Maciej Satkiewicz

⚠️ If you think dating Apps like Tinder are mostly for hookups, think again‼️ The results of this Green Flags study by the #datingapp giant, Tinder, confirms that as humans, we tend to fall into our own traps of thinking and limiting beliefs. According to the survey of over 8000 women and men, 65% of female respondents believed that men were looking for casual flings. And the reality was far from that! Only 29% males said they were looking for casual partners. So, on the one hand, we sabotage our own dating success by making the wrong assumptions, and on the other hand, I suspect the cause of many of these misunderstandings lies elsewhere: ➡ Singles are making basic mistakes on their own dating app profiles and sending the wrong messages! Dating on an app is an art and if you want to stand out and get ahead to meet the right partner for a lasting, committed relationship, you need to do your homework. I can help with that❣️ ➡️ Join my fast-track program "Master the Art of Dating on An App" in June and get ready for the season of amazing summer dates with great matches you meet on dating apps! In just one month, you'll have a profile that attracts the right people, and the strategy to transform chat conversations into dates, all while avoiding common traps and mistakes. Ready for the summer of love? The program (very small groups with my live coaching included during Zoom sessions) starts June 1, learn more and save your seat now (more information in the FEATURED section of my profile).

4

I have meant to write this for a long time because I believe it would help many founders in approaching database building. At LowCode Agency, we often see clients come to us with their own database designs that violate the First Normal Form (1NF) rule. 𝐖𝐡𝐚𝐭 𝐢𝐬 𝟏𝐍𝐅? First Normal Form (1NF) is a fundamental principle in relational database design. It ensures that the values in each column of a table are atomic, and that there are no repeating groups within a table. 𝐂𝐨𝐦𝐦𝐨𝐧 𝐕𝐢𝐨𝐥𝐚𝐭𝐢𝐨𝐧𝐬 𝐨𝐟 𝟏𝐍𝐅 One common violation we often see is the presence of repeating groups in a table. For example, having columns like "Item 1 ID," "Item 1 Price," "Item 1 Quantity," "Item 2 ID," "Item 2 Price," "Item 2 Quantity," and so on. This structure violates 1NF because it implies that each order can have multiple items, leading to redundancy and inefficiency. In one recent case, we had a client with up to 30 repeating groups in a single table. 𝐇𝐨𝐰 𝐭𝐨 𝐂𝐨𝐦𝐩𝐥𝐲 𝐰𝐢𝐭𝐡 𝟏𝐍𝐅 To comply with 1NF, you need to redesign the table to ensure atomicity and eliminate repeating groups. Here's an example of how you can transform the above table into a 1NF-compliant structure: 𝐈𝐭𝐞𝐦𝐬 𝐭𝐚𝐛𝐥𝐞: Item ID, Item Name, Item Description 𝐎𝐫𝐝𝐞𝐫 𝐭𝐚𝐛𝐥𝐞: Order ID, Order Date, Customer ID, Notes 𝐎𝐫𝐝𝐞𝐫 𝐈𝐭𝐞𝐦𝐬 𝐭𝐚𝐛𝐥𝐞: Order ID, Item ID, Item Price, Item Quantity 𝐁𝐞𝐧𝐞𝐟𝐢𝐭𝐬 𝐨𝐟 𝐂𝐨𝐦𝐩𝐥𝐲𝐢𝐧𝐠 𝐰𝐢𝐭𝐡 𝟏𝐍𝐅 Data Integrity: Ensures that each piece of data is stored only once, reducing redundancy. Scalability: Allows you to easily accommodate new data without altering the table structure. Efficient Querying: Simplifies querying and reporting by providing a clear and consistent structure. Maintenance: Makes it easier to maintain and update the database.

22

8 komentarzy

☝️ From tips 1-5, there’s an assumption that the model is fixed, but it gets even more complicated when you can change the model. On Day 6, we’re exploring model level optimizations, which include any method that changes the model or architecture itself. Since neural networks often contain many redundancies, it’s essential to exploit or streamline these redundancies to accelerate the inference process. What can you do to optimize inference at the model level?   𝐒𝐢𝐱𝐭𝐡 𝐭𝐢𝐩: 𝐂𝐨𝐦𝐩𝐫𝐞𝐬𝐬 𝐦𝐨𝐝𝐞𝐥𝐬 𝐰𝐢𝐭𝐡 𝐦𝐨𝐝𝐞𝐥 𝐝𝐢𝐬𝐭𝐢𝐥𝐥𝐚𝐭𝐢𝐨𝐧. 🧑‍🏫 Widely applicable in various AI domains, including NLP, speech recognition, visual recognition, and recommendation systems, model distillation is a training technique that trains small models to be as accurate as larger models by transferring knowledge. 🔤 A classic example of model compression can be seen in various BERT models that employ knowledge distillation to compress their large deep models into lightweight versions of BERT. For instance, BERT-PKD, TinyBERT, DistilBERT, and BERT-BiLSTM-based model compression techniques solve multilingual tasks using lightweight language models. – What’s the #11DaysofInferenceAccelerationTechniques? The Deci team is posting, for 11 days, a series of inference acceleration techniques for deep learning applications. If you’re looking for practical tips and best practices for improving inference, follow Deci AI so you won’t miss an update. #deeplearning #machinelearning #neuralnetworks #computervision

