Preparedness team, led by @aleks_madry
, will focus on evaluation of and protection for catastrophic risks that might be triggered by AGI-level capability, including cybersecurity, bioweapon threats, persuasion and more. Come join us – https://
openai.com/careers/search
?c=preparedness
…
AGI
-
OpenAI Launches AGI Preparedness Team Led by Madry
By
–
-
Dynamic Memory LLMs Will Obsolete Prompt Engineering
By
–
Now we have LLMs with a fixed-size context. Imagine LLMs with dynamic, expandable long-term memory. You'll align them to yourself through iterative conversations. This will render prompt engineering obsolete. The LLM will anticipate your needs. The challenge will be: if the
-
New AI Preparedness Team Evaluates AGI Risks Quantitatively
By
–
We are building a new Preparedness team to evaluate, forecast, and protect against the risks of highly-capable AI—from today's models to AGI. Goal: a quantitative, evidence-based methodology, beyond what is accepted as possible:
-
Accelerated Timeline Debate in AI Development Strategy
By
–
Would love to see an accelerated timeline, but might be quite a bit longer if we assume they're going to bide their time as usual.
-
Designing AI Intrinsic Objectives to Prevent Sociopathic Behavior
By
–
You are right.
You can't control if a human is a sociopath.
But we can make sure an AI system won't be a sociopath because we get to design their intrinsic objectives. -
AI Alignment Easier Than Aligning Humans Through Design
By
–
No.
But I'm arguing that
a) we are familiar with the concept of aligning intelligent entities, whether they are children, adults, or corporations (which are super-human).
b) it is *much* easier to align AI than humans because we get to *design* the intrinsic motivations of AI -
Governance frameworks for AI agents: lessons from human entities
By
–
We already handle this with humans, groups, corporations, and governments.
-
Designing AI Objectives: Alignment Challenge vs Human Nature
By
–
You are right.
AI systems are much easier to align because we get to *design* their intrinsic objectives.
We don't get to do this with humans. We only get to modify objectives slightly through education. -
Designing AI Intrinsic Objectives for Common Good Alignment
By
–
Kids have human nature and drives hardwired in their intrinsic objectives. With education, we can merely modify these objectives slightly.
For AI systems, we get to *design* their intrinsic objectives. This makes it *much* easier to align AI to the Common Good than it is to align -
Aligning AI with Common Good Easier Than Nature Modification
By
–
It is much, much easier to align AI with the Common Good, than it is for children and animals.
the reason is that we can't "hack" human nature. We can only modify it through education.
We can "hack" animal nature through selective breeding (which is pretty brutal).
We can