Speaking of ASICs - how likely is it that as models get better we'll see someone baking a whole model directly into the silicon? It's like having l0 cache.

▲

kaelwd 16 minutes ago | parent | next [-]

Only 8B currently but it's been done: https://taalas.com/products/

	▲	apimade a minute ago \| parent [-]
		The only _public_ example we know. This is definitely being done with private models by HFT/quant firms, data processing agencies/orgs (large intelligence agencies, _every_ data analytics org, etc).

▲

SJC_Hacker 43 minutes ago | parent | prev [-]

You could do it but there would be no point, The only advantage over would be power consumption. And it would be quite expensive.

At the rate models are improving, it would be obsolete in six months.

▲

HPsquared 21 minutes ago | parent [-]

Power consumption and latency are very important on mobile

	▲	naasking 5 minutes ago \| parent [-]
		They're important everywhere of course, but especially on mobile. If AI reaearchers figure out how to offload knowledge and expertise from reasoning weights, then a core reasoning ASIC linked to the knowledge would totally rock.