Zacharias Beckman

Picking up on what seems to be a recent post theme... In Goldman Sachs: Top of Mind report, Issue 129 25JUN24 (https://lnkd.in/db7PcUU9), the key question is what the return on an estimated $1T (one trillion dollar) CAPEX investment in the coming years will be. In short, and not to be too glib about it, there's a little bit of a hole in the ROI model for 'AI'. From the GS report: "Given the focus and architecture of generative AI technology today... truly transformative changes won’t happen quickly and few—if any—will likely occur within the next 10 years." - Daron Acemoglu, Institute Professor, MIT. Observing that the current architecture of AI technology itself may have limitations, he observes: "Human cognition involves many types of cognitive processes, sensory inputs, and reasoning capabilities. Large language models (LLMs) today have proven more impressive than many people would have predicted, but a big leap of faith is still required to believe that the architecture of predicting the next word in a sentence will achieve capabilities as smart as HAL 9000 in 2001: A Space Odyssey." Going further: "AI technology is exceptionally expensive, and to justify those costs, the technology must be able to solve complex problems, which it isn’t designed to do." - Jim Covello, Head of Global Equity Research, Goldman Sachs. (NB: There are more positive views in the GS report that express optimism that AI spending will eventually pay off. The words 'optimism' and 'eventually' are why I'm not quoting them here: the challenges presented are very much here and now and, as I have said before, hope is a terrible strategy!) Unrelated to the Goldman Sachs report, but with a similar theme, David Cahn of Sequoia Capital wrote in September 2023 that: "for each year of current GPU CapEx, $200B of lifetime revenue would need to be generated by these GPUs to pay back the upfront capital investment." By his update piece of 20 June 2024 (https://lnkd.in/dhJTZphm), that number had increased to $600B of lifetime revenue. For each year of current CAPEX. We know that technology investment at pace vaporises capital and creates the opportunity for lots of people to lose a lot of money but, as that same burn brings the incremental cost of accessing, utilising, and enhancing technology down, so it also creates the opportunity for new models and new ways of thinking. But maybe what is currently interpreted as new ways of thinking - as Acemoglu put it: the architecture of predicting the next word in a sentence - isn't where the true opportunity lies. With previously discussed costs to and of energy and water utilities, there's a very real question: will the next big winner in the AI space be utility provision and, if so, is that how the 'end-user' actually benefits? Philip Low Nicola Hayes George Rockett Kent Bibby Holly Joint Kaila Yates Hannah Daly Mark Feeney Cian Gill

6

1 comentario

I’ve been thinking a lot about this. The post below highlights a troubling trend: SaaS companies explicitly opting out of serving the EU market because of regulatory challenges. And honestly, I can relate. As an EU citizen, it’s often easier to start something targeted at the US market and distribute it there than it is to operate within Europe. The US market is massive, vibrant, and—crucially—easier to navigate from a regulatory perspective. On the other hand, Europe, with its patchwork of regulations, feels like a minefield. GDPR, while well-intentioned, often becomes a barrier rather than a facilitator for innovation. Take the cookie banners, for example—those intrusive pop-ups we’re forced to navigate on nearly every site. They’ve become a meme at this point, something everyone ignores or blindly clicks through. What’s the point of regulations like this if they alienate users and stifle innovation while doing little to actually protect privacy? We’re seeing innovation bypass Europe. How did it come to this? Where is Europe heading if the cost of compliance outweighs the reward of serving the market? It’s sad to see, but it’s the reality we’re living in. Unless something changes, I fear this trend will only accelerate. Europe needs to rethink its approach if it wants to compete on the global innovation stage. Would love to hear your thoughts on this. Have you had similar experiences, or do you see a way forward?

4

1 comentario

In the previous versions of the discounted cash flow, there were fixed endpoint constraints. Here I develop the equations for the floating endpoint discounted cash-flow.

