Valerie Nilsson

proximity bias in AI, fueled by "ecosystems dominance" SV is great, but super-concentration of SV as well as near monopoly of ecosystems like iOS, Android and Facebook/WA - the world's best experiences will always come to rich US cities first to the people who are building it. Because the developers are all there and it's easier to build great products for themselves. Apple folks, by large are living in a 30 miles radius around Cupertino - so US101 traffic will be the first to be predicted, english will prevail and typical bay-area centric apps will be the first to have actual good intelligence. Surprisingly, even a super exclusive Greater China Market didn't stop Apple from building a moat there. Although the power of Apple there is much small (mainly hardware advantages rather than software - Apple pay is weak, no Apple TV/Music strength, macbook adoption is much smaller...) Regulation might be the only hope here for regions to jump ahead and break the moat, but it'll be real messy. Until then, let's first marvel at how great soccer dad/mom's saturdays would be with the help of Apple AI. https://lnkd.in/gDkQnfZc

6

I'm sure you've heard or read about the controversy surrounding Apple's iPad Pro "crush" ad that the company has now apologized for (below). It certainly captured huge attention, though it was not good for Apple's brand -- which used to be about enhancing human creativity. I can understand the (misguided) thinking behind it: we're packing all these tools and lots of creative potential into an incredibly thin machine. But the ad was inherently violent. Also it stands as a perfect metaphor for the ascendancy of machines, machine thinking and synthetic creativity over human creativity. What Apple neglected in approving this pitch/script/approach was the larger cultural context and cultural moment we're in. As someone said, this might not have been so offensive even a few years ago. It illustrates how vigilant and aware companies now must be about their marketing messages and creative against what's going on more broadly in the culture and society. I immediately thought of the disastrous Kendall Jenner Pepsi ad in 2017, which resulted in millions of dollars of losses for the company. Similarly, Apple may have tainted its new iPad Pros.

4

Interesting conundrum: A kebab menu is more visible using white dots on the blue background than black dots on the same blue background. But the contrast ratio is bad with the white (2.76:1) and good with the black (7.6:1). I’ve asked three separate folks with low vision to tell me which was better for them. They all say “white dots”. Granted, it’s a very small sample. Still. This leaves my client with an interesting choice: pick the option that is more accessible, but not standards conformant, or stick to the standards but end up with an element that is harder to perceive. What would you do? #Accessibility #WCAG

12

23 Comments

Have been thinking about two quotes recently! Sam Altman (In talk with Bill Gates): "These are the stupidest the models will ever be." Andrej Karpathy (Talking about AI Agent development): “There is a large class of problems that are easy to imagine and easy to build demos for, but extremely hard to make products out of. For example, self-driving: It’s easy to demo a car self-driving around a block, but making it into a product takes a decade.” ChatGPT came out 18ish month ago. We have seen a ton of cool demos, a lot of money has been invested in AI companies. But the products are not there yet. I had an expectation that we will see a lot of AI products in last Autumn. This Spring. It takes time to build an actual product around technology. But the big AI product wave is not there yet. Sure, there are plenty of "AI first companies". What are some AI products which you use besides usual suspects - ChatGPT, CoPilot, Midjourney, Perplexity/Exa? p.s. Funny enough, Linkedin gives me an option to "Rewrite with AI" below this text. Nah, thanks, I'm good!

43

5 Comments

Hot take on the OpenAI and Scarlett Johansson altercation? If the allegations are true, OpenAI clearly violated the privacy and ownership rights of ScarJo (and hence didn't comply to the demands of a particular individual). This, in my opinion, raises some major concerns on how seriously these proprietary LLM services take intellectual properties (IPs). Considering the solutions are a complete blackbox to the user, are their "data retention policies" trustworthy? Another reason why we at ScaleGenAI vouch for private LLMs– no shared compute infrastructures, no reliance on third-party services– LLMs fine-tuned and deployed on your dedicated compute environments (on-prem or VPCs), where you own and control your data security and ownership. #privatellm #genaisecurity #dataownership #ip #openaivsscarlettjohansson #generativeai

25

Remembering those days as an analyst in the 2000s, when so much time was lost on accessing, cleaning and standardising data. How technology has changed. Wonderful to see that risk analysts these days can truly focus on more value added activities, while financial institutions can uncover deep efficiency gains in risk. #creditrisk #creditrating #ai #fintech #saas

