Emmanuel Egyam, CFA, FMVA®’s Post

Should AI Prompst be engineered or designed? I read Alex Klein's article about designing AI prompts, and I agree with him. The best results come from a series of iterative prompts, defining the process, and, in my case, using the same research skills I use for surveys and interviews. I'm interested in AI as a new technology for research. Not that I'll run my data on ChatGPT or others alike, but there are good AI tools out there. They will only get better. What is your approach to using AI and creating your prompts? https://lnkd.in/g4e_XVN2

Insightful article! I agree that designers and the iterative design process are crucial in creating AI experiences that are usable and valuable. Thanks for sharing the article Adriana Grande.

The Golden Age of Building in AI Imagine if we kept adding more engines to a car to make it go faster, and at some point, adding more engines didn’t help much. That’s what’s happening in the world of AI right now, according to Ilya Sutskever, one of the biggest brains behind modern AI like ChatGPT. Here’s what he said: 💬 “The 2010s were the age of scaling—just using bigger and bigger data and computers. But now we’re back in the age of wonder and discovery, looking for the next big idea.” This means just “making things bigger” in AI isn’t working as well as it used to. Even OpenAI, the company behind ChatGPT, has reportedly been a bit disappointed with the results of their latest AI model. But here’s the exciting part: This might be good news! Instead of rushing to keep up with new AI tools every month, we can focus on making the current technology really useful. It’s like realizing you don’t need a faster car if you can use the one you already have to build amazing things—better maps, stronger bridges, and even cooler gadgets. Even if AI stops getting smarter tomorrow, we’d still have at least 10 years of opportunities to make the world better with the tools we have today. That’s why I believe we’ve entered the golden age of building—a time where creativity, problem-solving, and teamwork will shine brighter than ever.

630 days ago we didn't believe AI could read and write like a human. Today the full of "AI Hype" - how did we get there? We had never seen anything like ChatGPT. Interacting with a customer support bot was it's own special brand of torture. Suddenly, we had a bot that understood what we wrote and wrote back in a more coherent manner than most of our colleagues. It blows my mind how quickly we get used to new technology. Even if AI can't ever do more than now, it is still incredible. If you could go back 630 days to tell yourself about what ChatGPT can do today, I doubt your slightly younger self would believe you. Despite how mind-blowing this new tech is, so many people and media are yelling HYPE. Several articles describe how we're at the top of the hype curve. Now there are tons of people claiming AI can do more than it can. Obviously, those deserve to be called out. But I don't really believe we're anywhere near where AI will go. Mostly because of three questions: - In the past, it has taken us decades to understand the real value of new technology. How are we supposed to know after only 630 days? - Look at the first release of GPT; how does it compare to GPT 4o? To any of the other models we've seen released in the past 630 days?  - Why would you expect that speed of improvement to slow down or stop? Which other technological developments have slowed down or stopped after just 630 days? I don't know if it's all hype, but I still haven't found any arguments that hold up to scrutiny. What are the best arguments you've seen for or against AI hype? ------- ♻️ Repost this to help others in your network. And follow Torben for more

Happy to be featured in Data News, where I talk about why AI isn't delivering the return we expected and how we can fix that. For individual use, tools like ChatGPT show promising results. But when we try to implement these tools across the whole company, we often don't see the same return. Why is that? There's several things at play, which I highlight in this article. I don't think AI is overhyped, but we're not implementing it right. I'll add a link to the article in the first comment.

Thank you, Bernardo Kastrup, for articulating what many of us have intuitively understood about LLMs like ChatGPT. Your honest and insightful reflections, backed by deep technological knowledge, are invaluable today. AI's ability to convincingly mimic human intelligence, creativity, and now even emotions, is indeed powerful. However, those familiar with AI understand that this is merely a very convincing simulation. And as LLM's become ever more convincing, we need to keep that in mind.

I've been reading and experimenting with the updates of LLM models. As OpenAI states, there are 5 levels of progress: The levels of AI progress according to OpenAI:  Level 1: Chatbots, Al with conversational language – we are here  Level 2: Reasoners, human-level problem solving  Level 3: Agents, systems that can take actions  Level 4: Innovators, Al that can aid in invention  Level 5: Organizations, Al that can do the work of an organization And as Allie K. Miller says, you can use different LLMs for different projects. The theme for me is that humans can optimize their personal performance by collaborating with AI, even it that means securing "reasoned" responses. .... Practical AI for you and your teams 😎 I'm Marti, a workplace futurist available for speaking, workshops, and training. Next CoLabs

I love AI, but I often remind people it is not foolproof! To illustrate, I recently asked ChatGPT to generate fictional data on marketing and sales so I could run a regression analysis to see the impact of marketing efforts on revenue. One of the fields was "Campaign Type", which is categorical. ChatGPT provided the code, however, I quickly noticed it missed a crucial step - which is transforming the categorical field into numerical form for modelling. When I pointed out the missing step, ChatGPT responded as shown below. This highlights the importance of always double-checking AI outputs and combining them with your own expertise!

How to make an AI actually respond in a specific # of words. Here’s an all-too-familiar problem many of us run into with chatbots: 1. You ask ChatGPT to write a 100-word email → it coughs up 72 words. 2. Then you say, “Make it longer” → it overshoots to 128 words. 3. Finally, you say, “Not that long, go shorter” → it drops to 58 words. ChatGPT’s response length varies based on the adjective you use to describe your desired word count. Not sure who would ever use the word voluminous (sounds like something Voldemort would say?), but keep this chart of adjectives in mind next time you need to fine-tune the # of words an AI churns out.

Muse for today. I've been contemplating sharing my thoughts on the two rules of AI, considering ChatGPT as a large language model trained on web data. The number one rule is that any AI tool should save time. Sometimes, these AI tools give biased information, but despite their limited knowledge of the world, they make you productive and increase your value. Also, AI tools should not add unnecessary complexity. As a senior engineer, you know what you want before AI writes it, and you can detect when it is giving you bad guidance. That is one differentiating factor between senior and junior engineers. Companies value these sets of people any day as they prevent complexities of a system, business, or program. What are your thoughts on this?