Links: https://finance.yahoo.com/news/openai-says-wont-release-model-172418235.html https://www.bloomberg.com/news/features/2024-10-27/russia-is-getting-nvidia-ai-chips-from-an-indian-pharma-company?embedded-checkout=true https://readwrite.com/coinbase-shares-new-ai-agent-crypto-creator-as-it-hones-in-on-ai/ https://www.theinformation.com/articles/google-preps-ai-that-takes-over-computers https://machinelearning.apple.com/research/homomorphic-encryption https://www.newsbytesapp.com/news/world/mumbai-based-pharma-firm-aids-russia-in-acquiring-nvidia-amd-s-tech/story https://www.openread.academy/en/paper/reading?corpusId=510746906 In today’s video, I dive into five key AI stories from the weekend, and you might want some coffee for this one—it starts with the concept of homomorphic encryption. If you’re not familiar, this tech allows computations to be performed on encrypted data without decrypting it first, keeping your data secure. Apple is rumored to be leveraging homomorphic encryption for image recognition, which means your photos stay private even when AI analyzes them. Imagine asking Siri to identify a picture, and only the encrypted data leaves your phone, protecting your privacy while still enabling AI-powered insights. The second piece of news revolves around the buzz about a rumored OpenAI model named Orion, which was said to be 100 times more powerful than GPT-4 and expected to debut in November. However, OpenAI swiftly pushed back, denying any imminent Orion release. Could this mean a different model is coming, or is it just a matter of timing? Speculation is high, especially with Google planning a significant Gemini model release for December. The talk is that Google’s new release could enable AI agents to do much more, possibly changing the game for AI capabilities in everyday tasks. Meanwhile, Coinbase is stepping up to launch autonomous AI agents on the blockchain. This move seems aligned with the recent trend of AI-driven meme coins, such as Goat.cs Maximus. Whether these agents will drive real change or face governance issues remains to be seen. In more controversial news, NVIDIA’s H100 chips are being smuggled into Russia through an Indian pharmaceutical company, despite export bans. It appears Russia is rapidly scaling up its AI infrastructure, using these chips for data centers. Lastly, a major report on AI usage reveals a persistent gender gap, with women using AI 25% less than men in most demographics. Even though U.S. tech workers showed nearly equal usage, the disparity is concerning, especially as AI fluency becomes increasingly essential. #product #productmanager #productmanagement #startup #business #openai #llm #ai #microsoft #google #gemini #anthropic #claude #llama #meta #nvidia #career #careeradvice #mentor #mentorship #mentortiktok #mentortok #careertok #job #jobadvice #future #2024 #story #news #dev #coding #code #engineering #engineer #coder #sales #cs #marketing #agent #work #workflow #smart #thinking #strategy #cool #real #jobtips #hack #hacks #tip #tips #tech #techtok #techtiktok #openaidevday #aiupdates #techtrends #voiceAI #developerlife #apple #russia #india #encryption #security #safety #gender #gap #coinbase #crypto #agent #LLM

Six pieces of AI news you might have missed this weekend. I hope you can pronounce homomorphic encryption because that is what we are starting with. I know we're going right into it. Homomorphic encryption. That is a cryptographic technique. It's not necessarily an AI thing. It's being used in AI here. So homomorphic encryption is basically we take this data, we encrypt it, and we retain the shape of the data. The ciphertext that's retained is computationally identical. And that matters because if you can do math on the encrypted data, you can use machine learning on it, which is exactly what Apple is proposing in their new paper explaining how they plan to use homomorphic encryption to conduct operations on sensitive images and text on your phone on cloud servers without violating your privacy. And so the example they're giving is that you take a picture of the Eiffel Tower. Maybe you don't know what the Eiffel Tower is. I don't know. You ask your Apple intelligence, what's the Eiffel Tower? And what Apple does to identify that image is it sends a homomorphically encrypted copy of that image. So the original image is not going over to their servers. It's just a homomorphically encrypted copy. They compare it computationally with lots of other examples of encrypted images in their database. And they come back and they say, computationally, this looks like the Eiffel Tower, except they say it in a way that's much more human, like LLMs tend to do. And they say, hey, it's the Eiffel Tower. And