Just saw Vitalik post this: he disagrees with the one-sided, heated sentiment around the “anti-data-center movement” that’s been circulating this week.
His reasoning is pretty straightforward: algorithms keep improving, so something that would require more and bigger data centers to do two years from now could probably be done with today’s data centers in four years—or even sooner. So even if anti-data-center activists achieved extreme success, it wouldn’t lead to anything apocalyptic. The more urgent priority is to address the cybersecurity vulnerabilities we can already see, along with potential future risks such as biological threats. And research getting put into practice is held up by many non-AI factors, so if AI is delayed by a year, these applications often won’t be delayed by a year too.
By that logic, slowing the growth of computing power could actually be a net positive, he thinks. He also went out of his way to say there’s no need to invoke controversial doomsday scenarios or recursive self-improvement; even if this judgment ultimately turns out to be wrong, it wouldn’t rank among the ten most pressing political problems we face right now.
#Vitalik #AI #DataCenters
His reasoning is pretty straightforward: algorithms keep improving, so something that would require more and bigger data centers to do two years from now could probably be done with today’s data centers in four years—or even sooner. So even if anti-data-center activists achieved extreme success, it wouldn’t lead to anything apocalyptic. The more urgent priority is to address the cybersecurity vulnerabilities we can already see, along with potential future risks such as biological threats. And research getting put into practice is held up by many non-AI factors, so if AI is delayed by a year, these applications often won’t be delayed by a year too.
By that logic, slowing the growth of computing power could actually be a net positive, he thinks. He also went out of his way to say there’s no need to invoke controversial doomsday scenarios or recursive self-improvement; even if this judgment ultimately turns out to be wrong, it wouldn’t rank among the ten most pressing political problems we face right now.
#Vitalik #AI #DataCenters
