Mac-Based Gaming & Apple’s AR Acquisition (Plus LLM Hallucinations) | June 9th, 2023
In Today’s Tech News:
- Mac Gaming Is Right Around the Corner
- Apple Acquires Industrial AR Startup Mira
- WordPress Gets an AI Jetpack
- LLM Hallucinations Proven To Be Deadly
Mac Gaming Is Right Around the Corner
The Worldwide Developers Conference has brought incredible news for Mac gamers.
The Game Porting Toolkit was announced bringing Windows gaming capabilities to macOS Sanoma using similar translation methods as Valve’s Proton.
Under the hood, Apple is using Code Weavers CrossOver source code, specifically recognizing the open-source project Wine as an incredible answer to Windows gaming on macOS.
Reddit has already gotten its hands on the Sanoma beta as its users have shared footage of Cyberpunk 2077, Diablo IV, and Hogwarts Legacy all running on Apple’s M-series devices.
The posts showing off the gameplay can be found in our sources below.
As a final bonus, Hideo Kojima announced Death Stranding Director’s Cut will be coming to macOS.
Sources:
Apple Acquires Industrial AR Startup Mira
Mira, an augmented reality corporation focused on lightweight headsets for front-line business and military applications, has been acquired by Apple just one day after the announcement of the Vision Pro.
It is currently unknown how much Apple paid for the Los Angeles-based organization, but they had already acquired $17 million in funding up to this point.
Mira is focused on business applications of AR and has contracts within multiple industries including manufacturing, mining, entertainment, and defense.
Most contract details are unknown as they are private, but according to public records, a contract with the US Navy is valued at over $700,000.
Another contract is with Nintendo World as the headsets used for the Mario Kart ride were developed by Mira as well.
According to The Verge, Apple has brought on at least 11 employees from Mira, begging the question, what is Apple up to?
Sources:
WordPress Gets an AI Jetpack
Automattic, the organization that owns WordPress, has released a generative AI plugin for WordPress called the Jetpack AI Assistant.
The “creative writing partner” will be free for a limited time for WordPress.com users while others get 20 free requests before paying a monthly fee of $10.
Features of the plugin include headline creation from post summaries, tone adjustment throughout the post, entire post and table creation through prompting, intense spelling and grammar checks, translation between 12 languages including English, Spanish, French, and Chinese, and finally a conversational UI, much like ChatGPT, Bing Chat, and Bard.
To use the Jetpack AI Assistant, install the plugin to your WordPress site, add an “AI Assitant” block to your post through the block editor, and prompt to your heart’s desire.
A major concern of AI is the amount of low-grade automated garbage created, but we will have to see how or if Automattic will handle the abuse of their generative AI.
Sources
LLM Hallucinations Proven To Be Deadly
Researchers at Vulcan Cyber have recently released a proof of concept describing how attackers can easily use ChatGPT to help them spread malicious packages into developers’ environments.
They are calling the new technique “AI package hallucination” as the attack vector is as follows…
- The attacker prompts a large-language-model, or LLM, such as ChatGPT or Bard with a coding problem
- The LLM responds with the hallucinated package that doesn’t exist publicly
- The attacker creates and publishes the package containing malware (with or without expected package functionality)
- A developer then prompts the LLM with a similar problem
- The LLM responds with the hallucinated package as a solution
- The developer then installs and uses the malicious package unknowingly
- Finally, the package is run and information is sent to the attacker’s Command and Control server
Techniques used within this attack include typosquatting, masquerading, dependency confusion, software package hijacking, and trojan packaging.
As this is a brand new technique, ways to mitigate your risk are limited but include best practices such as
- Validating and verifying all packages used when building out solutions before installing them
- Auditing and monitoring the libraries and packages already used within your solutions
- This includes static scans of obfuscated code as well as checking for common vulnerabilities and exposures or CVEs such as the OWASP top 10
- Auditing outbound links to external sites and assets
- Auditing your networking and communications for any unexpected outbound or inbound traffic
Sources:
What Are Your Thoughts?
- Is Apple’s true focus on AR a bet on itself over generative AI?
- Will Jetpack AI be a great tool or a hindrance for journalism as a whole?
- Why did the cookie go to the doctor? (It was feeling pretty crumby)
Let us know your thoughts in the comments at the bottom of the page!