14

Hey guys, if you want a really good random number generator for your usecases, do check this open source Library out! Do tag us if you're using and let us know your usecases!

4

How #AI Simplifies Complex Topics with Realistic Examples 💡 Professional Explanation of Dependency Injection (DI) in Angular. A design pattern where a class requests its dependencies rather than instantiating them itself. #Angular's DI system manages the creation, sharing, and lifecycle of dependencies, ensuring consistency and efficiency. Not clear, ah? Let’s ask #AI to simplify with an example Dependency Injection is a way of giving a class (or component) the tools it needs (dependencies) from the outside, instead of the class creating them by itself. This makes the code easier to manage, test, and reuse because it separates the "setup" of tools from their "use." Imagine you're a chef (the class) 👨‍🍳 To cook a meal, you need ingredients like vegetables and spices 🥕🥦🍅🥒🌶️🧄🧅🧂 (dependencies). Instead of growing your own vegetables or making your own spices, you ask a supplier to deliver them 🚚 (Dependency Injection). Without Dependency Injection - The Chef is tightly coupled with the specific spices or ingredients he can use. If you want to switch to a different type of spice, you must modify the Chef class itself. With Dependency Injection - The Chef doesn’t grow or create the spices; he simply receives them. This makes it easy to replace the spices with different types or ingredients without changing the Chef class. Crystal clear now, isn’t it? 😉

4

Running llama 3 by meta does not require you to have sophisticated system setup. I have set it up on my personal laptop which happens to have a GPU with it. I have used llama.cpp to run the int4 and int8 quantized model of llama 3. Running a language model for your personal use does not require you to have cloud access such as azure. I have demonstrated one use case of deploying personal language model. During my research purposes It so happened that chatgpt was very hard to access either cause it was down or the site simply would be buggy during which I used the llama model deployed by https://lnkd.in/gWfkaKZA where I think such systems can be deployed by organizations for internal research purposes. Below is the link to the steps I followed in deploying and comparing the int4 and int8 quantized model. https://lnkd.in/gzShCPCD

21

5 komentarzy

5 Fun Facts About Software Development History!  The First Compiler: Grace Hopper, a pioneering computer scientist, developed the first-ever compiler in 1952. This groundbreaking tool translated human-readable code into machine language, revolutionizing software development.  The First Line of Code: The first high-level programming language was Fortran, developed in the 1950s by IBM. It was designed to simplify coding for scientists and engineers, making it easier to write complex mathematical calculations. Fortran stands for "Formula Translation."  The Birth of Open Source: In 1991, Linus Torvalds released the Linux kernel, marking a major milestone in open-source software. Today, Linux powers everything from servers to smartphones!  Java's Original Name: Java, one of the most popular programming languages, was originally called "Oak." It was renamed after the developers saw an oak tree outside their office and realized the name was already taken.  The First Video Game: "Spacewar!", created in 1962 by MIT students, is often credited as the first computer video game. It was developed on a PDP-1, a computer the size of a small room!  Love learning about tech history? Dive deeper into the world of software with Zavod-IT: zavod-it.com #SoftwareHistory #TechFacts #ZavodIT #Programming #TechTrivia

6

Just heard Praktika.ai was doing $20m in annual run rate. Basically it's your AI tutor for learning English. It's still raw... but in my view it's gonna improve fast and be like an arms race among various players. I predict in 5 years... as a % of english language learners... 90%+ will be doing it via pure AI and only <10% will still have any kind of human tutor involved. And i predict this after having learned 5 different languages (Spanish, French, Japanese, Chinese, Russian) via human tutors in the past. And having created and run a language learning business based on human tutors nearly 2 decades back. There are just many reasons why AI is much better than a human in the language learning process... 1- far cheaper 2- do it whenever you have time... no mess with scheduling or time zones. 3- very accurate progression tracking and therefore focusing on what you need to progress... rather than what is often fairly random topic selection by an amateur tutor. Are there any advantages to a human language tutor? Yeah.. if you're single and they are cute... then that is still a very good motivation to study. lol But in my experience... that's about it. Can the language learning incumbents pivot in time to save their business? Do their competencies translate to this new business? We're gonna find out... https://praktika.ai/

10

5 komentarzy

A great post & article. A must read!

