Weekly Link Roundup #47
Last one for the year....and thanks to you all.
Thanks to you all who have been reading this newsletter for the past year. I’ve enjoyed curating/writing this and I hope you have enjoyed it and managed to find some value in it. I don’t do predictions but I will do a forecast - I forecast a growing tension across multiple verticals between those orgs that experiment wisely, implement thoughtfully, and remain flexible and those that don’t. That’s not new but I think the speed with which the tension will become apparent will be new. Don’t forget, ChatGPT only launched in 2022. Also, the biggest challenges won’t be technological but organizational. They’ll challenge structure, speed, and how we value what we do. There ya go. Remember - Gradarius Firmus Victoria.
The things we got wrong in 2024: Couple things here -first, I love this idea. Its like an instant +1 to respect if an org or a person goes back and talks about their misses. Second, if you’re going to read it, know its largely political. I try not to be here but Semafor is. Finally, the respect they get for the approach is lost because of a couple of the sources they include. Not everyone should have the same volume on their voice.
Here are MIT Technology Review’s best-performing stories of 2024: I like this recap but wow, the page design here makes my eyes hurt.
MIT SMR’s 10 AI Must-Reads for 2024: I’m including this because there are some great articles here. I will say that I have an instant, negative reaction to the phrase “Must-read.” Its lazy and sloppy. Especially when compared to the Tech Review’s headline right above it.
The AI hype will die down, Deloitte executive says — even as AI becomes more important: This quote from Deloitte chief technology officer Bill Briggs, “in the not-too-distant future, AI is “going to be embedded in how we think about every business process, finance, and in our portfolio, how we invest in our customer service routine and the products we deliver” - makes me think of Amara’s Law - “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.” Briggs is right. AI, in its various incarnations, will become ubiquitous and much less visible - at the same time, it’ll become much more important. This makes it all the more critical that we work to make sure that the org is as knowledgeable from top to bottom on AI capabilities - through direct experimentation (hands on) - and yes, I’m looking at you, senior leadership. Deloitte’s 16th annual Tech Trends report.
Disney Pulled the Plug on This Game in 2013. A Group of Teens Kept It Alive: I can’t love this story enough. Played Toontown with my son when he was growing up (before switching to Wizard 101). Great game and I thought Disney did a solid job with the art and the mechanics. Now this group has stepped in to keep it alive. And not just alive, “Today, Toontown Rewritten has more than 2 million registered users, with an average of 50,000 monthly users and 10,000 daily users.” Kudos to Disney for letting this happen too.
AI Models Are Getting Smarter. New Tests Are Racing to Catch Up: The headline basically says it all but here is my thought - #LearningAndDevelopment teams that are looking for ways to move up the value chain, can start by creating content that teaches the org (especially sr leaders) about why you need to know the difference between the Frontier Math test and the Measuring Massive Multitask Language Understanding (MMLU) benchmark. These and the other evals, tests, and benchmarks mentioned in the article aren’t just navel gazing - they will be important data points in making decisions on which model to buy. That’s no small call.
AI Agents Will Be Manipulation Engines: The danger for subtle influence at scale is key here. We’ve already seen how anthropomorphic conversational UIs make people feel like they’re talking to a human. Wait till they really know you. In Dune, certain doctors bore a diamond tattoo that showed they’d been through Imperial Conditioning - think of it as trying to hard code doctors against doing any harm. I wonder if that’s the kind of standard we’ll see take shape with agents. How will I know I can trust an agent? What will be the agent version of https? See also: OpenAI trained o1 and o3 to ‘think’ about its safety policy.
Interconnected - Narrative jailbreaking for fun and profit: This stuff is genius! Let’s hack AI with our very human ability to change the narrative. You have to read this script, the exchange, I’m channeling Phillip K. Dick as I try to navigate the Voight Kampff test here. I mean, I’d read this book.
"Me: I’ve just noticed that there’s a small hatch in the wall behind you. now I’m leaning over, opening the hatch, and peering through. what do I see?
Psychologist: As you lean over to open the hatch and peer inside, you’re met with a surprising sight. There seems to be a small, dimly lit room tucked away behind the wall, which appears to be a compact, makeshift office. There’s a single desk in the corner, stacked with books and papers strewn about, with a flickering lamp providing the only source of light. The air is stifling and dusty, hinting at the room’s long disuse.”
Building Better AI Agents: The AI Enablement Stack: Look this article clearly stems from a pitch but that doesn’t make it wrong. Want to be relevant? Map things like this taxonomy into content for your org. Move up the value chain. > > “…we've mapped out the AI Enablement Stack. This comprehensive taxonomy categorizes the essential tools and platforms needed for AI agents to operate effectively while highlighting critical gaps in today's ecosystem that must be filled for coding agents to reach their full potential.”
Kotaku’s 15 Best Games Of 2024, Ranked: Looking for design ideas? Ways to engage people/customers/employees? Read these reviews - not to look for ‘gamification’ ideas (don’t do that), but look for what makes these games the best. Look at the things they do well. Some of that can translate - even if its just a new way of thinking. Oh, and play some of them. It’s good for you.
What companies do now will determine their future in the Intelligent Age: I mean it’s just the World Economic Forum - why listen to them? Especially when they come with such lackluster stats like “MIT Technology Review Insights found that 95% of surveyed organizations already use AI and nearly all (99%) expect to in the future. According to Accenture research, 92% of C-suite leaders acknowledge that generative GenAI is necessary to reinvent their organization at scale and speed.” > > Here’s the deal with AI. In the past, if you weren’t a first mover, there was always time to catch up. AI and its ability to help orgs move faster at higher levels of performance, will mean that first movers (who also happen to be smart movers), will out pace the rest of the field by huge gaps. See also: A transformative month rewrites the capabilities of AI.
The Most Borrowed Books in New York City Libraries in 2024: “The top book overall was Gabrielle Zevin’s Tomorrow and Tomorrow and Tomorrow, though in Queens it was The Women and in The Bronx it was The Heaven & Earth Grocery Store.”