Apple owns a platform. So they can just implement this later and make sure the competition loses their edge.
OpenClaw is what Apple intelligence should have been
301–310 of 440 posts
Re: OpenClaw is what Apple intelligence should have been
#302Earlier quoted context omitted.
If you’re heavily invested in Apple apps (iMessage/Calendar/Reminders/Notes), you need a Mac to give the agent tools to interact with these apps. I think that combined with the form factor, price, and power consumption, makes it an ideal candidate. If you’re heavily invested in Windows, then you’d probably go for a small x86 PC.
Can you imagine giving an AI access to your messages, notes and calendar though? I use agentic coding, this is next level madness.
I guess what’s wrong with it? Let’s say it has read only access, new messages and calendar invites need approval. I’m not sure I understand the harm? I suppose data exfiltration, but like you could start with an allowlist approach. So the first few uses and reads take a while with allowing the ai to read stuff , but it doesn’t seem that crazy given it’s what we basically do with ai coding tools?
Re: OpenClaw is what Apple intelligence should have been
#303What are the actual numbers on these purpotedly all-encompassing mac mini sales?
Re: OpenClaw is what Apple intelligence should have been
#304[dead]
Re: OpenClaw is what Apple intelligence should have been
#305The main issue why we don't see AI agents in products: PROMPT INJECTIONS
Even with the most advanced LLMs and even sandboxing there is always the risk of prompt injections and data extraction.
Even if the AI can't directly upload data to the internet, or delete local data, there are always some ways to leak data. For example by crafting an email with the relevant text in white or invisible somewhere. The user clicks "ok send" from what they see, but still some data is leaked.
Apple intelligence is based on a local model on the device, which is much more susceptible for prompt injections.
Re: OpenClaw is what Apple intelligence should have been
#306According to https://1password.com/blog/from-magic-to-malware-how-opencla... , The top skill is/was malware. It's obviously broken, so no, Apple Intelligence should not have been this.
OpenClaw is not broken, it is just not designed to be secure in the first place.
It's more like a tech demo to show what's possible. But also to show where the limits are. Look at it as modern art, like an episode of Black Mirror. It's a window to the future. But it also highlights all the security issues associated with AI.
And that's why you probably shouldn't use OpenClaw on your data or your PC.
Re: OpenClaw is what Apple intelligence should have been
#307The main issue why we don't see AI agents in products: PROMPT INJECTIONS Even with the most advanced LLMs and even sandboxing there is always the risk of prompt injections and data extraction. Even if the AI can't directly upload data to the internet, or delete local data, there are always some ways to leak data. For example by crafting an email with the relevant text in white or invisible somewhere. The user clicks…
Surely this is the elephant in the room, but the point here is that Apple as control over its ecosystem, so it may be able to sandbox and make entitlements and transparency good enough, in the apps that the bot can access.
Re: OpenClaw is what Apple intelligence should have been
#308The main issue why we don't see AI agents in products: PROMPT INJECTIONS Even with the most advanced LLMs and even sandboxing there is always the risk of prompt injections and data extraction. Even if the AI can't directly upload data to the internet, or delete local data, there are always some ways to leak data. For example by crafting an email with the relevant text in white or invisible somewhere. The user clicks…
Surely this is the elephant in the room, but the point here is that Apple as control over its ecosystem, so it may be able to sandbox and make entitlements and transparency good enough, in the apps that the bot can access.
Like I said: sandboxing doesn't solve the problem.
As long as the agent creates more than just text, it can leak data. If it can access the internet in any manner, it can leak data.
The models are extremely creative and good at figuring out stuff, even circumventing safety measures that are not fully air tight. Most of the time they catch the deception, but in some very well crafted exploits they don't.
Re: OpenClaw is what Apple intelligence should have been
#309I keep seeing posts about OpenClaw (I still didn’t try it myself) but I don’t get the constant references to the Mac Minis. Why are people needing the Mac Minis? Isn’t OpenClaw supposed to run locally in your laptop? And if it actually should run as a service, why a MacMini and not some docker on the local NAS for instance?
You need to run it on MacOS if you want it to interact with iMessages and such. And people are likely not proficient enough to set up and connect everything without GUI.
Re: OpenClaw is what Apple intelligence should have been
#310Earlier quoted context omitted.
I don’t believe this was ever confirmed by Apple, but there was widespread speculation at the time[1] that the delay was due to the very prompt injection attacks OpenClaw users are now discovering. It would be genuinely catastrophic to ship an insecure system with this kind of data access, even with an ‘unsafe mode’. These kinds of risks can only be _consented to_ by technical people who correctly understand them, le…
The more interesting question I have is if such Prompt Injection Attacks can ever be actualy avoided, with how GenAI works.
They could be if models were trained properly, with more carefully delineated prompts.