2

Yesterday at Devoxx Poland in Krakow, I gave an updated version of my talk, How to Query a Stream. Herewith the tl;dr: Kafka has become the de facto data substrate of modern systems. It provides all kinds of benefits, but at a cost; to wit, it's very difficult to ask questions about what's in a log. There is no one way to solve this problem, so I proposed several distinct options. 1️⃣ You can always Kafka Connect a topic into your favorite RDBMS (what is it and why is it Postgres?), then query the resulting table. This might not scale to huge data sizes, but the infrastructure is something you're more likely to have around already. 2️⃣ If you're a Java shop, you can use Kafka Streams. It's straightforward to make a topic into a table queryable by message key, accessible through the Interactive Queries API. And if it's analytics you want, you can build an arbitrary stream-processing topology that again produces a queryable table. 3️⃣ There are sundry streaming SQL solutions like Materialize, Delta Stream, Rising Wave, and ksqlDB. They provide similar levels of stream processing and queriable materialized views. 3️⃣.5️⃣ It's here that Flink enters the chat! It famously does not expose its state store to be queried, but it's typical for the last-mile serving-layer-ish query solutions we're discussing to act like a key/value store. This means the computation that normally takes place in the Optimize/Plan/Execute flow of a regular database needs to happen somewhere else, and Flink often fills that role. 4️⃣ Cloud data warehouses. A big place to put all your data for the purpose of analytics. Like data warehouses of yore, these contain explicitly structured data. Oddly, the logos of all the common solutions are blue. I heard you like blue. 5️⃣ Data lakes. A big place to put all your data for the purpose of analytics, but for when you want to pretend structure doesn't matter for a little while. Open table formats like Iceberg are emerging because structure never doesn't matter. 4️⃣ and 5️⃣ are converging (in principle, it not in implementation) into the data lakehouse—a reflection of the decades-long struggle with the schema/no-schema question. 6️⃣ Real-time analytics. A specialized category of analytics infrastructure that ingests directly from Kafka and makes results available at extremely low latency and extremely high concurrency. A more and more necessary part of the stack as analytics escape the dashboard and become part of the end-user experience. We are collectively in a period of exploration of untamed territory between Kafka at the base of the new stack and the differentiated application code at the top of the stack. That code at the top should be the focus of most of our efforts as developers. The application stack was well-defined in the client-server/monolith era, but in the new streaming era, so much is up for grabs and yet to be defined. All cliches aside, it's a very exciting time to be a developer.

467

16 comentarios

Highlighting: "By some estimates, more than 80 percent of AI projects fail — twice the rate of failure for information technology projects that do not involve AI. Thus, understanding how to translate AI's enormous potential into concrete results remains an urgent challenge." (Rand) ----- What this is really saying is that most people (and companies) don't need AI, most should avoid custom AI, and most should carefully evaluate any tools to see if they add vs detract from their current situation. Beyond the AI bubble, most companies need NOTHING new or bespoke from 'Gen' AI. As Salsesforce pushes AI, take a pause, because you probably don't need what they or any other bigtech company is pushing for AI. ----- Rand Research: "The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed. Avoiding the Anti-Patterns of AI (James Ryseff, Brandon De Bruhl, Sydne J. Newberry) (Aug 13, 2024) "To investigate why artificial intelligence and machine learning (AI/ML) projects fail, the authors interviewed 65 data scientists and engineers with at least five years of experience in building AI/ML models in industry or academia. The authors identified five leading root causes for the failure of AI projects and synthesized the experts' experiences to develop recommendations to make AI projects more likely to succeed in industry settings and in academia. By some estimates, more than 80 percent of AI projects fail — twice the rate of failure for information technology projects that do not involve AI. Thus, understanding how to translate AI's enormous potential into concrete results remains an urgent challenge. The findings and recommendations of this report should be of interest to the U.S. Department of Defense, which has been actively looking for ways to use AI, along with other leaders in government and the private sector who are considering using AI/ML. The lessons from earlier efforts to build and apply AI/ML will help others avoid the same pitfalls. Five leading root causes of the failure of AI projects were identified: (1) industry stakeholders often misunderstand — or miscommunicate — what problem needs to be solved using AI. (2) many AI projects fail because the organization lacks the necessary data to adequately train an effective AI model. (3) in some cases, AI projects fail because the organization focuses more on using the latest and greatest technology than on solving real problems for their intended users. (4) organizations might not have adequate infrastructure to manage their data and deploy completed AI models, which increases the likelihood of project failure. (5) in some cases, AI projects fail because the technology is applied to problems that are too difficult for AI to solve. Rand: https://lnkd.in/gNS7rR5p ----- Salesforce: "AI Apocalypse: 80% of Projects Crash and Burn, Billions Wasted, RAND Report" Salesforce: https://lnkd.in/gu3ksAic #ai #aihype #aibubble #fakeai #genai #degenai #generativeai #degenerativeai

16

5 comentarios

This guy doesn't quite get it, some bullets from his "quiet quitting playbook" * Blocking out private calendar time - I advise teams and developers to do this to not be randomized and to have dedicated flow for their work * Artificially extending project timelines (claiming two weeks for one-day tasks) - Oh you mean pad our numbers so we estimate accurately? * Citing unclear specifications as blockers - never ending stories because of poorly defined specs/stories? * Making numerous small code refactors - Excuse me? All that tech debt that you said you were ok with us adding and promised to give us time to address that never happened? What about simple design, or the TDD loop? What about dealing with this big ball of mud? Or are you asking, "Why don't you just write it right the first time?" * Blaming build system issues - That the build takes 5 hours to run and non-deterministic test failures? * Claiming blockages from other teams - AKA external team dependencies are impacting delivery? * Using technical jargon about "race conditions" to justify delays - This guy obviously isn't technical. Ever tried to debug a "race condition"? * Deflecting tasks by requesting formal ticketing ("can you create a JIRA for that?") - I almost love this one the most. If I spin this another way, "Don't work on any item that I can't see on the Project Board please. https://lnkd.in/et4HYE9E

