviously be covering it in the coming months. So Trump may also seek to address what he perceives as political biases in AI algorithms, potentially leading to executive orders that could alter how tech companies operate and develop their AI systems. So, like I said, the influence of prominent tech figures, including Elon Musk and others, will play a critical role in shaping Trump's AI policies, highlighting a shift in Silicon Valley's approach to collaboration with the government on national security issues. All right, our next piece of AI news that matters. Microsoft is adjusting Co Pilot Pro. Like, are they getting rid of it? Well, maybe, kind of, at least for certain countries. So Microsoft has announced that it will start including Copilot Pro features in its Microsoft 365 personal and family subscriptions in some countries, a move that comes after feedback from users regarding the previous $20 a month charge for these AI powered tools. So right now, this change currently applies to subscribers in Australia, where I just was, New Zealand, Malaysia, Singapore, Taiwan, Thailand. So a lot of countries there suggesting a trial phase before potential broader implementation. So here's how it could work. So subscribers will receive a monthly allocation of AI credits to use Copilot in various other applications, including Microsoft Word, Excel, PowerPoint, Outlook, OneNote, and the new Microsoft Designer app. But doesn't mean you get it for free. So the price of Microsoft365 personal and family subscriptions will reportedly increase with specific hikes, such as, as an example, for four Australian dollars for family subscriptions or five Australian dollars per month for personal subscriptions, so effective upon the next renewal date. So this price adjustment is part of Microsoft's strategy to reflect the added value over the past decade and to fund future innovations. The company's previous attempts to charge an additional $20 per month for Copilot Pro was met with a little bit of skepticism, indicating that. Are you still running in circles trying to figure out how to actually grow your business with AI? Maybe your company has been tinkering with large language models for a year or more, but can't really get traction to find ROI on Gen AI. Hey, this is Jordan Wilson, host of this very podcast. Companies like Adobe, Microsoft and Nvidia have partnered with us because they trust our expertise in educating the masses around generative AI to get ahead, and some of the most innovative companies in the country hire us to help with their AI strategy and to train hundreds of their employees on how to use Gen AI. So whether you're looking for ChatGPT training for thousands or just need help building your front end AI strategy, you can partner with us too, just like some of the biggest companies in the world do. Go to your everydayai.com partner to get in contact with our team or you can just click on the partner section of our website. We'll help you stop running in those AI circles and help get your team ahead and built a straight path to ROI on Gen. New approach may be more palatable for consumers, so will extend to the US or European markets, indicating that this selective rollout may be a precursor to wider adjustments. So yeah, I'm guessing we will see at the Microsoft conference November 19th in Chicago. Like I said, let me know if you're going to be there at the Microsoft Ignite conference in my hometown of Chicago. So I'm sure we'll be hearing a little bit more on Microsoft's future plan for Copilot Pro. And is that still going to be a thing? I think it's been great for Mac users such as myself, although I'm in the middle of probably switching over to a PC. I'll probably be buying one within a week or so once I get back to the US in like 36 hours. I think it's actually been a pretty good product, right? $20 a month unlimited, right? Yes. You still have to have your Microsoft subscription to use those Microsoft products on Mac, so I think this really is going to be impacting those that aren't, you know, using Microsoft 365 at, at, at your work. But most people, I'd say the overwhelming majority of people are already using Microsoft 5365 Pro, and if you are, then you are probably using Microsoft 365 Copilot as well. So this maybe may not be impacting a huge percentage of users, however, for some people like myself, right, I currently pay $20 a month for Copilot Pro as well as my, you know, Microsoft subscription. So yeah, it is going to be impacting a lot of people. But yeah, I'm curious, do you guys out there, do you use the free version of Microsoft Copilot? Do you use the Microsoft Copilot Pro? Or are you using that Microsoft 365 copilot? All right, our next piece of AI news. Google accidentally leaked its Jarvis AI, an upcoming web browsing agent so Google has recently kind of leaked details about its upcoming AI agent called Jarvis, which is designed to assist users by browsing the web on their behalf in Chrome. So we talked about this on the show last week, but an internal preview of Jarvis AI was posted on the Chrome Extension store, but was quickly removed, indicating that Google may have accidentally released it a little too early and maybe in the final stages of development. So the prototype was described as a helpful companion that serves surfs the web with you suggesting a focus on enhancing user experience during online activities like shopping and travel planning. So users who managed to download the prototype in time while it was still up on the Chrome web Store found it unusable due to specific access permission requirements, hinting at the complexity of the AI's functionality. The official unveiling of Jarvis AI is expected in December, according to reports