It pretty bad. Just ask it to quote something like the first paragraph of moby dick (i.e. "call me ishmael...") it barely gets past 5 words before it completely gets it wrong. Even the worst normal 4B models get 3-5 sentences.
That’s because diffusion models struggle with memorizing sequences while it’s essentially what autoregressive models excel at.
I’ve tried to fine tune a diffusion model to memorize an MD5 hash and after days of hacking on it, it failed to completely memorize it. Autoregressive models of any size learn it over <1000 iterations.
the headline chart was lazy on our part, thanks for the flag and I’ll update it.
each individual benchmark that is combined into agentic coding index was compared on xhigh between ua / codex / pi, and headline improvement was calculated on xhigh, but then agentic coding index pareto chart by default includex codex max, hense the confusion.
(to finish your thought) Which is important for consumer hardware to be better suited to running these models. Cloud providers are already able to batch as many requests as they want together to improve resource utilization, so they will not see a big benefit from diffusion models.
> A redesigned inward-facing microphone works with voice enhancement algorithms to better recognize and articulate your voice, so it sounds more natural when you’re on phone and video calls.
Also where are you getting "moving away from bluetooth"? That was a rumor (that they might move to their U1 tech for better audio quality) but they didn't mention that in the presentation nor is it mentioned on their site.
Ahh, that makes way more sense. I didn't apply the "nothing on" to the second thing and while I think moving off bluetooth is a plus I know there are others who see it as a negative and I assumed that's what they meant since moving from BT was rumored ("Apple doing something proprietary", even though clearly the market isn't moving quickly to fix the issues with BT).
UWB would provide plenty of bandwidth at close range with better energy efficiency, but it will not match bluetooth for distance. With sufficient software complexity you could switch back and forth between them depending on conditions… so that sounds like a business decision on cost to implement and benefit.
Or some clown has a patent conjoining some aspect of UWB and audio, or protocol switching. Apple never says "we don't do this because there is a patent in the way". It isn't their image.
Was approached by Amazon and Google somewhat recently, they both wanted me to relocate to Dublin. Perhaps FAN hire for 100% remote, but AG don't seem to.
