93 | Apple signs deal with OpenAI for iOS, more big news from Microsoft Copilot, Open AI and Google AI for this week's AI News

0:00Hello and welcome to a not so short weekend news episode of the Leveraging AI, the podcast that shares practical ethical ways to improve efficiency, grow your business, grow your business, and advance your career with AI. Hi, this is Isar Maitis, your host, and I thought we're going to get a week of slightly less crazy news after the weeks of the big announcements from OpenAI and Google and Microsoft, and just a little bit ahead of the announcements that are expected to come from Apple at their worldwide developer conference, but that did not happen. So there is a lot of stuff to cover and to talk about this week, so let's get started.

0:44Since we covered a lot of big news from Google and Microsoft and OpenAI in previous weeks, I will start this week with Anthropix. Those of you who don't know Anthropix, Anthropix are the company behind Cloud, one of my favorite tools, probably the one that writes the most natural language, that does the best summarization of topics that I use literally every single day for multiple tasks. They just announced a new capability to develop AI agents through their API. They call it tool use or function calling, which enables to connect Claude to external APIs and use that in order to take action and do different things with external software.

1:24So this could be used to build tools for almost anything you can imagine, from email assistant to helping you with shopping on different topics, to a categorization and summarization of data across multiple sources, etc. Literally anything you can imagine. The other interesting aspect that this tool also enables to work with images, which means you can provide visual data through the API and use that to build your applications and agents. It is currently available through Anthropix Messages API and on Amazon Bedrock and Google Vertex AI, which is all the platforms that host this tool. This tool has been in beta since April, and now they're opening it to the public.

2:04And most people on the beta site was using their smallest, fastest, and most cost-effective model of Cloud3 called Haiku, which costs$0.25 for every million tokens of input and$1.25 for every million tokens of output, which is a very reasonable amount for the capabilities that it provides. Again, that's for one million tokens, meaning about 750 ,000 words. So while this is still not available on the chat platform that would compete with tools like GPTs from ChatGPT and shortly Gems from Gemini from Google, I assume that's the direction where they're going to go next. And from OpenAI to Apple, as I mentioned, Apple's WWDC Worldwide Developer Conference is coming up fast.

2:52It's going to take place on June 10th. And we expect a lot of big news from Apple, a lot of them around AI. Apple has stayed very quiet on their AI research compared to the other big players like to Microsoft and Google and Anthropic and Meta. And we expect that to change dramatically in this coming conference. Apple has been in discussions with both OpenAI and Google to integrate their tools and their AI capabilities into Apple's offering. And as it stands right now, they have a signed agreement that was announced this past week with OpenAI to integrate OpenAI's capabilities into multiple different Apple services.

3:34The rumors are saying they're still in discussions with Google to potentially integrate them into some of their tools as well. This process is not something that is new to Apple. They have done something similar with integrating Google Maps into Apple products in the past, and they did that while they were developing their own products. So I believe Apple feels right now their AI capabilities are not to par with the most advanced models out there, whether that's Gemini or ChatGPT, but they are going to integrate the most advanced models right now to provide the best experience to Apple users across probably everything Apple, and then in parallel, develop their own tools that over time will replace the external tools.

4:14The AI tools are expected to, first and foremost, make Siri significantly more powerful and conversational and allow Siri to do a lot more things. But it's expected to also be integrated to a lot more aspects of iOS 18, including summarization features of news, summarization features of iMessages, maybe a chatbot within iMessages and the capability to create images, similar to what Meta has in both Instagram chat and WhatsApp. And so the exact list is not known yet, but it's very obvious that is coming. And that might also be integrated into some summarization of news in search in Safari, similar to what Google is doing right now.

4:57So lots of opportunities and we'll know within just over a week. Now, the company that's probably not too happy about this is Microsoft, which was the only large company that had to deal with OpenAI so far to run all of its products in the background. So as Microsoft is fully committed all in on AI and everything that it's doing, right now, OpenAI's technology is what's driving Copilot more or less across the board, even though they have been developing their own models and investing more and more in that and hiring people from different companies and bringing them over. But as of right now, they are running OpenAI tech in the background.