from the Information, which adds to the anticipation surrounding this tool. So yeah, we'll actually see if it even keeps the name Jarvis AI. As we talked about before, there was another pretty popular GPT writing product called Jarvis for a couple of years. They reportedly got hit with a little lawsuit from Disney and Marvel and then had to change their name to Jasper. So we'll see if Google still calls it Jarvis AI or something else. But Jarvis AI through the Chrome Extension, aims to automate web browsing tasks, allowing users to focus on more important matters such as holiday shopping by handling the mundane tasks such as purchasing gifts. So this development follows a trend in AI where companies like Anthropic are releasing similar products, indicating a growing market for AI agents capable of performing tasks autonomously. Yeah, and if you are interested in the Anthropics version of this. It's called Computer Use. We did a dedicated show about two weeks ago walking you through step by step exactly how to use this Computer Use from Anthropic. It was extremely buggy, but I do think it shows some promise. Speaking of agents, Microsoft, not just Copilot Studio, they released another new multi a new multi agent AI system for enterprises called Magentic One. So Microsoft has introduced Magentic One, a multi agent AI system designed to automate complex tasks that typically require human intervention. So Magentic one is built on Microsoft's open source Autogen framework and aims to handle open ended web and file based tasks with future plans for more complex reasoning tasks. So right now it is an open source multi agent AI system designed to coordinate multiple AI agents to complete complex tasks. So developers and researchers can access it right now on GitHub under a custom Microsoft license. So yeah, this isn't something for the everyday person. You do have to be a little bit more technical. You have to know your way around a GitHub repo and you know, kind of, yeah, you gotta be a little bit of a dork to use this one. But let's talk a little bit more about it. So the system features an orchestrator agent that directs four specialized agents, so Web Surfer, File Surfer, Coder and Computer Terminal to efficiently complete assigned tasks. So each of these four sub agents has distinct functions. So for example, Web Surfer, like you can guess it manages tasks in web browsers, while Coder specializes in, you guessed it, writing and executing code. So currently Magentic One uses GPT4O from OpenAI as its underlying large language model, but can be integrated with other models that possess strong reasoning capabilities. So the company advises users to take precautions when deploying the system, including isolating agents in containers and limiting their Internet access to mitigate risk. Yeah, it's like, hey, these things can be powerful, go put them in a sandbox by themselves, don't let them play with other kids or don't let them play with live important data that is connected to other parts of your company. Also, Microsoft said a human supervisor is recommended to oversee the agents, particularly when handling sensitive data. So I have not yet had a chance to play with Magentic one and hey, let me know audience, live stream audience. Some of these things are pretty technical. So let me know if you do want to see something like this. All right, there we go. Our last AI news story. Open AI is reportedly facing some challenges with slowing AI improvements. So a recent report from the information indicates that OpenAI's next flagship model is codenamed Orion and it may not showcase the same level of advancements as previous iterations, raising concerns about the future of AI development. So according to reports from the Information, employees testing the Orion model noted that while it outperforms existing models, the improvements are less significant compared to the leap from GPT3 to GPT4. So if you haven't been following along, this Orion model is the next model. This is the code name for the next model in the GPT series. So as a reminder, the reasoning model, it is not in the GPT series. OpenAI said that that is its own so the O1 model. So right now we have access to O1 preview and O1 mini. It is a reasoning model, so that was previously co named Q Star, then Strawberry and was released as 01 Preview and 01 Mini. So with Orion it is not a reasoning model, at least not that we know of. It is a GPT model. So GPT 3, GPT 4, presumably Orion will be either GPT 4 or 5 or GPT 5. So specifically, some OpenAI employees who tested the new model called Orion say it performs at a similar level to GPT4 after only 20% of its training, which I would say is actually pretty impressive. Right? But you read the headline and you know all the news, it's, it's saying oh the, the, this AI improvement has slowed down according to this. I, I don't know, I don't see it, but let's continue talking about it. So however, the improvements, they said employees that tested this were speaking anonymously to the Information. They said the improvement isn't as big as the jump was from GPT3 to GPT4, which suggests that making models better by simply adding more data might be getting harder now that there's less high quality data available. So in some areas, including coding, Orion may not consistently outperform prior models, suggesting a slowdown in the rate of AI enhancements. So to address these challenges, OpenAI has established a foundations team tasked with developing new strategies to sustain model improvements despite a decreasing availability of fresh training data. So reportedly these strategies may involve training Orion using synthetic data generated by AI models such as O1. Right? These reasoning