Diogo Neves

Here's how to use AI for your #ABOG articles for MOC. And no it's not to cheat on the quiz! Use Notebook LM to make learning enjoyable. Here's what to do 1. Go to NotebookLM: https://lnkd.in/eFCbixdV 2. Then paste an article you are interested in. 3. Click Notebook Guide to get a summary or study guide. Or generate a podcast on the article that you can listen to on the way to the hospital. The AI generated podcast is surprisingly easy to follow and informative.

6

1 Comment

So no doubt everyone has been talking about the latest llama3.1 large language model that has been released by Meta today. It's supposedly equal or better than ChatGPT 4o or Claude Sonnet 3.5. I've just taken it out for a spin on our GPUs and despite all the energy used to built this model it still can't do simple maths! What to know why? Then like and comment below and that'll keep me motivated to finish my next presentation on this topic!

13

3 Comments

Looking for resources on how to make effective use of LLMs in a developer workflow, if anyone has any good ones they can share. My GitHub copilot trial has finished and I’m trying to decide if I should continue my subscription. The results I’ve had so far have been a bit underwhelming but I’m willing to accept that I might need to learn how to use it effectively to get the most out of it (hence this post). Its definitely a better autocomplete (at least in certain cases) and so I think it has meant I’ve typed less (always a good thing for these battered old fingers) but I don’t think it’s actually saved me time, as I spend the time I would’ve spent typing reading/checking the things it’s generated. I’d probably be won over enough by the better autocomplete to keep the subscription if it weren’t for the environmental impact. But just like I get the train where possible rather than flying, I’m wiling to accept a certain level of friction to limit my carbon footprint. LLMs would have to demonstrate a step change in productivity to justify this for me I think. I haven’t seen this step change yet, but maybe I’m looking in the wrong places?

3

11 Comments

I think right now I'm generally more interested in how AI will impact existing organisations than the new products it will enable. Those who have worked with me will know I kind of love building little automations, syncing stuff up, creating workflows, that sort of thing. And I have a strong hunch that AI is just going to enable so many more possibilities. I've been playing with building stuff with Airtable, Asana, Notion, BuildShip, and Gumloop (all with OpenAI in the background doing the heavy lifting) and I want to hear more about what others are trying out, how they're building them, what's working, what's not. So I'm super excited that CodeBase is doing our new AI Navigator course for SMEs in Aberdeen and the Borders. We all need to learn together and share our experiences. If you're in those areas, do check it out below - and perhaps tag any organisations you know who might be interested! https://lnkd.in/efYCZ7XR

15

2 Comments

Now that our open source development project, for TGN VCS equipment for use in any modeler, is well underway and starting to show results, I’m thinking of ideas for first use to show off what it can do. Why not show TGN at work on an icon? We could show TGN VCS equipment in the Farnsworth House to really TELL THE STORY of how this building is put together. The virtual tour at the link gives a nice impression. But the medium (the virtual tour, as with any kind of model) has limits. The understanding developed remains kind of stuck at surface skimming level. Of course there have been countless attempts made at addressing that, attempts to dig deeper. This one is a great effort: https://lnkd.in/dVDtDMnY Look at all the information at that website, which is dated 1994, documenting the house, various kinds of description, portions modeled in 3D, images of the original construction drawings, videos, various analytics, photos… It’s a useful archive. I’m grateful to those who made it. Their names appear on a Credits page at the site. And no doubt, many have modeled this house since then. How many students have made it their project to 3D model the Farnsworth House, in Sketchup or Rhino or whatever favorite modeler? But take ALL OF THAT as a given and ask yourself, let’s say you were going to replicate Míes van der Rohe and build one of these, a Farnsworth house; sure, copy a modernist masterpiece, verbatim. Why not, in the spirit of ‘AI’! But I digress. Let’s say, you’re going to build something different but inspired by this house. How well do you understand this thing, what this house IS, how it’s actually put together, step by step, idea by idea, system by system, piece by piece? So you can learn from that and use the concepts on a different project, and differently.. Look at the drawings here: https://lnkd.in/dbRXdgcm Do you really understand where those drawings are, in context, and therefore what they mean? What about the 3D model? Take the best one you can find. Do you understand how this house is built by flying around in the model? Do you have any idea whether or not anything that matters is missing anywhere in the model? Where? And how can you even look and see how things are put together? The state of the art, of software supporting this, is just a joke. Still, in 2024. It’s stupid. What has the industry been doing all this time? I still just can’t believe how counter-productive and self-defeating the whole situation is. Come on now, let’s fix this once and for all. Set this industry on a new path that makes sense. Anyone interested in joining our open source project developing TGN VCS equipment for use in ALL modeling platforms, message me. The TGN VCS development outline and spec: https://lnkd.in/dUMj8Eqa

