Daily Tech News 21 October 2023

• Instagram has apologised for accidentally labelling terrorists as terrorists. (Ars Technica)

  Not everyone has accepted Meta's apology. Director of Amnesty Tech—a branch of Amnesty International that advocates for tech companies to put human terrorist rights first—Rasha Abdul-Rahim, said on X Twitter that Meta Instagram apologizing is "not good enough."

  "You're not allowed to tell the truth", said Abdul-Rahim. "You have to lie, and you have to tell the lies we feed you."
  
  
  

• Zotac's ZBox Pico PI430AJ uses Frore's piezoelectric Airjet cooling system. (AnandTech)
  
  These are often called solid state, but they do have moving parts. They're nearly silent because they vibrate too fast for you to hear - the only sound is the moving air.
  
  The catch? This device has a 7W CPU and would run fine without a fan anyway. Though a quick look at the benchmark results suggests it might be thermally limited, so it's possible the fan does help.
  
  
  
• Samsung's new HBM3E memory is big and fast. (AnandTech)
  
  A single package - not exactly a chip, but a stack of 12 silicon dies - holds 36GB, runs at 9.8GHz, and is 1024 bits wide, giving a bandwidth of 1.2TB per second.
  
  
  
• Mediatek is no longer a second-tier manufacturer of chips for cheap phones. (WCCFTech)
  
  The company used to be synonymous with budget devices, but its Dimensity 9300 is just a hair slower - 1.7% - than Apple's best mobile chips in multi-core tests. The difference is wider in single-core benchmarks, where Apple still holds a convincing lead.
  
  I'd like to see chips like these in NUCs. Windows Arm devices have been pretty underwhelming so far, mostly using chips that are badly out of date by the time the products hit the market.
  
  
  
• Reddit is considering blocking search engines and other web spiders - particularly those used by AI companies - though exactly how it plans to do that is an open question. (The Verge)
  
  Google and Microsoft are big targets with lots of money that you can sue if they breach their own terms of service related to web indexing. If you flag them in your robots.txt file - which they promise to obey - and they ignore it, that's grounds for a great big settlement.
  
  AI companies are much smaller and less scrupulous, and if your content is still public there's no simple way to block them. Instagram has been doing this for years, and people still scrape its content.
  
  And the AI companies need Reddit far more than Reddit needs them. You can only train AIs on AI content for a very short time before it all turns to sheep.

posted by Pixy Misa at 04:00 AM

| Access Comments