What's in a World Sim?
World simulators are universally fun as hell. So many great video games fall into this category. Stone cold classics like SimCity and the Sims, Civilization and MS Flight Simulator, and the tragically under-appreciated Spore in which you take the reins on natural selection itself (OK, Spore isn’t an amazing game, but as a world simulator it’s fantastic). Or take modern epics like Grand Theft Auto and Red Dead Redemption — and the transcendental Goat Simulator. Playing a world simulation game is like taking a break by sitting on the throne of God.
World simulators extend beyond gaming. I would categorize Bartosz Ciechanowski’s Mechanical Watch simulator as a World Sim. Yes, it's a blog post, also a pedagogical tool, and a neat web technology demo — but Bartosz's visuals are based on legitimate simulations of mechanical behavior. Virtual springs and gears in his simulation operate a lot like the real thing. The spring mechanics in particular are a joy to interact with because they feel so real. If I needed to quickly get up to speed with the functions of a mechanical watch, I can’t imagine a better approach short of playing with a physical example (incidentally, some 3D printed mechanical watch simulators can be found on AliExpress or built in Lego).
[ Keep reading ]Who will be first to charge a Salary for an AI?
Subscription software has been on a steady trajectory to overtake the entire software market. I don't have data handy, but it's easy to run through landmark software products, from Microsoft Office to Apple iTunes/Music to Adobe Photoshop, and map out the transition. Once upon a time, we purchased these products once, up front, and we'd get a CD-ROM or MP3 file in return; today they are all subscriptions. Video games have resisted to some extent (Apple Arcade and Google Stadia, R.I.P., are not world beaters). But when is the last time you purchased a disk for anything else? The subscription model is so common that one might say it has reached its middle age — and we can start to think about what comes next.
AI is the latest craze, so perhaps we can learn something about future business models from AI startups. However, if we look around, AI startups are consistently following the established model: charge $10-100 per seat per month for a suite of AI tools, accessed by web & mobile apps — like a SaaS in all ways but the underlying AI technology layer. I imagine that smart minds at these companies see subscriptions as a comfortable transition strategy to something new. And, if AI reaches even a fraction of its enormous promise, that something new could be much much, more expensive.
[ Keep reading ]I just want a ChatGPT to teach me something
OpenAI has a new model. But it's not GPT-5 (please, please, we just like how it feels increment a number!). It's called "o1-preview", and it's a GPT that's fine-tuned for chain-of-thought reasoning.
On one hand, chain-of-thought prompting has been possible all along with GPT-3+, and is widely used. Heck, you can find it on site called LearnPrompting.org.
But OpenAI has taken things a step further by (1) training this model to use chain-of-thought all the time and, (2) more importantly, hiding the chain-of-thought from users. We do get to see a kind of summary of the reasoning steps (at least in ChatGPT), but not the gory details of the LLM output. This might provide OpenAI some protection from other models training on o1 output, but it's a bummer.
I don't have access to op1 via API, but I do have it in ChatGPT. And it has no problem with the classic strawberry problem (LLMs struggle here because tokenization swallows letter-counts):
So, at least we have that. I like to challenge these models to teach me something; let's try that.
[ Keep reading ]