8

AI is rapidly transforming product development across industries, yet many companies still grapple with how to effectively integrate and leverage this technology. At Meero, we've gained valuable insights from our work applying AI to real estate, automotive, and fashion visuals. These experiences resonate with many of the counterintuitive lessons shared in a recent article from Lenny's Newsletter which gathered advice from over 20 top product builders on creating AI products. One key lesson is the importance of identifying high-impact workflows where AI can provide substantial time savings and invite repeat use. For instance, our AutoRetouch product automates fashion photo editing and creation tasks, allowing professionals to focus on higher-value activities like complex production shoots and creative direction. The initial user effort to adopt the tool yields significant ongoing rewards. We've also learned that proprietary data and industry-specific knowledge are becoming increasingly crucial differentiators. Generic AI models are becoming commoditized, but combining them with deep vertical expertise and carefully curated datasets allows for truly transformative applications. Our CarCutter product leverages our automotive industry knowledge, including integration with most dealership management systems (DMS) in the US and Europe, to deliver AI-powered vehicle image editing tailored to dealerships' specific needs. Additionally, we've found that thoughtful user experience design is critical for AI product success. The promise of AI to create anything from a simple prompt can be daunting for users. Providing the right starting points, visual options, and intuitive interfaces to guide users is essential. With ProperShot, our real estate visual content tool, we've iterated extensively on the UI to make AI-assisted property imaging more approachable and effective for agents and property managers. As AI capabilities rapidly evolve, Meero is focused on building flexible systems that can seamlessly incorporate model improvements. This allows our products to continuously enhance performance without requiring major reworks. Fanny Kauffer Maxence Bernard Steven Petitpas-Polselli François Pallier Clément Boutignon Ludovic Dupuy Christof Opresnik Alexandre Castera #AI #generativeAI #innovation #product

80

2 Comments

What is *agentic reasoning*? Most of us are pretty familiar with asking ChatGPT questions and getting answers back. But can we really build "autonomous reasoning" on top of what looks at first like very sophisticated "next token" prediction? In this very simple demonstration, I show what happens if you ask the GPT-4o LLM, via API call, what is the current time? It starts by saying that it doesn't know. Then I proceed to give my agent a "tool" which includes a function for executing system commands. That function has no special purpose, it's description says simply "Executes commands on the underlying linux system inside the container." Now the key is that I simply ask the same question again: "What is the current date and time?". I haven't given the agent any instruction about how to answer that question, but it _reasons_ that it can execute system commands to determine the time, and that is what it does. This logical reasoning sure looks way more advanced than "text completion". To top it off, I then remove the system commands tool, and give the agent a web browsing tool instead. Then I ask my same question again, and this time the agent figures out the time by browsing the web. Again there are NO special instructions, just basic functions exposed to the agent. #ai #supercog #aiagents

14

2 Comments

Three Things we learned during the past two days at All for Content 1️⃣ Generative AI is Everywhere but... nowhere: Two-thirds of the talks during the trade show were about Generative AI. BUT when I asked a room of 50 people who had a Generative AI use case in production, only 3 people raised their hands. Just 3! 😱 2️⃣ In the Generative AI Era, Branding is Key: The main question we received from visitors was, "How can I make my brand AI-ready?"  It's not a simple question, as a brand is defined by so many little (and big) details. This is precisely what we are working on at Newtone, and we are going to release a whitepaper on it very soon. 🤗 3️⃣ After the Excitement (2023) and the Disillusionment (2024) about Generative AI, here comes the Maturity: Only a few companies have successfully implemented Generative AI solutions in production, but the impact for them has been huge - see Klarna. They started with small, impactful projects that actually produced results. Some brands told us, "My IT team is implementing an AI that will fulfill all our needs 🤦‍♂️... "  Identify the right use cases, focus on easy implementation, and achieve huge impact. Thank you so much to the All for Content team for organizing this event. Special thanks to JULIE DARDOUR ✨, Anne Pinto 💬, Virginie VOVARD, Jean-Luc Prigent for our enriching discussions. 😊

44

11 Comments

In the continuously accelerating landscape of modernity, the ability to seize and streamline our cognition is ever more pivotal. I'd like to steer your attention towards 🌀 Vemo AI, a platform pushing the boundary within this dynamic environment. ➡️ What is Vemo AI? Think of it as not just a gear, but a highly articulate framework that improves how we accumulate and arrange ideas, information, and insights. 👉 It's a solution sought not just for individuals but corporations navigating the knowledge economy! My perception of this news teems with both anticipation and opportunity: 👉 The first is a prediction. With Vemo AI, Machine Learning and Artificial Intelligence aren't mere buzzwords. They're fundamental for setting the stage for new opportunities, intricate insight encoding, and evolving as per user behaviors. 👉 Second, as a content aficionado, the longitude such tools gift in an overly information-saturated world is noteworthy. As they will assist us in not only capturing and processing information, but more importantly, deriving meaningful interpretations from it. Their impact on all spectra of our personal to professional lives will continue to amplify: 🔹 For individuals, halting information overload and chiseling wisdom out of awareness. 🔹 For organizations, fostering culture of knowledge sharing and breeding innovation. I see Vemo AI, and inherently, AI-based solutions following this trend, act as protectors of cognitive capital. ☑️ Clarity in chaos ☑️ Productivity from deluge ☑️ Innovation amid information So, while others remain tangled in the unending labyrinth of distraction, with Vemo AI, have ourselves a savvier, more efficient, and insightful future. As I see it, it's not just something to embrace; it's something to be excited about! Tools like Vemo AI underscore the mantra: Information is not power, harnessing it contructively is! Here’s to embracing this new wave of cognitive empowerment 🥂. Have a thoughtful day! #VemoAI #AI #machinelearning #futureofwork #productivity #innovation.