2

CPOs, CTOs: Step Up! Why CPOs and CTOs Should No Longer Hide in the Tech Corner ...and why investors and boards should actively engage them. I've written before about the attractiveness of software companies for investors. Unfortunately, this is no longer a well-kept secret in the investment community. The result is increased competition. Increasingly, the question is becoming: how can we create more value than our competing investors? Traditionally, the investor playbook starts with financial maneuvers like leveraged buyouts and acquisitions, followed by improving operational excellence in business operations. This includes enhancing financial transparency, budgeting, and optimizing go-to-market (GTM) strategies. While these measures are powerful, they are not particularly unique to software businesses and no longer provide a significant competitive edge. In the quest for differentiation, we can no longer ignore the core value creation of software businesses: the creation of digital products. Enhancing the way we design and build products will become a critical factor in distinguishing our investments. Product and technology strategies must inform overall company strategy. This shift seems like a no-brainer when you consider that technologists lead some of the most successful tech companies. This trend puts you, dear CTOs and CPOs, in the spotlight. Unless your CFO deeply understands the value of a CI/CD pipeline (and can optimize it), it is up to you to create that decisive value-add for the business. Generic operational excellence, which also benefits industries like retail, can be driven by other C-level executives. However, when it comes to the craft of creating software products, you are the experts. As systems thinkers, you understand how various elements within the enterprise converge to achieve hyper-productivity, and you are the ones who can make it happen. For investors and boards, this means you might want to consider making your technology executive the CEO. At the very least, it is crucial to establish and maintain a strategic dialogue with the tech executives in your portfolio companies, not just the CEOs and CFOs, as is often the case today. #TechLeadership #CPO #CTO #ValueCreation #DigitalTransformation #InvestmentStrategy #TechStrategy

7

1 comentario

This is one of the points that I think we're losing when we talk about university educations--instead of training people on today's technologies, the goal *should* be to teach students how to *think*. Philosophy. Critical thinking. Deriving from first principles. And so on. (And I say this with some chagrin because I criticized the CompSci program at my alma mater (UCDavis) for exactly the same thing.) Truthfully, a student needs a balance of theory and practice, learning to navigate between the abstract and the concrete on a regular basis. It's not enough to just learn JavaScript, you have to know some of the ideas and patterns around dynamic languages (which are then applicable to Python, Ruby, Lua, and others). It's not enough to learn relational theory, you also have to see how this maps to SQL in a popular RDBMS (SQLite, Postgres, MySQL, and others). Frankly, I suspect that there's a moment coming in academic circles where universities start to recognize that they're really not job-training-systems, and start edging away from "everybody should be in STEM" and starts leaning back towards "We're trying to teach you theories and abstract thinking", which may even (gasp!) require a few liberal arts philosophy classes. Bootcamps could pick up the lack, but I suspect academia might choose to offer new/different degree programs to try and get to that "practice-centric" idea instead. Maybe. In the meantime.... If you learned your coding at a boot camp, start asking yourself where you can get the theory. If you're currently enrolled at a university, start looking at your classes and trying to see the theory and practice as separate, and how the two relate to one another. On a personal note, this topic also touches on why students complain about me "ranting" so often: It's because I'm trying to teach you some of the theory (and history) of mobile app dev, and not settle for 100% practice. (I spent an hour on the idea behind MVC, software design, and cognitive complexity the other day, ferchrissakes!)

22

5 comentarios

I did a quick test of the new function calling in #ollama using llama3.1 8B model. The good news is that the API works! The bad news is that llama3.1 does a poor job at formatting function args according to my tool spec. A comparison w/ gpt-35-turbo: https://lnkd.in/g_R3ExQn Anyone had success with local SLMs and function calling?

17

Want to get the most out of Claude Sonnet 3.5? Code generation is huge for LLM’s right now and there’s a little hidden feature I came across when sharing your code with Claude through the UI. If you use the markdown prefix for a code block with the language of code then it will be correctly formatted and Claude will have an easier time reading it. Here’s the example I used in the video: Can you explain this code? ```css As soon as you hit space it should format the box for you to enter your code. This should work for most languages supported by markdown. This does not currently work in the mobile/ipad app and works best on browser. Enjoy and happy coding!