5:34And obviously, the fact that their biggest competitors on the computer side, Apple, is going to run similar technology is not great news for Microsoft. But staying on the OpenAI topic, OpenAI signed two huge licensing deals with two large and distinguished news sources. The biggest one is News Corp, which is the company of Rupert Murdoch, that includes major publications like the Wall Street Journal and the New York Post and the Times of London and many others. This deal is a multi-year agreement between OpenAI and News Corp. It does not include other aspects of News Corp like financial analysis tools and stuff like that, but it does include all their news media outlets in the United States, United Kingdom and Australia.

6:21and the goal is obviously to provide ChatGPT's future models and potentially search models. This is something that has a lot of rumors around it. So it will provide these tools, the access to all the historical data from these news outlets, as well as access to real-time data and updates that can be summarized in ChatGPT and other future OpenAI tools. This deal is as important to News Corp as it is to OpenAI. as you probably know, the news world and media world in general is not doing great and media buying has been shrinking. They face a lot of competition from social media and other platforms.

7:03And so a deal like that, that now providing them a paid way to monetize their news is helpful for both sides. The other big licensing deal of OpenAI this week is with Vox Media and The Atlantic, which gives them access to additional publications such as The Verge, Eater, and The New York Magazine. In this particular deal, in addition to providing open AI access to everything from Vox Media, it also works the other way around where Vox Media and The Atlantic are going to implement various types of open AI tools and working with them to develop dedicated tools for them in order to benefit and become more efficient as a news outlet.

7:46This is not the first time OpenAI signs a licensing deal, they've done similar deals with, as I mentioned, News Corp this week, but also with.dash, Meredith, and the Financial Times and Reddit. And what we're seeing is potentially how the future of news and magazines is going to look like as we start consuming more and more of our information through these GPTs and various chat bots versus directly with the sources. And again, this will allow these organizations the financial lifeline they so desperately need in order to survive in this new future. But not everything is unicorns and butterflies in the OpenAI world.

8:23As I mentioned, last week was a really bad week for OpenAI as far as news, and this week is not much different, and a lot of negative winds are blowing against OpenAI. Two former board members, Helen Toner and Tasha McAuley, both have raised serious concerns on the lack of transparency and the aggressive pursuit of revenue and domination and power by CEO Sam Altman. They were part of the board that kicked Sam Altman out. So obviously, there's a history there. So I don't know how much of it is accurate. But based on them and an interview they've done in The Economist, they've expressed serious worries about Altman's behavior in the past and in the current present.

9:07again, with his relentless drive to get more power and more influence, and with the fact that it's a self-governed organization right now, as he changed the formulation of the board, which allows him a lot more flexibility to do what he wants in the path to AGI. Helen Toner also had an interview this week on the TED AI show. It's a podcast that was aired on this Tuesday, and she accused Altman of being deceptive and withholding information from the board in multiple occasions. She even said that the board heard about the release of the original chat to the public on Twitter versus hearing it from him first before it was released.

9:49In addition to this, and obviously he's significantly increased of power by the fact that now he controls the board and by the fact that his technology is running everything in Microsoft and that his company is now also going to run and power the AI functionality behind everything Apple. It's been rumored, and these rumors have been shared by the information, which has been a very reliable source of information about everything in general, but definitely what's happening in AI. They shared that Sam is potentially looking for ways to transform the capped profit structure of OpenAI to a full for-profit company.

10:30So if you don't know the history, OpenAI was created as a nonprofit research organization with the goal of creating AI that will benefit humanity. After a few years, they've understood that in order to raise enough money to actually enable them to do that, they need some kind of a for-profit capability. So So they created a weird structure where the nonprofit organization is controlling a capped profit organization. That capped profit organization is the one that is currently powering Microsoft in a deal that provided a few billions of dollars, 13 to be exact, at this point, to OpenAI to actually run its operation.

