I've Used Fable and I Know AGI Is Here
It even managed to fix some bugs
Ok, that was just a catchy title to get the algorithm's attention.
I have been using Fable, that's true. I used it for some difficult tasks where the other models at my disposal simply couldn't get the results I wanted.
And, yes, Fable did well. It's a very accurate model.
Fable managed to do a better job at the things I felt the other models were struggling with. Which is exactly the kind of incremental improvement you'd expect from any technology.
You buy the newest iPhone, and you expect it to be a bit better than the previous one. You expect the things that used to bother you to have been worked out.
Does that always happen with any new version of a product? Of course not. But isn't it what we expect from something we're paying good money for? Absolutely.
So I guess I'm just confused by the hype. People are marveling at the fact that this newer, wildly more expensive model is actually better than anything we've had before. But why? What did we expect? A worse model?
At this point there's nothing magical or hidden about LLMs. We know about the Scaling Laws, reinforcement learning, pre-training, post-training, harnesses, the lot.
It's all fairly well known, which is why smaller labs are coming up with their own models all the time, and why open-weight Chinese models are catching up fast.
Despite the hype, and as far as we know, no new technique or weird black magic was used to train Fable. It's just good old, incremental, solid engineering work. If you're the lab that came up with Opus, with some time, money, and talent you should be able to come up with Fable.
This is all super interesting from a tech and engineering standpoint. LLMs are genuinely exciting. But unfortunately we're being distracted away from that admittedly nerdy conversation in favour of a manufactured, end-of-the-world, panic-driven one which is scaring a lot of people away from even giving AI a try.
And then you have the attention-seeking crowd on social media talking about every new model like the Messiah has just come.
None of it is helpful, none of it is productive.
For those of us who actually work with these technologies, it's incumbent upon us that we offer a more balanced, optimistic view. We know there's nothing crazy about LLMs, it's just computers. It's a very interesting, powerful, exciting technology that can and should be used for good.
And yes, the Dario's of the world will try and push all the buttons to make more profit from it than is reasonable, but that has happened before.
And yes, you can be both greedy and brilliant at the same time.
🔥 Sponsored by QuickTable
QuickTable is the refreshingly simple way for restaurants to take online bookings. No clunky dashboards, no endless setup. Just clean, intuitive design that works. It’s built with the same technical principles we celebrate in this newsletter: simplicity, resilience, and focus on what really matters.
And here’s the best part: friends of this newsletter get a full year free. That’s right, twelve months of effortless booking, on the house. Don’t miss it, use this signup link.