tadala

Everyone wants to use less compute to fit more in, but (obviously?) the solution will be to use more compute and fit less. Attention isn't (topologically) attentive enough. All these RNN-lite approaches are doomed, beyond saving costs, they're going to get cooked by some other arch—even more expensive than transformers.

Is this a popular view? I think mathematicians can be odd, but usually they communicate quite well. I think as far as popularization of their fields go, mathematics is probably doing the best out of the lot: numberphile, 3blue1brown etc.

I think this is a little bit of a stretch. They have been using transformers for search since the paper came out—which is partially why they dominated the space, and also at this point they're caught up with OpenAI. Google simply had to be more careful about releasing something like ChatGPT because they're a public company.

They're the only company that is fully end-to-end (not Apple anymore, sorry Steve): energy, silicon, data, cloud, also serving to both consumer and enterprise. Frankly their only problem is the sentiment towards them, and their juvenile, cheap branding (though the tech-illiterate philistine managers they keep importing are of no help).

I had an idea that they should rebrand to E-corp and go all black.

You could fill a form and request them not to train; they usually approved it fairly quickly, but did not advertise it well enough!

Shocking comment. Do you think the scientific method is inbuilt in our DNA or something? Where do you think it all comes from?

On the other hand, the other response to OP's comment is a perfect display of what he means. A lot of tech bro hubris/idiocy.

Not within mathematics, where it is the entire sport, and which is the point of contention.

Ah the nature vs nurture debate, we meet again!

Give me a Neural Net in its first epoch and I shall mold it into anything!

Not a typo but an incorrect use of that word.