11:06And now Sam Altman is potentially looking for a way to completely eliminate the nonprofit organization. So if you combine that with a few different things that happened this week and several different announcements from Sam saying he will do everything on the path to AGI because he believes it will benefit the world dramatically and he doesn't care how much it's going to cost, whether it's$5 billion or $50 billion or half a trillion. Again, when you hear these kind of things and then you hear those thoughts from people who left OpenAI, you understand that there might be a problem there. The problem that we have is that we have very little impact on that.

11:42But I think we at least need to be aware of what's happening or potentially happening. Again, I don't know the personal vendetta that these people have with Sam and the fact that they were trying to kick him out and he came back and they had to leave. So I'm putting all of this with a question mark in the end. But the overall view of what's happening right now doesn't look good. Another small piece of news on the same topic, Jan Leakey, the person that was running the super alignment team in OpenAI, which is the team that was in charge of safety of these models, he resigned and obviously shared serious safety concerns about OpenAI's safety focus or lack of it.

12:24He just joined Anthropic, which is not very surprising as Anthropic has always presented itself with the company that is focusing on the development of safe AI systems. So while these board members, as well as Jan Leakey and other people who left OpenAI, are claiming lack of control and that the fact it's self-governed is a problem, there is maybe hope. So the European Union's ChatGPT Task Force has released its first guidelines for AI-generated content, focusing on ethical and legal implications of such technology. The key recommendations that they have put in place are the following. Transparacy, so AI-generated content, should be clearly labeled as such, and users should be informed about which algorithms were used in order to create the content.

13:16Accountability is the second one. AI systems should be designed to ensure accountability for the content they generate, including the ability to identify and correct errors. Third one is explainability. AI systems should be able to provide explanations for the decision-making process and the data used to generate the content. And the fourth one is data protection. They want all the AI-generated content to comply with the EU data protection regulations to ensure privacy and security of user data. Now, this is not a law yet. These are recommendations from this task force. But that being said, there are zero capability to do any of these things with the existing models.

14:00Meaning right now, there isn't a silver bullet that knows how to identify AI content always. There's definitely no accountability because there's still a lot of hallucinations in these models. Nobody really knows to explain how these systems generate what they are generating, even though there are some advancements. And if you want to hear what Anthropic is doing on this topic, you can go to the episode from last week. And data protection, again, there are serious concerns on how these companies collected and are collecting data in order to train their models. So while this sounds really good, I would say this is unicorns and rainbows at this point.

14:38And the chances that this is actually happening, at least with the current technology, are very close to zero. And I don't think that the European Union's goal is to prevent people in Europe to use these models, meaning they will have to compromise somehow on all these things or wait a while until the models can actually achieve what they're claiming. I'm very happy that they're pushing in that direction. I would love to see similar approaches in other places around the world, and I would like to see a collaboration internationally between governments and regulators together with the companies that are creating this to protect us all from the potential negative harms of this technology so we can benefit from the amazing positive aspects of it while reducing the risks.

15:25Now, we briefly mentioned Microsoft as a concerned party to the deal between OpenAI and Apple, but there are a lot of news from Microsoft themselves. One of the things Microsoft announced this week is a full release of GitHub Copilot as a fully integrated capability into Visual Studio. So Copilot was initially introduced as an extension of Visual Studio a few years ago, back in 2021, but the capabilities of GitHub Copilot were not even close to what they are today. Today, GitHub Copilot is probably the most advanced and most widely used coding co-pilot in the world. And right now it becomes a integral part of Visual Studio.

16:06So developers working on top of Visual Studio, which is a huge percentage of developers around the world, will be able to enjoy the benefits of this amazing platform. It is going to be able to do a lot of stuff that I'm not going to explain right now because it's way too technical and way to focus on software development. But if you are a software developer, I highly suggest you go and check out the detailed news about this topic. Something that touches all of us, or at least most of us, if you are in business, is Microsoft just shared that they're going to add a lot more AI functionality to Excel.

