On Existential Risks
The degree to which a nuclear war between the US and Russia could escalate depends on how many of their nuclear weapons would survive a first strike. For decades, both the US and Russia have been able to maintain a secure second strike by hiding their nuclear weapons on submarines, armored trucks, and aircraft. If improvements in technology allowed... See more
Would US and Russian nuclear forces survive a first strike? — EA Forum
But at the point where a model becomes significantly more capable than GPT-4, we think evaluators need to be checking closely whether it meets some minimum capability threshold. Currently, we define that capability threshold as whether the model could plausibly autonomously repli -cate itself, assuming no human resistance.
Asterisk Issue 03: AI
By 1960, the U.S. war plan called for launching the entire nuclear arse -nal — at the time, 3,423 weapons, exploding with the blast power of 7,847 megatons — against 1,043 tar -gets in the Soviet Union, its satellite countries in Eastern Europe and Communist China. 3 This was not a plan to strike back if the Soviets launched a nuclear attack on the... See more
Asterisk Magazine Issue 01 Inaugural Issue
Somewhere between 2 percent and 3 percent of global calories were disrupted, but that led to a spike of about 40 or 50 percent in grain prices.
Asterisk Magazine Issue 02 Food
That means that in principle, humanity could crowdsource threat identification. Instead of warning the world about a new threat, scientists who are very concerned about a particular way that biology could be used to cause harm could contact a curator of the secure DNA system and say, “I’m really worried about this.” If the curator agrees, they can... See more
Asterisk Magazine Issue 01 Inaugural Issue
DELAY
you could instead train the model to make snide commentary in its hidden thoughts and then see if your alignment techniques were sufficient to remove the snide commentary. So the case is structurally the same, where the model has this hidden behavior and you want to make sure that training the observed behaviors also affects the hidden behavior in... See more
Asterisk Issue 03: AI
In the U.S., it took only 326 days from the first U.S. laboratory-confirmed case on January 20, 2020, until the first FDA-authorized vaccine on December 11, 2020. The COVID vaccines likely saved nearly 20 million lives worldwide by the end of 2021.
Asterisk Magazine Issue 02 Food
Why despite global progress, humanity is probably facing its most dangerous time ever
80000hours.orgwe do produce more food per capita than ever before, which is an incredible achievement. The challenge today is making sure that we can get it to people. And the challenge in the future will be making sure we can always produce that food.