mpantone

Hopefully someday Apple Intelligence will run spell check before articles to AppleInsider are submitted...

  

If only someone would create a standalone, manually invoked spell check in the interim... If LLMs already help students write term papers and cubicle monkeys write e-mails, why not news articles? Ah, to dream...

  

iadlib said:

Just release the director's cut in the first place... after all... he is the director

Hollywood studio executives are understandably allergic to films with really long running times. Remember that the "perfect running length" varies by person. The problem is when too many reviewers say a movie is "too long" or "too slow paced" which will dampen ticket sales.

Ridley Scott knows this. He is well aware going into these projects that there will be a theatrical release that is shorter than his own ideal version. It's not like Napoleon is his first film. Remember that the first people he needs to gain acceptance from are studio execs, not AppleInsider forum contributors.

And when Scott releases a new film we (well at least those who aren't naive) can expect a longer "director's cut" in the future. Hell, he probably scripts these films knowing that some scenes aren't going to make the theatrical cut but are important to the longer version. It's not like he goes back to reshoot additional footage six months later.

And he's definitely not the sole practitioner of this. There are Michelin star chefs that offer tasting menus will differing numbers of courses. Taylor Swift has done this with great success, releasing different versions on an album with additional content. Her most recent The Tortured Poets Department has been released in a different forms including the Anthology edition which includes an entire second album.

king editor the grate said:

mpantone said:

king editor the grate said:

In-person?!? Awesome! Hopefully, it will be interesting to watch again.

I'm afraid you have misunderstood this event.

The actual presentation will be the pre-recorded video as Apple has done since the pandemic. This will be projected in the Steve Jobs Theater and streamed worldwide (Apple TV, YouTube, etc.).

After the video is over, event guests will be given the opportunity to try the newly-announced products in-person in a demo zone outside the theater. Likely some Apple executives and PR types will be available to field questions from the attendees (mostly media). This portion of the event is not broadcasted.

Aw, crapola. 

Apple is more or less done with live presentations with rare minor exceptions that are very situationally specific.

All of their major announcements are now pre-recorded which better suits their "control freak" mentality: there is no chance of an embarrassing moment caused by balky demo hardware/software or poor audio caused by a poorly affixed lav mike. Everything goes smooth and by the numbers. Also the streaming video quality is better quality since they aren't compressing on the fly live.

Streaming live events is very commonplace in 2024. It wasn't the case back in 2007 when Steve announced the iPhone.

There should be no surprise about this.

king editor the grate said:

In-person?!? Awesome! Hopefully, it will be interesting to watch again.

I'm afraid you have misunderstood this event.

The actual presentation will be the pre-recorded video as Apple has done since the pandemic. This will be projected in the Steve Jobs Theater and streamed worldwide (Apple TV, YouTube, etc.).

After the video is over, event guests will be given the opportunity to try the newly-announced products in-person in a demo zone outside the theater. Likely some Apple executives and PR types will be available to field questions from the attendees (mostly media). This portion of the event is not broadcasted.

blastdoor said:

Scrolling through the entries so far it seems that the Neural Engine works great for the quantized version of the benchmark and so-so for half precision. But it seems Apple hardware in general is kind of lame for the single precision version (with the best hardware being the GPU in the M3 Max). 

Happy to be corrected, but I take this to mean that Apple hardware is good for on-device inference but crappy for model training. 

I wonder how much of the issue is CoreML needing more optimization versus Apple needing beefier GPUs... 

I got similar results (GPU being best for single precision, Neural Engine best for quantized benchmark) on my Mac Mini M2 Pro.

The fact that current Apple hardware (which is consumer targeted) isn't optimized for model training comes as no surprise.

This just leaves many questions and perhaps some of them will be answered in the future. My hunch is that Apple will eventually provide developers some sort of access to cloud servers that have more hardware resources for model training. There's also a small possibility that Apple could release some sort of add-in-board AI accelerator for the Mac Pro that is better suited for model training.

But for sure, Apple won't prioritize model training on their regular MacBook and mobile offerings.

I ran the benchmark on one of my Windows PCs equipped with a GeForce RTX 3060 12GB graphics card. That GPU had nearly double the performance of my M2 Pro for single precision and +50% better performance at half precision (compared to the Mac's Neural Engine). Now the 3060 is a low end card from the Ampere generation of GPUs so I'd expect even better performance from higher-end models or those from the Ada Lovelace generation.

All that said, benchmarks are a generally poor barometer of real world performance. Since iOS 18/Sequoia isn't even officially released, it will be some time until we see whether or not the Geekbench AI benchmark reflects actual performance in any sort of meaningful way. In the same way, there aren't many consumer AI applications running locally on typical PCs yet so running the Geekbench benchmark on Windows PCs.

So much of the usefulness is unleashed from the software (Apple considers itself a software company first) so just benchmarking hardware using synthetic tests might be extremely marginal in terms of usefulness or accuracy vis-a-vis real world usage.