TheDudeMan

Referenced in this paper:

"Overall, while approaches such as FNet, Performer, and sparse transformers demonstrate that either fixed or approximate token mixing can reduce computational overhead, our adaptive spectral filtering strategy uniquely merges the efficiency of the FFT with a learnable, input-dependent spectral filter. This provides a compelling combination of scalability and adaptability, which is crucial for complex sequence modeling tasks."

And a comparison section after that.

Maybe he was joking about the Roaster 2, also.

With Lost, I knew they had no direction in mind, and that bothered me. But I also knew that whatever they were doing, they were doing a damned entertaining job of it.

They are trying to convince the world of the problem. What else can they do?

> I don't think LLM systems will be competitive with everything that humans can do for a long time

It's looking like about 3 years. That will be another ~100x cost reduction, $100s more billions in infra, new training algorithms, new capabilities.

Did the radiologists suddenly get better than the AI at reading images? Or is the system simply unchangeable?

I'm guessing that's mostly wifi congestion.

What panels? What is this about?

I suggest improving the probing a little. Something like

  idx = (idx + 1) % map->cap;

becomes

  iterCount++;
  idx = (idx + iterCount) % map->cap;

What you're calling "gamed" could actually be research and progress in general problem solving.

Do you have millions of strings that need to be manipulated in the same way at the same time?

This is why peer reviews are so important. Amazon does it right.

Could you say a bit about how the poker AI works? I know it's not modern, but just curious. Thanks.

100% this. This would be a dream job for the right candidate, too.

Stop trying to mimic brains. Do what works best for transistors.

Weird that this isn't getting traction on HN. This idea is going to go far.

So all forms of search are dumb in your view?

They are moving to Discord.

Or wheels.

Those are unfortunate, IMO. But miners would support those because they increase transaction fee revenue. Hopefully they'll get priced-out of the fee marketplace eventually, either by the network getting more legitimate use (moving money instead of creating NFTs) or by people losing interest in NFTs.

$2.3T market cap seems like a lot for a dead thing. Market cap is crude metric for anything, but still.

https://coinmarketcap.com/charts/

Yes, it is a public ledger. This ensures that no inflation is occurring (other than the expected decreasing miner subsidy).

Most cryptocurrencies are controlled by a small number of insiders. Bitcoin's high level of decentralization makes this not the case. Although the increasing centralization of mining, mining pools, and mining hardware is becoming a concern.

Bitcoin's strength is censorship-resistance.

Same here. People are confused because the author gave two options:

  1. Bundle the .dll as a resource into the app and make it extract it to some temporary folder, then use P/Invoke and call the compiled .dll from within the C#/.NET app.
  2. Use C++/CLI.

And then a single "problem" paragraph that didn't make it clear which problems go with which approach.

What is the standard solution to this type of phish? This problem did not exist in the ASCII world, of course. Unicode is useful, but what is the best way to prevent this malicious use of it?

AI and bitcoin consume similar amounts of energy, but bitcoin uses more green energy than AI:

https://wired.me/science/energy/ai-vs-bitcoin-mining-energy

Is the problem the data center or is the problem carbon emissions?

How about tax the carbon emissions.

Which is better -- subsidize what you perceive to be the solution or tax what is known to be the problem (carbon emissions)?

(people sing a different tune when the power consumer is bitcoin miners)