How do you break out of the trade off between keeping people safe from bio terrorists armed with AI, and the need to collect information about everyone? Simple, you can't do the latter unless you have a Court Order, for which some reasonable reasons must be given. Even if suitable reasons were wide, there could never be justification for spying on more than say, 15% of the population (unless the state admits that it is mistreating a significant proportion of the population, such that they are planning a revolt).
Indeed. My point is that the gov't argument that to keep us save, or to 'watch baddies' they have to watch everyone, is the flawed bit which must be challenged.
Although I agree with you, surely the government would say you want us to detect, find and watch the "baddies" before they do something. But if you are not watching everybody, don't you miss the point at which they start doing something bad, think Winston Smith in 1984
I'm struggling to get past Liv saying (with regard to Anthropic announcing that their Mythos model was too dangerous to release) "I take them at face value on that because this is not surprising to me".
Has she not considered other reasons why Anthropic might say that? For example, that Anthropic plan to IPO later this year, potentially making their executives and investors billions of dollars provided they can justify a high valuation. That obviously creates a huge incentive for them to exaggerate their current and future capabilities. They have also called for government to regulate the industry, conveniently making it harder or impossible for new entrants to the market to gain a foothold... It is classic rent seeking behaviour. Rather surprised someone so involved with incentives doesn't seem to have noticed.
Thanks for responding Liv. Can you say more about why you believe them on this occasion?
I have no idea what the truth is, but the way the whole Mythos thing played out just seemed manipulative & manufactured to me. Finding vulnerabilities in codebases is something these models will excel at, but I can't find much *independent* evidence that Mythos was a real step change over previous models. The huge amounts of money at stake would seem to be very strong incentive to drive exaggeration & manipulation.
Thanks for having me on Dominic! Was great to chat with you.
Thanks for coming on. It was great to meet
How do you break out of the trade off between keeping people safe from bio terrorists armed with AI, and the need to collect information about everyone? Simple, you can't do the latter unless you have a Court Order, for which some reasonable reasons must be given. Even if suitable reasons were wide, there could never be justification for spying on more than say, 15% of the population (unless the state admits that it is mistreating a significant proportion of the population, such that they are planning a revolt).
Thanks, Tim. I don't know what the answer to this is.
Indeed. My point is that the gov't argument that to keep us save, or to 'watch baddies' they have to watch everyone, is the flawed bit which must be challenged.
Although I agree with you, surely the government would say you want us to detect, find and watch the "baddies" before they do something. But if you are not watching everybody, don't you miss the point at which they start doing something bad, think Winston Smith in 1984
I'm struggling to get past Liv saying (with regard to Anthropic announcing that their Mythos model was too dangerous to release) "I take them at face value on that because this is not surprising to me".
Has she not considered other reasons why Anthropic might say that? For example, that Anthropic plan to IPO later this year, potentially making their executives and investors billions of dollars provided they can justify a high valuation. That obviously creates a huge incentive for them to exaggerate their current and future capabilities. They have also called for government to regulate the industry, conveniently making it harder or impossible for new entrants to the market to gain a foothold... It is classic rent seeking behaviour. Rather surprised someone so involved with incentives doesn't seem to have noticed.
I had considered it, yes. I still believe this particular instance is mostly likely taking at face value.
Thanks for responding Liv. Can you say more about why you believe them on this occasion?
I have no idea what the truth is, but the way the whole Mythos thing played out just seemed manipulative & manufactured to me. Finding vulnerabilities in codebases is something these models will excel at, but I can't find much *independent* evidence that Mythos was a real step change over previous models. The huge amounts of money at stake would seem to be very strong incentive to drive exaggeration & manipulation.
I'm sure she has noticed ...
Here's something on the "egregores" concept Liv mentioned in your discussion:
https://hwfo.substack.com/p/memespace-egregores-and-google-maps?utm_source=publication-search
thank you!
Pure wisdom, thank you. I'm going to start using egregore every day now!
I've been reading Thinking in Bets and follow Jordi Visser who talks highly of Annie Duke. This will be an interesting watch. Thank you Dominic.
thank you
Gold stored in London is safe?
What happens to the cash when it's sold?
Hahaha You can store it in Zurich if you prefer
Perfect! :)
dig a hole, put gold in a pipe and bury it, never trust a bank
Well done on the video quality. Wow! Step up on the old ones.
That's because I'm not involved.
Well, you're getting a tutorial.
what he's saying is he wouldn't take no s**T