The article vastly underestimates how often no was said before ZIRP. Getting to yes was very, very, very hard. Zirp moved the default to build it snd they will come.
We always knew the limits to Moore‘s Law are first and foremost economic. Given an industry used over decades to predictable lowering of price per compute function and thus swallowed any advance for new user functions and overhead when the limits are reached there is going to be a squeeze. AI scaled up at the time the production capacity became more inelastic.
Maybe it is time not just shrink transistors but also software bundles. I can see decades of possible progress hiding in plain sight behind a browser screen.
It would make memory-poor phones more viable. Like why can't we have a 512MB, or even 256MB RAM phone. Although I doubt that the software effort would be cheaper than just buying the extra RAM. It's definitely much more uncertain.
The OpenMoko Freerunner only had 128MB RAM, it was able to run a Linux desktop of the time, Englightenment/E16. There were lots of apps for it too. IIRC the cut down QtMoko distro ran best though.
That's newly fabbed memory vs. an existing stock. The stock is quite massive, so optimizing existing use and enabling it to be repurposed can be meaningful.
If the existing MacBook needs less memory, it can use its current memory spec for higher-level uses that formerly required a MacBook Pro. That meaningfully affects the market.
If your point is purely about supply and demand for datacenter HBM and LPDDR, you're probably right. Local model inference (using the existing memory stock) can make a dent in current use, but not in projected future uses that will plausibly involve much larger models.
There is a reason moderation decisions are not perfectly transparent: They are gamed otherwise. So there needs to be legal recourse with discovery and meaningful liability attached to submitting to the role of acting as the agent of a foreign government.
While I agree with adding code contributing to complexity is problematic there is lots of code in existing code basis which is overly complex due to past outdated requirements or less than perfect human coders. The current flood of AI driven security fixes demonstrates that AI can be pretty good in detecting security edge cases. It is not inconceivable to use it to also reduce code complexity.
Yesterday in Lidl I was a bit shocked seeing the coupons offered by their app. They did a really good job with their mixture of stuff I had bought or might buy.
reply