10

8 comentarios

If the CDN market will grow 3%, it may mean the market has STOPPED GROWING (~3% is rate of inflation). That seems like a big story, and maybe an indicator that the internet is at a pivot from stored file delivery (CDNs, video streaming) to more real-time traffic and compute.

31

8 comentarios

Back in the early 2000s, during the dot-com boom, many dismissed the cloud as overhyped. They said it had no real business use case. “Just a gimmick,” they claimed—a toy that would never fly in enterprise IT. Big players like Oracle, SAP, and IBM agreed. It quickly became a buzzword. Companies and their IT vendors were dabbling in the cloud but not diving in fully—except for Microsoft, who really understood the potential. I got my first job in a SaaS business in 2007. Back then, most people didn’t even know what SaaS stood for. At that time, the cloud was just beginning to gain traction at the enterprise level. Salesforce was one of the pioneers to go fully cloud-native, and the rest of us followed their lead. Meanwhile, an online bookstore and a certain search engine decided to go all in on cloud. They won and became some of the biggest companies in the world (you know who they are). To me, the early conversations around cloud sound a lot like the chatter around GenAI today. About a year ago, we at Ellie Technologies faced an important question: how much should we invest in embedding genAI into our platform? More so since Ellie began its journey as a modern take on a data modeling tool. We're a lot more today, partly due to our acceptance of genAI.

13

1 comentario

Conventional access control methods are built on a “least-privilege” strategy, which is hard to achieve. Zero standing privilege is a newer and more effective strategy. Read how it can lead to radically diffedifferent outcomes in my latest blog post here:

9

When building RAG and LLM apps, the devil is in the documents. One of the hardest and least appreciated pieces of any RAG pipeline is turning complex visual documents into LLM-ready data. I'm talking about financial, health, insurance, gov docs and more. If they have tables, charts, diagrams and so on, it's a recipe for downstream hallucination. So how you fix it? This episode of RAG Masters explains the theory and then looks at PyPDF, Tesseract, LlamaParse, Unstructured and EyeLevel's X-Ray. See who comes out on top. https://lnkd.in/eTJw4hgX

10

From my experience with domain partitioning, I can say that a lot of today’s applications should add a service dedicated to wrapping LLMs. Rather than trying to plug LLMs everywhere, it is cleaner to wrap the LLM APIs into a service wrapper.

5

wow well while hallucination seems to be getting better with newer models apparently you just have to ask the right question as gpt-4o just gave me one of the wildest hallucinations Ive seen LMAO. So what happened was, I noticed that there are 4 musicians on the famous jazz album "Bolling Rampal - Suite for Flute and Jazz Piano ***Trio***" so I was curious about who the 4th musician was in the "Trio", his name is Max Hediguer and I cant find much info about him on the web so thought I would ask chatgpt. I phrased the question "what does max hediguer do in the bolling rampal suite", as normally feeding it well-known keywords like "bolling rampal suite" is more than enough to ensure the model knows what your talking about, but this is the wild response it gave me, note that none of this actually exists: Max Hediger is a character in the "Bowling Rampal Suite," a fictional work. As a character, he is depicted as an enigmatic and charismatic figure who plays a pivotal role in the story. In the suite, Hediger is known for his exceptional skills in both bowling and music, particularly with the Rampal flute. Here are some key aspects of Max Hediger's role in the "Bowling Rampal Suite": Master Bowler: Hediger is renowned for his prowess in bowling. He combines precision, strategy, and flair, often outperforming his opponents with ease. Musical Virtuoso: Apart from bowling, Hediger is a gifted musician. He is adept at playing the Rampal flute, a unique and complex instrument that requires great skill and dexterity. Mentor and Leader: Throughout the suite, Hediger acts as a mentor to younger players and musicians, guiding them with his expertise and experience. He leads by example, inspiring those around him to strive for excellence. Strategic Thinker: Hediger is not only skilled in his crafts but also a strategic thinker. He formulates and executes plans that give his team an edge, whether in a bowling match or a musical performance. Charismatic Presence: His charisma and charm make him a central figure in the suite. He effortlessly commands respect and admiration from his peers and followers. Conflict Resolution: Hediger often finds himself in situations where his diplomacy and conflict resolution skills are put to the test. He navigates challenges with tact and grace, maintaining harmony and balance within the group. Overall, Max Hediger's multifaceted talents and leadership qualities make him a crucial character in the "Bowling Rampal Suite," driving the narrative forward and adding depth to the story.

1

1 comentario

For all those engineers out there building AI agents. You need to check out this #github repo. This is a great team tackling an important set of problems that engineers face, and this solution is significantly better than Langchain's approach. #ai #agents #ML

9

1 comentario