On-device LLMs/edge AI tooling risks
Key Questions
What on-device AI update did Adobe release for Photoshop?
Photoshop 27.7 adds on-device AI processing for M1 and later Macs. Users should note potential limitations or catches with the new feature.
How is Google advancing on-device LLMs for agents?
Google presented fine-tuning techniques that boosted tiny LLM performance from 46% to 90% for on-device agents. This supports privacy-focused edge deployments on Pixel devices.
What is Apple's strategy for on-device AI?
Apple emphasizes on-device intelligence combined with Private Cloud Compute to advance sovereign AI. The approach prioritizes privacy in its expanding AI ecosystem.
What hardware risks come with increased edge AI deployment?
Expanded edge deployments broaden the attack surface, including vibration side-channels and LLM compression vulnerabilities. Local hardware choices can introduce unexpected security issues.
What new AI chips or tools did Google unveil at I/O 2026?
Google announced the TPU 8T and Gemini Intelligence features for on-device use. These support broader integration across Android and PC hardware.
How are companies like Ubuntu and ASUS approaching local AI?
Ubuntu is prioritizing local AI over cloud-first integration, while ASUS introduced an on-device Router Assistant. Both reflect a shift toward edge-focused tooling.
What challenges exist with older Pixel phones and new Gemini features?
Gemini Intelligence may leave some recent Pixel models unsupported despite long update promises. This highlights gaps in Google's 7-year device support commitments.
What portable AI hardware options are gaining attention?
NVIDIA's palm-sized AI computer and similar compact devices are emerging for on-device workloads. These tools aim to bring powerful local inference to users.
Adobe Photoshop 27.7 on-device (M1+); Gemma fine-tuning on Pixel; LLM compression; vibration side-channels. Continued edge deployment expands attack surface. Apple on-device AI strategy signals privacy-focused expansion.