6

NoFrills also makes mistakes sometimes (though rarely), but our test data shows NoFrills is more stable and accurate with live data. My previous posts highlight why we should consider NoFrills on par with other global solutions, moreover it proves that it is not a GPT wrapper or an open-source app. When it comes to Indic languages, NoFrills is miles ahead of all the models mentioned above. I will provide some examples in the coming days.

19

I write a lot of texts in three languages, and by default, of course, with errors. Running them through a separate ChatGPT window every time is too lazy. So I came up with the idea of making a contextual tool, especially since Mac OS X allows it. "Contextual" means that it doesn't matter which OS application the text being proofread is in: in email, in forms on websites or services within Chrome, in PyCharm comments, Slite, TextMate, ... I just select the desired fragment, press Cmd+Shift+0, and in 2 seconds the text changes in the same place to one well-written by a native speaker. It's super convenient. Everything is done based on OS X Automator + a small script working through ChatGPT. An example of the script and prompt is in the screenshot. I wrote a little article about this on Medium ages ago (I'll provide it again in the comments). But somehow it didn't generate much interest among people, only got 12 claps. The sources are here: https://lnkd.in/dpakKrAE The details are on Medium: https://lnkd.in/dNK-hX7K

15

Loom - My biggest timesaver of 2024 I rarely participate in these "games" but I was curious - how much do I use Loom? According to their report, I recorded 557 Looms last year. It says that equates to avoiding 135 meetings, but I would argue it was more like several hundred. What is Loom? If you're not familiar with Loom, you can use it on your desktop or phone to record your screen, keystrokes, mouse movements, voice and face simultaneously (face is optional). When do I use it? I started using it in 2022 for troubleshooting tech issues so I could show and explain to my developers what I wanted them to change on our website. I still use it heavily for this, but then I started using for training purposes with new team members to show them how to do something in our system. I also use it regularly to communicate with all of my team members when I'm delegating projects or trying to explain something where visuals help, like how to change a graphic for social or our site. Honestly, I use it for almost everything. Why? Simply because it saves me time, enhances communication, and documents what I requested and when. The recipient I send the link to does not have to have any account to access it. It automatically transcribes all of my instructions. I have my videos organized into a library and can easily search by title if I'm looking back for something. I do pay for the monthly subscription, but it is no brainer for the amount of time it saves me in a month. I should have an affiliate link, but I don't! lol. I am just simply sharing this piece of software for anyone not familiar with it because its a non-negotiable in my "tech stack" and I love learning what people use to run their life and business more efficiently! Trello, ChatGPT, and Notebook LM are right up there too. What are your top three pieces of tech driving efficiency? I'd love to know!

8

2 komentarze

🚀 Exciting news in the world of code language models! OpenCoder is redefining standards for open-source LLMs with a focus on transparency and reproducibility. 🌐 Here's what makes OpenCoder stand out: 🔍 RefineCode Dataset: A rich dataset of 960 billion tokens from 607 programming languages. 💡 Advanced Pretraining: Choose from two model variants (1.5B and 8B parameters), trained with stringent data cleaning and deduplication methods. 🔥 Instruction Tuning: This two-stage process aligns theoretical computer science with real-world coding skills through diverse instruction sets. 🏆 State-of-the-Art Benchmarks: OpenCoder shines in multilingual code generation and debugging, confirmed by tests like HumanEval and McEval. OpenCoder's unmatched transparency and performance empower researchers and engineers to explore, enhance, and innovate in code intelligence. Curious to learn more? Dive into this groundbreaking initiative and see the future of coding intelligence! https://lnkd.in/enTyYWbh

3

3 komentarze

Except they aren't the same. Despite me trying to convince myself they are. I just finished reading Aurimas' newsletter where he talked about memory for agents. Side note, I'm organizing the AI Agents in production virtual conference on November 13th. We will go in-depth on AI agents. Register here https://mlops-l.ink/0srs Some key non-AI generated thoughts and takeaways from my side: 1. Simplified definition and diagram of agents - basic recipe is knowledge, LLM, short-term and long-term memory and tools. (check out the diagram if you are a visual learner. 2. Goldfish agents are worthless - I think lots of early agent frameworks ran into problems because they couldn't properly manage short-term memory. ie continuously updating the context in the system prompt is a recipe for failure. 3. Long-term memory to save the day - lots of focus is being shifted to long-term memory to remedy these short-term memory problems. 4. Hive mind memory - interesting thought experiment is to consider that multiple instances of the same agent interacting with different humans could tap into this memory independently creating a sort of hive mind. 5. Could episodic memory give us collaborative mode - how can I make my agents available for other folks to use? how can I give you an agent workflow that you can tweak to your needs? 6. Semantic memory can get you far - mainly for grounding and reinforcing something that you hold important. (it also resembles a rag system a bit more) 7. Procedural memory for reproducibility - this is where the agentic frameworks value props shine. anything that will help codify the process so that you can track changes and better reproduce successful results should be tracked. Diving into the AI agents world makes me so stoked for the Agents in production virtual conference. https://mlops-l.ink/0srs