7

1 Comment

(Gift link) - Nice oped from Martin Casado and Katherine Boyle at a16z VC on the blatant bias for Big Tech in D.C. on all things AI. I've never seen such government favoritism for incumbent tech companies in my 40+ year career. It's frankly obscene, self-destructive behavior. It may be some part exploitation of ignorance on the part of Big Techs (you've been duped), it may be some part faulty ideology, but I don't think there is any question it's also political corruption. It's one of the most extreme examples of government capture we've ever seen. To clarify my position -- I don't think it was necessary to even mention open source in this op-ed as it muddies the water with two issues, even if related (I support some open source, though in LLMs it does have security considerations -- all LLMs do). I'm also not against all regulation in AI, but I am fiercely opposed to regulation that favors dominant monopoly incumbents and/or new ventures with multiple billion dollar valuations that don't even have a product. As I repeat often, the super majority of problems and risks that need regulating in AI are caused by LLM firms and their Big Tech partners. That's where the focus should be, and since they are conflicted, the government shouldn't be relying on them for advice like they are, or internal AI advisers who came from them who are on the payroll of non-profits funded by tech billionaires with a myriad of conflicts. It's ugly. Bottom line: While the U.S. has a serious problem with the Big Tech oligopoly and their oligarchs, we are not quite yet an oligarchy, but we're getting very close to meeting that definition. We better fix it before it's too late. ~~~~~ Quote: Although the public-facing argument for AI regulation is to promote safety, we believe the true purpose is to suppress open-source innovation and deter competitive startups. Many emerging technology companies are well-equipped to help solve national-security challenges, but of the 22 members on the board, none represent startups, or what we call “little tech.” Only two are private companies, and the smallest organization on the board hovers around $1 billion in value. The AI companies selected for the board either are among the world’s largest companies or have received significant funding from those companies, and all are public advocates for stronger regulations on AI models.

1

Hyperoptimisation from a product that works perfectly well seems to be incredibly value destructive for all concerned. Not sure why it happens so much in tech I learned today that something coming up imminently on our engineering road map is updating all our lambdas to be compatible with new AWS standards. Apparently this is something that will happen every 18 months or so forever and GCP is apparently even worse. If we don’t do this by the time AWS deprecates this functionality Spot Ship will stop working. However doing it will lead to zero improvement in functionality for us. AWS has literally spent money on engineering that has had no effect but to waste our engineers time and subsequently p*ss me off. They are not alone. There comes a time in a product life cycle where it feels engineering dollars are being spent for the sake of it. Why not either downsize workforce, or give the guys and girls a very long very well paid holiday (if trying to keep them out of competitor hands)??? While our product has got so much better over the last year (primarily by ticking off the most crucial of existing paying customer asks), I want to stay aware that at some point it will be essentially “done”. Any further “improvement” will just annoy clients. Engineers love to say SAAS is alive and never “done”…but I think this may be largely talking their book I’ve seen it a lot in consumer (e.g. having to update the Uber app when you just want to get a car back from the pub once every two weeks…despite them having added no noticeable new functionality for 10 years). This is the first time I’ve noticed it so clearly in B2B. #shipsandshipping #maritime #founder #vc

121

7 Comments

This is a really interesting article about the eventual commoditisation of AI. We at Jiva.ai have been banging this drum for several years. All good technology eventually becomes like a life/business necessity, sometimes not always visibly so. AI is becoming a core component of almost everything you interact with. Which means those who know how to build and manipulate it (rather than package it and sell it) are the ones with the real power. AI Will Eventually Fade From View, By Design https://lnkd.in/daN4bzpu

27

5 Comments

Well said! As parents, we all want to ensure our children’s online safety, especially when it comes to popular games like Roblox. KidsNanny’s Screen Scanner feature adds an extra layer of protection by automatically detecting inappropriate content on your child's screen in real-time, whether it’s through conversations, in-game interactions, or suspicious offers. This ensures parents are immediately notified if any harmful behavior occurs, allowing them to take swift action. KidsNanny.ca - AI-powered Parental Control App

7

Here's how I use Claude.ai to SAVE HOURS: I have to create an R&D doc for UK Tax. I have loads of docs, files, writeups, tech specs, etc. So, I go to Claude.ai and upload those docs. I ask Claude if they understand the project. I then explain the situation: "I need help explaining the timeline of events and our work, invoices, team members' time and the technical achievements we made". I could write this manually. But it would take hours. Probs DAYS! 😫 Sifting through docs from 2022/23. But I have a timeline and answers to all my questions in MINUTES. And I have a solid document to send to my tax advisor, hitting all their points. I cannot stress how much time I've saved. AI isn't always about the "new and shiny". It can (and should) be used practically. -- What mundane tasks do you use AI for?

