michelb76

Apple is known to try to be neutral on a lof of things, while being decidedly not neutral on other stuff. It's confusing to be honest. I guess it's all to protect business interests, which makes it all the more inconsistent. I'm guessing they need Apple TV to be as politically bland as possible.

Makes sense. You can say Apple lags behind, but all AI tools are still not ready for primetime AT ALL. None of them have a level of polish that allows a regular user to use them, and all of them often still fail spectacularly. You can look at the LLM scores but none of that matters when a lot of them still can't do simple calculus.

I use several AI tools daily, but I'm a nerd and find the use cases for them, AND am able to see when it hallucinates. Apple is simply going to wait for this to mature and in the meantime use other services to supplement its offering. People went berserk when Apple Maps sent them in a wrong direction, can you imagine the backlash if they just let any LLM answer something?

Google still has the better web index, as OpenAI regularly fails on that front. Pretty sure Google will power Siri for web requests, and maybe Apple will use their own LLM for Apple related services.

VictorMortimer said:

twolf2919 said:

While definitely a vulnerability, I wish the author had highlighted how difficult it is to exploit.   Just telling me that the attack "must also be used on the same chip cluster as the cryptographic target app in order to function, and both must use the efficiency cores or the performance cores at the same time" doesn't really tell me that.  It sure sounds very esoteric - I'm not running a cluster on my MacBook (cluster of one?) nor do I really know what a 'cryptographic target app' means.

As most attacks, this requires you - at the very least - to download and run an app of unknown origin.  Don't do that.

So, stick to open source only?

Because the reality is that this can be hidden just about anywhere in a closed source app, and your precious app store can't protect you.

Good luck getting an average mac user to understand source code.

The ANE/NPU is only for inference, which is hardly used at the moment by any app, since inference is also fast enough on CPU+GPU for small models. So I'm wondering what the M4 will change. Double/triple the memory, memory bandwidth, GPU size? Auto-converting models to support MLX?

Since smaller models will be optimised for fast inference anyway, what are they going to bring that will be significant?

welshdog said:

Diletante said:

This could be a very interesting collaboration. Wardle’s tools are interesting in the way that they detect various types of malware but they currently require too much user involvement which can make them false-positive fire hoses. If AI could be used to filter that torrent to those alerts I really need to examine more closely, that would be a big win.

Not sure I understand your point about the Objective See apps. I have a number of them and they work well.

BlockBlock for example pops up for about everything, even things that are just fine (this is expected behaviour). As a regular user it's very difficult to determine what is a false positive and what isn't. If they can make this (more) foolproof for the average user, it would be great.

hets said:

This doesn't make any sense. The app store shouldn't have the power to destroy the web (apps). Apple has no right to kill the open web in EU.

They won't. It only kills the open web on iOS. This has zero impact on Android users. If web apps become popular, Apple will adjust. Safari is one of the worst browsers to develop a web app for, since Apple has not implemented many of the capabilities to protect the App Store. Apple will simply be a late adopter again, and none of this really matters.

There are many Apple features we never had in some parts of Europe, like Apple News in many countries, and we don't miss them. This will simply be another one. Limit access to many features in Europe and the market will adjust. Switching costs will go down anyway.

Redefining 'pro' again.