4

1 Comment

The AI journey needs to be guided by purpose. And, as I write in an upcoming article the idea of "better" versus "more". https://lnkd.in/eDcjRpVv

3

This article has all sorts of reassuring quotes That is, reassuring me to continue to doomsday prep and prepare for a battle with humanoid terminators *changes name to John Connor* "AI interpretability is an especially pertinent topic, given the heated debate surrounding AI safety and the risks of having an artificial general intelligence go rogue, which to some experts represents an extinction-level danger for humanity." https://lnkd.in/g_U-92mk #AI #LLM #OpenAI

6

4 Comments

I just cancelled my ChatGPT subscription and moved over to Anthropic's new family of LLMs, here's why: 1) Context window and recall: ChatGPT is lazy (Google it), especially when dealing with large files. Claude 3 has a 200k context window (around 150,000 words) and achieves 99% recall accuracy when tested using the Needle in a Haystack evaluation. Meaning: It can digest ~150,000 words and remember ALL of them. 2) Computer Vision: One of the main drawcards to ChatGPT was the computer vision capability. I work with images + ChatGPT all the time. I don't know if Claude had this before but it does now! 3) Code: I only work with pretty basic HTML and CSS but I still use ChatGPT quite a bit for help with coding. From what I'm seeing, Claude's new models are SCARY good at coding. Creators are using it to make new video games with a single prompt (they actually work), functional web apps, and even simulations of solar system physics. I'm seeing the term "Sparks of AGI" starting to get thrown around again. 4) OpenAI Playground: If all else fails, OpenAI playground (GPT4o API) is just way better, cheaper, and doesn't suffer from the massive system prompt like ChatGPT. I won't be surprised if OpenAI announce something big very soon to get back in the spotlight (as they tend to do), but at this point, it's going to take some convincing for me.

5

I just published a new post "Innovating with conviction: lessons from Marc Pincus (Zynga founder)" https://lnkd.in/epTFRxWD We learn an important innovation framework from Marc, including: - The goal of innovation: getting to true signal - Develop your core idea thesis, and create multiple idea variants. Keep your idea variants separate from your idea thesis. - Cycle through your idea variants quickly to take the maximum number of shots on goal - Get to minimum viable idea state, not even minimum viable product, to test your idea variant Lots of practical innovation guidance from Marc Pincus, I've learned a lot and I'm immediately putting his framework into practice. I encourage anyone working on 0 -> 1 projects to do the same!

16

1 Comment

Our #AI chat bots are becoming more human. And this will change how we interact with them. It some cases it with change how we interact with the world. I’m just not sure we are ready… https://lnkd.in/gPJWPTve

1

How Clear Context Elevates Your AI-Generated PRDs After using ChatGPT for the past 1.5 years, I've discovered that the key to getting high-quality Product Requirement Documents from it, or any answers in fact,  lies in one crucial step: providing crystal-clear context in your prompt. Treat ChatGPT like your team members & stakeholders. The more context you provide, the better they can support you. Here's why it matters: 1️⃣ Accuracy: Detailed context helps ChatGPT understand your product's unique needs & situation. 2️⃣ Relevance: The more specific you are, the more tailored the output for you. 3️⃣ Efficiency: Clear input = fewer iterations = time saved! 4️⃣ Consistency: Comprehensive context ensures alignment with your larger product vision. 5️⃣ Garbage In Garbage Out: Wrong context can lead to inaccurate output. Remember: ChatGPT is a powerful tool, but you are the product expert. Your insights are the fuel it needs to shine! What's your experience with AI-assisted PRD writing?

11

3 Comments

I created an MGMT Boston GPT. I have no idea if it really works but what I do know is it is mostly giving me better answers than I can give myself*. It's trained on 100k+ words! Now that GPT access is open to all, you should check it out: https://lnkd.in/eE6kRNHv And please give me feedback. Better yet, you should make your own. The future is here! *The outputs could be a bit more live & actionable. If you're asking which of the startups I've highlighted are hiring...that does not work as well. Give me time!!

32

7 Comments

The launch of ChatGPT was a "Sputnik moment". In making tangible decades of progress it shot AI to the fore of public consciousness. This attention is accelerating AI development as dollars are poured into scaling models. And just like Sputnik, ChatGPT demarcates a starting line, rather than the finish point. What is the next stage in this journey? And where is the destination? I'm privileged to be joined by the brilliant Nell Watson. Nell offers a broad perspective on the possible trajectories. She sits in several IEEE groups looking at AI Ethics, safety & transparency, has founded AI companies, and is a consultant to Apple on philosophical matters. Nell makes a compelling case that we can expect to see agentic AI being soon adopted widely. We might even see whole AI corporations. In the context of these possible developments, she reasons that concerns of AI ethics and safety — so often siloed within different communities — should be understood as continuous. Along the way, we talk about the perils of hamburgers and the good things that could come from networking our minds. https://lnkd.in/gr68g5uC

16

1 Comment