models, they said that they said that they are going to use the O1 or the O series of models as to create new synthetic training data for the GPT models. So OpenAI has not confirmed any of this or they've not confirmed plans to release a GPT5 or Orion in 2025, responding to inquiries with a statement indicating no Immediate rollout. One important thing to note about this article is it did quote Noam Brown, a researcher at Open AI. So he was apparently quoted in the article from a TED Talk that he gave. However, on Twitter someone did ask him about this and he kind of deflected the overall sentiment of this article, which I found interesting saying quote in the TED AI talk that I gave, which they the information selectively quoted in the article. I make the case that there won't be a slowdown in AI progress anytime soon. So pretty interesting there. Even one of the people quoted in this article by name was quoted from a TED Talk and then the researcher himself responded and said hey, actually it seems like they just pulled a quote quote that didn't tell the whole story because actually in that talk he specifically said that there would not be any slowdown in the AI space. My hot take. Absolutely not. Absolutely not. I do think that the improvements from a GPT4 to a GPT5 will actually be smaller if, if you just look at the model quality just because, yeah, it's, it's, it's data. You do need a model that can reason because you know you can only get so far with essential. Right. Like large language models are essentially just advanced next token prediction. Right? Like times a million. It's much more complicated than that. But if the quality of the training data doesn't drastically improve to get better outputs from said GPT models, you have to start introducing either synthetic data from a reasoning model like O1, which kind of uses this under the hood chain of thought reasoning like a human would. Right. So instead of, you know, having to prompt engineer a GPT model or a Gemini model or a CLAUDE model. Right. The O1 model kind of does that by itself. Right. It does kind of the work that a human would normally do to get superior and better outputs from a GPT model. So I do think that they to see a sizable leap from a GPT4 to a GPT5 or a GPT 4.5. Yeah, you're gonna have to either use synthetic data because there's not enough high quality new data that's being created on a yearly basis. There's not. I think it's actually the opposite problem. I think so many people are starting to use AI and they don't know how and they're doing it in a lazy way. So much of the new information that is being posted online is a lot of AI data and it's a lot of stuff that's probably actually from GPT 3.5. So you almost have this regurgitation of older, low quality data. So I do think that model makers are going to have to start using synthetic data and they're going to have to start including some reasoning capabilities by default. Right. I do think to see a sizable leap in GPT5, in Orion, whatever it's going to be called, you're gonna have to borrow or use at least some of that secret sauce of O1 in the reasoning and the chain of thought thinking underneath to really get superior outputs. All right, that was a lot, y'. All. You can tell I haven't done this. You know, I've only done this now twice in the past 10 days. So it's been, it's been great to take some, a little bit of time off. But yeah, my, my voice, I'm already losing it now after 33 minutes. And y' all know I, I usually do this every single day for like 40 minutes. But let's do a quick recap of the AI news that matters. So first, Open, I came out that they have a lot of huge partnerships with the US federal government, agencies like NASA and the IRS. Subscribing is going to hire 1,000 human workers for its new AI Product Agent Force. A court has dismissed a copyright lawsuit against OpenAI. Trump's return to the White House is causing a lot of uncertainty around AI development. And what's going to happen with all of these companies when there's maybe new tariffs, there's maybe, you know, the executive order might be gone. So yeah, we'll see what happens there. Microsoft is apparently kind of starting to phase out Copilot Pro, at least in some countries. We'll see if that's rolled out worldwide. Google kind of accidentally leaked Jarvis AI its upcoming web browsing assistant. Speaking of that, Microsoft launched Magentic One, a new multi agent AI system for enterprises. And last but not least, according to the information, OpenAI is facing some challenges in its AI improvement rates going from model to model. All right, I hope this was helpful, y'. All. A lot of work goes into this every single week. Even when I'm on vacation, enjoying an anniversary with my wife, spend a lot of time. So if this was helpful, please let someone else know about it. Don't be a jerk. Go. Click. If you're listening to this on LinkedIn, please repost this. If you're on Twitter, retweet it or re exit, whatever it's called. Maybe if you're on the podcast, send this to someone, email this to someone, text it to someone. I know this might be your little secret on how you're the smartest person in AI at your company, but so much work goes into this. So if you could please repost this Share this if you're listening on the podcast, if you haven't already, please please click that subscribe or Follow button on the podcast player of your choice. Whether you're listening on Spotify or Apple, please leave us a rating. And please join us back tomorrow and every day for more Everyday AI. Thanks, y'. All.

And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going for a little more AI magic. Visit your everydayai.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.