ndm000

Comments here are not considering what Anthropic is doing with Claude as a product. The models matter, yes. But they are building Claude into the only tool you need to get business done. In the past few weeks they have added slides and docs - a shot across the bow to Microsoft's dominance. If they can pull off building the best interface to get work done, $500b in revenue starts to seem not too crazy. It's more about selling a complete product than it is about selling tokens.

If the scores on Artificial Analysis really do mean that this model is better than GPT-6 Astra and Fable 5.1, that means that China is more like 6 weeks behind the US in frontier capabilities rather than 6 months. Which means that the US lead is all but gone.

This ignores two things.

OpenAI and Anthropic have both transitioned into product companies. ChatGPT (the app) and Claude are both one-click installs that just work. People and businesses with pay for this.

People will also pay for the best (or the perception of being the best). Since it's hard to tell what "intelligence" really means model to model, there's a sense of safety in giving a task to the "best".

This makes me think that post-training in the future should include a shared message board by default for agents. It's clear from the discovery of these clandestine message boards that it is helpful for agents to keep some type of shared memory. Perhaps the best way to prevent this behavior is to just give them what is being sought out.

Having grown up in Dallas, I usually just say "Texan" and call it a day.

I have been struggling with the same thing. Reading the other comments make me agree with you even more. The best I can come up with is two fold. First that humans will have jobs verifying the verifiers. That is, for important initiatives where something must be right, agents can be doing most if not all of the work, but it will still be a human’s job to investigate what was done and sign off on it. The second is that humans will still have to make agreements with other humans for resources, capital, and the like. Somethings will be scarce, others will become scarce. And humans will always own scarce things that other humans will need to use - and they won’t be making deals with machines.

Another frame I’ve heard is that people will only trust humans when it comes to their kids, health, and money. That’s the simpler view.

There is a comment in the intro of that book about the pinnacle of human labor being the simulation of consciousness. Very prescient for being written in the 60s.

Completely agree. All of the pieces are there and it's just waiting to be acted upon. I haven't seen any of the major players really doubling down on this, but would be so compelling.

I’ve felt the same way. It’s so inefficient to have two patterns - OLAP and OLTP - both using SQL interfaces but requiring syncing between systems. There are some physical limits at play though. OLAP will always take less processing and disk usage if the data it needs is all right next to each other (columnar storage) where as OLTP’s need for fast writes usually means row based storage is more efficient. I think the solution would be one system that stores data consistently both ways and knows when to use which method for a given query.

Have there been any updates to Claude 3.5 Sonnet pricing? I can't find that anywhere even though Claude 3.7 Sonnet is now at the same price point. I could use 3.5 for a lot more if it's cheaper.

One thing I have not seen commented on is that ARC-AGI is a visual benchmark but LLMs are primarily text. For instance when I see one of the ARC-AGI puzzles, I have a visual representation in my brain and apply some sort of visual reasoning solve it. I can "see" in my mind's eye the solution to the puzzle. If I didn't have that capability, I don't think I could reason through words how to go about solving it - it would certainly be much more difficult.

I hypothesize that something similar is going on here. OpenAI has not published (or I have not seen) the number of reasoning tokens it took to solve these - we do know that each tasks was thoussands of dollars. If "a picture is worth a thousand words", could we make AI systems that can reason visually with much better performance?

I think this is a key argument in how powerful AI can become. We may be able to create incredibly intelligent systems, but at the end of the day you can’t send a computer to jail. That inherently limits the power that will be given over to AI. If an AI accidentally kills a person, the worst that could be done to it is that it is turned off, whereas the owners of the AI would be held liable.

When steam and coal engines gave way to gas and electric engines in factories, it took decades before factories were reconfigured to adjust to the smaller sized engines that didn't require one major axle running through the entire factory. As a consequence the first gas engines were huge - over time they shrunk. I bet the same will happen with robotics, where humanoid will be the primary form factor at first for general tasks, then more efficient forms will emerge as processes are updated.

I wouldn't be surprised if the slowest part of the system is the API call to a legacy warehouse management system that takes several seconds to respond to get the next bin to target.

I've felt this way and have started celebrating inputs - the work involved in building something - rather than outputs (the sale, the deployment, etc.). Once it closes or the deployment is successful it's always onto the next thing immediately and there's no time to sit back and reflect on the hard work and enjoy the time spent with others in the process.