you have no idea how it did that. Well, they used homomorphic encryption to keep your information private. So it's a really landmark paper, if only because Apple is, I think, trying to shift the industry toward being more transparent about privacy practices when it comes to LLMs. And we will see if they're able to encourage others to reveal what they're doing to protect privacy or not. All right, number two, news and counter news. Orion is the codename for the new chat GPT, or so The Verge was reporting. It's certainly not the first one to report that the codename is Orion. It was the first one to report that November 30th is when it will release. To corporations first, apparently. And OpenAI immediately pushed back and said, no, no, no, no. That's not the release date for Orion. Now you can read that multiple ways. It could be, we are releasing something November 30. It's just not codenamed Orion. You could read it as, nah, it's not the date for Orion. Or you could read it as, we have another model that's coming out in December, and we don't want to talk about it. We will see. I still expect some kind of OpenAI release toward the end of the year. And part of why is news item number three, which is Google. Google is apparently dropping a new model of Gemini publicly. And thanks to the success of this agentified CloudSonic 3.5, where Cloud can drive your computer, well, they're releasing a version of Gemini that can do that. The agents are coming. They're coming by the end of the year, just like I've been sort of calling out for a bit here. And that's going to put pressure on OpenAI to release something. This is very much a sort of an arms race scenario right now. Okay, piece of news number four. Coinbase has seen the success of GoatCoin and other things on the blockchain, and they are claiming they are releasing an agent on blockchain, or an agent builder on blockchain, so you can build autonomous AI agents on blockchain very, very quickly and easily. Why they decided to do this after they saw major news of a rug puller over the weekend, it was one of the scandals that broke. To no one's surprise, there was news of a rug puller for a different AI meme coin over the weekend. And this is just what you're going to get with meme coins. Okay, then after we talk about Coinbase, we need to talk about Nvidia. Nvidia's H100 chips are the generation behind the current state-of-the-art, which is Blackwell. H100 is, I believe, it's a different code name, whatever. I'm tired. It's morning. So anyway, those chips, piles of them, 1,100 Dell servers worth of H100 chips are going to Russia via Shreya, which is a medical devices company registered in India. They are not supposed to be going to Russia because they have been on the no import list that the US has issued for certain sorts of chips. And the reason for that is that they're trying to block the development of AI models in certain countries. And H100 chips can obviously be used for AI model development. And so $300 million worth of those chips has gotten into Russia and it is a national security issue now. So there's that. Last but not least, I believe this is news item number six. There is a gender gap in AI adoption. And so the paper I read found 25% on average less adoption by women of AI versus men. And it was widespread. It was everywhere from Kenya and India all the way to the US to Sweden. And the gender gap was really consistent. The only place where there wasn't a gender gap observed was US tech workers where women actually outpaced men very, very slightly on adoption. And if you're wondering what that number is, roughly two thirds of US tech workers apparently are using chat GPT or a similar thing. So that's sort of a depressing way to start the Monday. You have sort of the Russia import thing, you have conflicting news on Orion, you have news of a gender gap in AI adoption. I don't get to make the news, sadly. That's what you got. And hopefully, what we will see is clarity on these AI releases soon. And we will see an understanding of why gender adoption isn't the same. It is a weird one. It's not necessarily, perhaps it's tied to prevalence at work and so that it ties into other disparities. I don't know. I'm really curious why it persists. And I'm concerned because usage tends to translate into fluency. And this is one of those foundational skills for the future. So it's actually something where we really need to see a gender gap on this close early so that we avoid perpetuating disparities. All right, well, that's all the news that's fit to talk about. I don't know if it was all fit to talk about, to be honest with you. It's weird, especially the AI on-chain agents. All right, cheers. Happy Monday.