Shane Jones’ Post

Does Microsoft Copilot (and others AI) really have security issues?  Short answer: No. People have security issues. Acceptable answer: The first factor to consider is responsibility. If I use a text written by someone to offend a certain category of people, or to discriminate against a specific group, the fault is mine, not of the one who wrote the original text. If I use a paper written by a scientist to build a dangerous object and harm someone, the responsibility is entirely mine, not the scientist’s. If I use AI irresponsibly, the blame and consequences continue to be mine, and no one else’s. We try to assign non-existent responsibilities to the AI itself, to those who design it, to those who train it, to those who run it, but we forget the most important responsibility, the individual one. AIs are also giving us the opportunity to open a constructive debate on the contemporary excesses of political correctness. The fear of being able to unintentionally offend someone is making us dumb, asocial and easily manipulable. I can’t write something because it could be misinterpreted and used incorrectly. I can’t express an assertive opinion because it could be used by someone else as a personal attack. I can’t create a work of art because someone could use it as an offensive propaganda tool. This obsession with achieving a “safety” that does not exist and cannot exist only distracts us from the real dangers arising from the development of increasingly performing AIs. AI is a catalyst and accelerator of human passions, and we are not ready for all this speed. The real security problem, as always, is man. If AI makes dangerous wrong decisions, it is because we need immediate decisions that free us from all responsibility. If AI generates dangerous content, it is because we are not able to manage our passions. The censors applied to most generative engines are already too many, and despite it being evident that hardcoded rules worsen the user experience without guaranteeing more safety, they continue to be used extensively. What we should focus technical and philosophical efforts on is the distinction between real and synthetic, are the shared definitions of intelligence and consciousness, is the education and empowerment of users. Full answer: Too much for LinkedIn, I must write a book 🤷🏻

I am not usually gushing with praise for anything technological, especially in the economic wasteland developing in the wake of artificial "intelligence". I do, however, have to say that Microsoft has done a bang-up job with their Phi-3 AI model. I was not expecting much after Phi-2 was a great disappointment, but 3's the charm, apparently. It behaves as I hoped other, larger models would for my use case. For a couple decades I have been rather disappointed with just about everything they have produced, disappointed with the culture working at Microsoft Research--"Where great ideas come do die," one said--and less than impressed with their involvement in open source and what they produced. To be fair, now I find greater disappointment in the code for the current tooling around AI that I am constantly needing to modify or code around. If the MS engineers during my tenure there were merely throwing code over their shoulders into the darkness, these other codebases are written by juiced raccoons who are trying to create the ultimate automated dumpster-diving robot. I would rather this success be the start of a growing trend instead of lonely flower in the wasteland destined to wither away.

While studying for my Microsoft AI-900 certification, I decided to make some YouTube shorts about various topics I thought folks might want to learn about. Here is my first post about Confusion Matrix on a Classification Model. Let me know what you think. https://lnkd.in/gTnHhAGa

This week's edition of The Upgrade newsletter: Unpacking the White House's AI Executive Order! 📜 The recent 100+ page EO issued by the White House aims to balance the potential and risks of AI. It introduces measures aimed at ensuring safety, addresses DEI and ethics concerns, and involves cloud providers, too. I unpack what it does and does not do! Plus, did you hear about the latest AI misstep by Microsoft news? 🤦♂️ 📰 Dive deep into this week's top AI stories, case studies, and more in the latest issue. From AI tools to legal challenges and workplace implications, stay updated on the intersection of AI, media, and storytelling. ✍️ Read it below, and subscribe to The Upgrade for weekly insights: https://lnkd.in/dsAQXraN #AI #WhiteHouse #ExecutiveOrder #TheUpgrade #Microsoft #AInews 

Today at Microsoft Build 2024, Thomson Reuters showcased the combined power of CoCounsel with Microsoft Copilot for Microsoft 365 setting the vision to combine the deep expertise and trusted content from Thomson Reuters with Copilot's deep knowledge of how and where professionals work to deliver a personalized AI experience - taking GenAI assistant know-how to the next level. https://gag.gl/MLl2Jq

More than half (55%) of top technology officers say they plan to purchase enterprise-level gen AI software in the within the next six months. But at $30 a user and a minimum commitment of 300 users currently required by Microsoft, what do you do if you can't afford a $9,000 monthly bill but you're desperate not to get left behind? Talk to us! Copilot is powered by GPT-4 yet the more powerful version, GPT-4 Turbo, costs just $0.01 per ~750 words entered and $0.03 per ~750 words outputted (source: https://buff.ly/3NXUnC6)! You don't need Copilot to access the world's most powerful AIs in a safe and secure way!

Today at Microsoft Build 2024, Thomson Reuters showcased the combined power of CoCounsel with Microsoft Copilot for Microsoft 365 setting the vision to combine the deep expertise and trusted content from Thomson Reuters with Copilot's deep knowledge of how and where professionals work to deliver a personalized AI experience - taking GenAI assistant know-how to the next level. https://ow.ly/jfoa50RPlqp

MAIVA - Microsoft AI Value Accelerator Playbook: This playbook aims to address this critical business problem by offering best practices and strategies to toward successful implementation and scaling of AI initiatives, ensuring they realize expected investment returns. https://lnkd.in/e_TuvFZs

OpenAI and Microsoft will continue to dominate in AI advancements, leveraging groundbreaking features and strategic offerings like GitHub Copilot Workspace. Plus, with OpenAI’s launch of powerful new tools like enhanced file search and vector stores, and Microsoft’s integration of cutting-edge AI, their leadership in the AI space is clear. The upcoming Microsoft Build event in May and Apple’s developer conference in June will further highlight the disparity in AI capabilities. Additionally, the new capabilities in cost management and file handling further secure their competitive edge. This contrast could become especially evident in the upcoming months. And, it's only May.