11

6 Comments

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 Comments

The hype cycle around Generative AI has been needling at me for some time, and I think Benedict Evans hit the nail on the head why: at the moment, it feels like SQL, not AGI. ChatGPT, Claude, Midjourney, etc. are fantastic tool structures for asking good questions (like SQL is), you just still have to know your use case, and you still have to ask the good question. SQL's a great tool for a financial analyst, but it won't inherently turn you into a solid investor. My career got built off SQL, as did many others, and Oracle has made a tremendous business with that technology as its backbone. Perhaps in the future we'll see lots of LLMs as a common tool in the research stack, in much the same role SQL plays now. https://lnkd.in/gKXbdbDh

2

Does Screenshotbot save on build times? This was a question that surprisingly came up a couple of times at #dcbln24. It probably had something to do with the large number of platform and DevEx folk at the conference. In any case, the answer turns out to be "Yes". More specifically: You'll save 19% in build times on average, and 49% in build times in the P95 case. The savings come from avoiding re-runs when screenshots change. Read my blog post with the analysis, and I'd love to hear your thoughts: https://lnkd.in/eEh4zqkN

6

With the rapid globalization of movies, series, and OTT content, the demand for high-quality subtitles has never been greater. Subtitles are no longer just a tool for accessibility—they are essential for connecting with audiences across different languages and cultures, ensuring that stories resonate globally. To explore this crucial element of content creation, I’ve recently written an article for The Broadcast Bridge detailing best practices for subtitle creation. The article dives into the key techniques and stages necessary for crafting accurate and effective subtitles, drawing on both industry standards and my own experiences. For anyone involved in content creation or interested in the subtitling process for global media, movies, series, and OTT platforms, please explore the article and share your thoughts. https://lnkd.in/giUx8FTV Special thanks to Manik Gupta and Sana Afsar for their constant support and guidance. #SubtitleCreation #TheBroadcastBridge #BestPractices #GlobalMedia #AI #NLP

21

2 Comments

Who would have thought PDFs would be the main source I’m turning into embeddings for firms looking to get accurate data to their LLMs? Big thanks to WillowTree for giving me the space to experiment and get real-time feedback, while I was there. It’s amazing to see how these technologies evolve and the new opportunities they create for efficiently extracting and utilizing business-specific data. #AI #LLMs #TechInnovation #ResearchandDevelopment #PDFtoEmbeddings #WillowTree #ArtificialIntelligence #DataEngineering #ForeverLearner

3

Poked around on some ideas over the weekend and ended up using the new OpenAI Realtime API to create a translation dictation app for improved conversational translation. I've found translation apps, even the new ChatGPT to be a bit difficult to use. Without creating a translator and instead transcribing those words down in multiple languages you can keep the conversation flowing. Check it out and look in the video notes for a link to the demo application which you can run with an OpenAI key! https://lnkd.in/gJ9M72-c #aitranslation #aicode #openai #realtimeapi #chatgpt

8

1 Comment

I recently finished a Novella by Brandon Sanderson title "The Emperor's Soul". It was written about 12 years ago but basically has a magic system where people can prompt-engineer the world. In this system, if you can understand something's history and make subtle tweaks to it, then you can change what it is now. The heroine's primary task is a complex one where she has to remake a person. The approach and complexity of the task struck familiar chords with how when developing an AI agent we have to carefully craft its persona and orchestrate its thinking processes for the variety of scenarios it may encounter as it interacts with users and the world. An enjoyable read and interesting parallels to consider. https://lnkd.in/dj9ybBc9

12

2 Comments

Anthropic just announced their computer use feature, and my mind is blown 🤯 Anthropic introduced upgraded Claude 3.5 Sonnet and a new model, Claude 3.5 Haiku. Sonnet delivers strong coding abilities, while Haiku now rivals Claude 3 Opus in performance, speed, and cost. But here’s the real kicker: computer use  Computer use is in public beta. Available through API, where Claude can now interact with computers like humans do:  - looking at screens - clicking - typing - and more. This is a game changer—next-level scraping, form filling, automation what will you build with it?

1

Cryptographic agility should be part of your mind set as you plan out your risk and threat mitigation strategies. To achieve “security by design”, it is crucial to consider this well in advance. Cryptomathic can help 🫶

19