16:41Right now, the functionality of AI in Excel is really limited. It can write functions for you and do things like pivot tables and create some basic charts, but that's roughly where it ends. The reality is there are a lot more capabilities you can do with this. I'm using large language models right now through a script that I've created to bring in large language models into Excel and Google Sheets. And that works amazingly well as far as analyzing data, both structured and unstructured. So qualitative and quantitative data, you can do real magic with it. OpenAI has introduced the capability to bring CSV files and Excel files straight into ChatGPT and use it in order to analyze and do really cool stuff with that data.

17:23So this kind of functionality is still missing from Excel. And there are rumors that they are bringing these capabilities to Excel. I think that's the promise since day one that AI can help you in everything that you're doing, including and maybe focusing even on data analysis. And right now, Excel is one of our top quantitative data analysis tools. But adding the capability to bring a large language model into it to actually query what's written in the cells can also allow doing qualitative data analysis. As I mentioned, it's something I've been doing for months and it's extremely powerful. And it will be awesome to have this as a built-in capability, both in Microsoft Excel, which was announced, and also in Google Sheets for Google users, which is something that was not announced yet, but I'll be really surprised if it's not in the pipeline.

18:09Sadly, Microsoft has not provided a specific timeline or the specific details of what's going to be included in that release. But I think we need to expect it's going to happen in the next few months. Another place that Microsoft Copilot has been integrated to is Telegram. So if you are a Telegram messaging app user, you are now able to bring Copilot straight into the messaging app in order to ask it question and chat with it and get information and summarize stuff, just like you can do in Copilot itself, but do it straight within Telegram. This is a similar functionality to what exists in meta different chats now running Llama in the background.

18:46In order to do that, you have to go to the relevant page on the Microsoft platform and agree and accept to the terms and conditions and then provide Copilot the phone number using for Telegram. And then it will appear on your Telegram messaging platform and you can start using Copilot straight in there. The other cool thing is that it's completely free right now, so you can benefit from it in an app that you're potentially using every single day while not paying anybody anything and enjoying the benefits of having it. In follow-up to some of the announcements they made last week in a Q &A session, Microsoft also shared that they're starting to deploy what they're calling Copilot Plus.

19:23Copilot Plus is a suite of several different next generation AI features that is going to be available and running locally on new AI-driven PCs that Microsoft announced last week. So this will include AI edits to compatible Adobe programs. It will include real-time live caption and translations on videos that you're watching, suggested visual and contextual edits of different things you're doing on the screen, local image and text generation, and so on. And as I mentioned, this is only going to be available on the Copilot Plus PCs, which are the new line of PCs that Microsoft announced last week.

20:02By the way, if you're looking for one, these PCs are supposed to be launched this month in June of 2020. And from Microsoft to Google, in a recent leak, it was revealed that Google is planning to introduce a significant upgrade to Google Photos, which is a very cool video feature. It will allow you to edit videos using AI straight within Google Photos. The feature codename is Piper, and it's supposed to allow you to create professional-looking videos using really advanced video editing capabilities just by using natural language and prompts. This includes finding relevant, interesting segments, automatic editing, audio enhancements, color corrections, and many other professional-level capabilities built straight into Google Photos, most likely for free and using AI to do this without the knowledge on how to actually do professional editing.

20:56This is great news for people like me and hundreds of millions of other people who are using Google Photos to manage all the images that they're taking. I'm an Android user for years, so I have a huge database of images and videos and literally every video I shoot with my phone shows up there. So the ability to edit that straight within the platform quickly and create professional-looking videos is a huge benefit. This is supposed to roll out in the next few months. As I shared with you last week, the rollout of Google AI Summarize News has not rolled out very well. And a lot of people over social media has shared multiple really funny and disturbing examples on how the AI summaries get stuff terribly wrong, such as adding glue to a pizza sauce in order to prevent the cheese from sliding off of it when you're baking it.

21:44one person that really enjoyed that and enjoyed laughing about it is the CEO and co-founder of Perplexity, Aravind Srinivas, who has obviously a competing platform that allows you to do Google research, but also use large language model capabilities. I have been using Perplexity for a very long time and I love using it. It's a phenomenal tool. I must say that I think in the long run, Google will win the war, but at least Perplexity can enjoy the fact that they won one of the battles. And the last piece of news from Google, not really from Google, but they're very interesting because it's the former head of China in Google.

