Elon Musk, one of the most influential visionaries of our era, expressed his fears about the potential dangers of such an advance, even attempting to acquire DeepMind to have control over its development. However, Demis Hassabis remains focused on his goal,
SAFETY
-
The possible consequences of Demis’s success
By
–
So, what could happen if Demis succeeds? We could solve the world's greatest problems overnight. Or we could create something we cannot control. The future of humanity could depend on Demis's success.
-
OpenAI Hiring Research Scientists Engineers AI Safety
By
–
We are hiring Research Scientists and Engineers for safety research at @OpenAI
, ranging from safe model behavior training, adversarial robustness, AI in healthcare, frontier risk evaluation and more. Please fill in this form if you are interested: -
GenAI breakthroughs accelerate speed of opportunities and threats
By
–
Directionally the same, but the speed/opportunities/threats are coming much faster due to breakthroughs in GenAI.
-
Confronting AI’s Ultimate Existential Questions
By
–
You cannot avoid the ultimate questions forever.
-
Truth-Seeking Systems: Mutation, Trial, Error, Conjecture
By
–
Evolution works by mutation and selection. Innovation works by trial and error. Science works by conjecture and criticism. All truth-seeking systems work roughly the same way.
-
Nonverbal Turing Test for Robots: 188 Behavioral Benchmarks
By
–
How could we define a nonverbal Turing Test for robots? We asked people to use their bodies to test what agents could and could not do. We found they tested physical, emotional and social abilities with 188 unique behaviors! https://
arxiv.org/abs/2409.11602 https://
youtu.be/hc27aTLBv7U?si
=rjJhEKk4a-BaYBlL
… -
Affective Empathy Missing From Foundation Models and AI Ethics
By
–
Affective empathy, a building block for morality, is missing from foundation models (and also missing from people with psychopathy). While I argued for it in 2014, I am similarly concerned about building it into machines. It means defining pain and linking it to physical function
-
o1-mini Struggles to Prove Matrix Multiplication Distributive Property
By
–
I tried to get o1-mini to prove distributive property for matmul and it tried like 3 times before basically saying "ok I can't get this but like trust me it holds"
-
ChatGPT vs. Claude: Best AI for Content
By
–
ChatGPT vs. Claude explored. Choose the best AI for content. • Assess flexibility
• Review safety
• Check versatility Read more: https://
buff.ly/401y0Sj