Subj : OpenAI's democratic AI ¨,
To   : All
From : Mike Powell
Date : Wed Feb 07 2024 09:14:00

 * Originally in: DoveNet TechTalk

The task of aligning AI is tricky as it is difficult to determine who
gets to decide what values to align with 

From TLDR 2024-02-06

 INSIDE OPENAI'S PLAN TO MAKE AI MORE `DEMOCRATIC' (15 MINUTE READ) [6]

 The task of aligning AI is tricky as it is difficult to determine who
gets to decide what values to align with. OpenAI is searching for a
path toward AI alignment upon which large swaths of the public can
agree. While it is consulting with the public and adjusting its
strategy based on the voice of the people, the process is not
democratic as the input is advisory, not binding. Sam Altman has
previously said that his company would 'respect' it if the public
unequivocally said that OpenAI should stop or slow down, but experts
think it would be unlikely. 

 NEW E. COLI STRAIN WILL ACCELERATE EVOLUTION OF THE GENES OF YOUR
CHOICE (4 MINUTE READ) [8]

 Researchers at Cambridge University have created a synthetic
orthogonal DNA replication system in E. coli that can be used as a
risk-free way to generate and study mutations. The system is
completely separate from the system that the bacteria use to copy the
genes they need to survive. It turns the bacteria into hotbeds of
evolution that have the potential to greatly facilitate the
development of enzymes and other proteins that will have applications
in research, medicine, and industry. The system was inspired by a
similar system that exists in yeast. 

 IT'S EASIER TO FORGIVE A HUMAN THAN A ROBOT (2 MINUTE READ) [12] 

 We hold robots to a higher standard than we hold humans. Robots will
have to be virtually perfect before we accept them for tasks like
driving, medical care, and even customer support interactions. For
example, even if self-driving cars prevent a significant number of car
accidents, they'll still be responsible for the accidents they cause,
which will have legal ramifications that may be impossible to
overcome. Humans may be more forgiving if AI is adopted widely enough,
but in the meantime, we'll have to work out what a tolerable error
rate is for robots when comparing them with humans. 

 META CUTS OFF THIRD-PARTY ACCESS TO FACEBOOK GROUPS, LEAVING
DEVELOPERS AND CUSTOMERS IN DISARRAY (3 MINUTE READ) [17] 

 Meta is deprecating its Facebook Groups API in favor of its new
Facebook Graph API v19.0, which still has features that allow
developers to privately reply in Facebook groups. 

 SNAP IS CUTTING 10 PERCENT OF ITS STAFF (2 MINUTE READ) [20] 

 Snap has struggled to expand beyond its core social networking
product while facing the same challenges as other tech and media
companies in recent years. 

Links:
------
[6]
https://time.com/6684266/openai-democracy-artificial-intelligence/?utm_source=tldrnewsletter

[8]
https://arstechnica.com/science/2024/02/new-e-coli-strain-will-accelerate-evolution-of-the-genes-of-your-choice/?utm_source=tldrnewsletter

[12]
https://world.hey.com/dhh/it-s-easier-to-forgive-a-human-than-a-robot-d4a97b3a?utm_source=tldrnewsletter

[17]
https://techcrunch.com/2024/02/05/meta-cuts-off-third-party-access-to-facebook-groups-leaving-developers-and-customers-in-disarray/?utm_source=tldrnewsletter

[20]
https://www.theverge.com/2024/2/5/24062044/snap-layoffs-10-percent-staff-snapchat?utm_source=tldrnewsletter

 -- https://tldr.tech/tech
--- SBBSecho 3.14-Linux
 * Origin: capitolcityonline.net * Telnet/SSH:2022/HTTP (1:2320/105)

.