22:23He is a scientist called Kai Fu. And he just shared that he believes that AI will replace at least half of human jobs by 2027. That's within the next three years. In the interview with him, he shared that he believes that so many jobs are going to be eliminated by AI or replaced by AI because of the repetitive nature of these jobs. But the interesting thing about his prediction, he was not talking just about knowledge, white collar jobs. He also mentioned stuff like trucking and assembly lines employees, because a lot of these tasks are highly repetitive. Because of that, he's emphasizing the critical aspect of AI in our lives.

23:05And he's saying it's going to be more impactful to humanity than electricity and the internet combined. Now, if you think about each and every one of those separately and how much it impacted our lives, that's a very bold statement. But Kai-Fu Lee has held executive positions in Apple, Microsoft, and Google, as I mentioned. So he knows one or two things about technology and its impact. So maybe we should listen. And to help you understand his point and maybe provide some proof from several different sectors, here are a few things that were announced this week. So a group called AI for Finance, in collaboration with Columbia University and NYU Shanghai, have developed what they call FinRobot.

23:48FinRobot is an open source platform that is built from several different language models, both large and small language models that are combined together in order to provide a very highly capable financial tool for businesses. It is built out of several different agents and layers. One is called financial AI agents, financial LLM algorithms, LLM ops, data ops, and multi-source LLM foundation models. So that's a mouthful, but all it means that they've built a tool built out of multiple layers using several different AI capabilities that can perform precise and efficient financial analysis addressing specifically transparency, global market adaptation, and real-time data processing that are all required for financial organizations.

24:37Now, if you're not sure if you will ever be willing to give your financial analysis to an AI, researchers in a recent research that was published this past week that was done by the University of Chicago, they have tested AI models, specifically GPT-4, that was not trained And in addition to its baseline capabilities on financial capabilities, they tested it against several different financial tools, as well as human financial analysts in doing analysis of companies' financials. And they have found that the tool was as good and in some cases surpassed the human capabilities and was as good as specifically trained models on finance.

25:17So that tells you that a general model, not even the latest one, because now they have GPT-4.0 and GPT-5 is supposed to come out sometime in the next few months, can now do financial analysis out of the box that is as good as the average human analyst. So if you think that a specifically trained model like the one I shared before cannot do it better, you're probably mistaken, which means very quickly we'll be able to do highly detailed, accurate, transparent financial analysis using AI tools with a lot less human analysts. Another example from a completely different direction in a completely different company, Klarna is a huge international corporation.

25:59They have went all in on AI. I shared with you earlier this year their huge success in deploying a support chatbot that is now replacing the work of 700 full-time employees answering millions of customer service requests over chat. So Klarna just announced that they are now using generative AI capabilities in streamlining and improving the efficiency of marketing campaign that is going to save them approximately$10 million per year. So Klarna are stating that they have increased the number of marketing campaigns while managing to save significant costs because they are using AI tools like MeJourney, Dali, and Firefly to create images instead of looking for images and editing images, which is what they have done before.

26:47That part saves them about$6 million a year, and that enables them to update their app and website and images to weekly reflect events such as Valentine's Day, Mother's Day, summer sales, and so on at a much higher frequency that they've done before and with a lot less money. In addition to money, it also saves them a lot of time. So the process that previously used to take them six weeks now takes them seven days. Now you're asking where the other$4 million of saving come from. It comes from cutting spending on external marketing suppliers, companies who help them with translation, social media agencies, and so on.

27:26They're now not using them. They're using AI to do the work of them, again, saving another$4 million. dollars. So if you have any doubts about the fact that AI is going to take a lot of jobs, these three simple examples all from this past week, and there are a lot more every single week, are a clear proof that this is where it's going. Now, there's no doubt in my mind AI will also generate a lot of jobs, but my fear is that it's not aligned as far as timeline, meaning a lot more jobs are going to be eliminated in the short run before new jobs are going to be created, if they're going to be created, because this particular revolution is very different than anything we've seen before.