172

24 komentarze

Is it me or does the documentation of CrewAI and LangChain seem quite similar (Overkill). 😣🤡 Interesting! 😶 In contrast, Swarm by OpenAI seems promising in terms of ease of use, modularity and too generic that it can be customized. Unfortunately, no idea if it will be supported by them or will it have any updates or fixes down the line. Sad. What is your recommended framework for autonomous production grade AI agents/assitants? Share your thoughts and insights in the comments 👇

22

27 komentarzy

Cultural Mindset in Action A striking example of how different cultures approach questions and collaboration: Responses to a question in a Russian chat: - First of all, you asked the wrong question - Why can't you just use search? - I don't know the answer, but you'll probably fail anyway The same question in a Finnish chat: - I don't know myself, but I've heard something about it. Let me ask around - Here's a helpful link about this topic - Great initiative, guys! Good luck! It is widely believed that the level of trust between people largely determines economic well-being. However, another equally significant factor is how we support one another. Let's support people around us in their initiatives! ❤️ --------------- Hi, I'm Dima. I'm building Karhuno AI - the AI-powered search engine that empowers B2B SMEs to target ideal clients through advanced semantic analysis and sales signals. #entrepreneurship

7

3 komentarze

Many times when I am presenting someone will say, "If GenAI is so good at knowledge-level tasks what will kids need to know?" I always answer that these tools are not very helpful if you have zero knowledge base or lack curiosity and a desire to learn. The two posts below reinforce this and should be considered as we roll these models out to learners. We should consider some of these ideas as students engage with an LLM: 💠 Ensure that we teach them how to prompt engineer. This skill set will need to be updated as the tools emerge and work in new ways. 💠 Be clear on what the requisite knowledge is as you begin the unit of study and ensure that learners have the prior knowledge needed. 💠 Utilize prompts that are already defined for some tasks. 💠 Require some learning to happen offline and outside of the LLM. 💠 Rethink how we assess learning. 💠 Remember that learning happens in many ways and one of them is collaboratively. Provide opportunities for group work with well-defined roles. These are just a couple of ideas, but I encourage others to share. As always, thank you Leon Furze for getting me to think deeply this morning. Richard Charles, Ph.D., GSP Amos Fodchuk Kourtney Bostain Jon Gregori Dr. Theresa Perry Paige Johnson (she/her) Clay Smith Tom Ryan Joe Ableidinger Emma Braaten Marlo Gaddis

18

2 komentarze

Is A.I. going to replace programmers? The answer is yes, but there’s some nuance to it! A good way of looking at programmers is as diplomats between humans and machines. Skilled translators that use their mastery of language to “negotiate” with computers on behalf of companies trying to turn an idea into a product (I know a lot of programmers create their own products, but they're a minority as far as worldwide employment goes, so stay with me here). But the thing is that A.I. is great at cutting middlemen - and programmers (for the most part, globally) are middle-men between someone’s ideas and the final product. And with the way things are going, it’s pretty safe to assume that we’re venturing into a world of fewer and fewer programmers. It will start as it has now: where A.I. isn’t replacing programmers directly, but making today’s programmers much more efficient. I’d say that A.I. has probably made me about 2-3x better at coding tasks right now - but now it means that I am considerably less likely to hire an extra programmer to help me out. When you have a continuously evolving hive mind at your disposal at $15/mo, when is paying a programmer $100/hr worth it? But where we’ll end up is also just as interesting: we’re entering a world where everyone will be a programmer. Where if you have an idea, you can simply talk to an A.I. system and it will build it for you. A world where where A.I. systems will be able to do the coding AND the diplomacy/translation/logistics work that programmers currently do. So is A.I. going to replace programmers? Yes. But by making everyone a programmer. If you’re looking for an advisor to your company, I’m finally open for business after years of people telling me I should do it. I can help you navigate this space and save you a loooot of money. #artificialintelligence #coding #programming #futurism #innovation #engineer #ai #coder futurism

71

34 komentarze