Lorde Celsius

international music sensation, doctoral student in physics, philanthropist, millionaire

[Lorde Celsius @LordeCelsius]

@growing_daniel skill issue

[Lorde Celsius @LordeCelsius]

@nostalgebraist sending this twete back in time ten years

[Lorde Celsius @LordeCelsius]

@_aidan_clark_ This is incredible bait

[Lorde Celsius @LordeCelsius]

@Aella_Girl @estherzelda0514 the ultimate scissor is whether or not you believe in the existence of positive sum trades

[Lorde Celsius @LordeCelsius]

@GoodmanAric @Aella_Girl but you can! You can become a teacher! your time is fungible with property taxes! By way of dollars!

[Lorde Celsius @LordeCelsius]

@a5eef72a @Aella_Girl @pmarca I would not want to be in the business of arguing the efficiency of various interventions with a radicalized activist!

[Lorde Celsius @LordeCelsius]

@a5eef72a @Aella_Girl @pmarca This is not an argument that “we should not use violence to stop the people trying to build it”, this is an argument that “we should only resort to violence if other, more efficient means of convincing don’t work”

[Lorde Celsius @LordeCelsius]

@Aella_Girl @pmarca I’m not sure it’s compatible to simultaneously believe “if anyone builds it, everyone dies” and “we should not use violence to stop the people trying to build it”

[Lorde Celsius @LordeCelsius]

@EigenGender @gptbrooke I worry that this will not be enough

[Lorde Celsius @LordeCelsius]

@Miles_Brundage not to mention the whole Elon directly intervening to change model behavior regarding Elon thing

[Lorde Celsius @LordeCelsius]

@tszzl I dunno man, it seems like there was at least one contractual obligation that was so unacceptable to the federal government that they were willing to detonate a company over it

[Lorde Celsius @LordeCelsius]

@CharlieBul58993 @Miles_Brundage The edge vs cloud deployment question practically doesn’t prevent autonomous weapon usage. If current frontier language models can play video games, they can pilot drones. There are latency and connectivity constraints, but nothing insurmountable.

[Lorde Celsius @LordeCelsius]

@littmath Given these are, on the surface, identical to Anthropic’s redlines, it probably behooves us to understand exactly why the redlines are acceptable under OpenAI’s contract, but not Anthropic’s

[Lorde Celsius @LordeCelsius]

@sama If the government wanted to used chatgpt to analyze lawfully collected private data of American citizens at scale, would that be allowed by your contract?

[Lorde Celsius @LordeCelsius]

@thsottiaux It’s honestly shameful watching openai rank and file fall in line for what is absolutely contractual safety-washing. You’ve been lied to, and you will have no one’s empathy when the government uses these models in ways that are already allowed by the contract you think is “safe”.

[Lorde Celsius @LordeCelsius]

@boazbaraktcs It will not be a comfortable update for you, but the contract, as written and posted, does not forbid the government from using your models for scaled domestic surveillance, or for usage in autonomous weapons. Your red lines are not actual lines. The contract literally allows it

[Lorde Celsius @LordeCelsius]

@itsclivetime I think you should update considerably on the possibility that you are being systematically lied to by your leadership

[Lorde Celsius @LordeCelsius]

@boazbaraktcs @ChrisPainterYup if the OpenAI agreement has the same red lines that DoW had with anthropic, then why did the DoW cancel their contract with Anthropic and designate them a supply chain risk, and then enter into the same contract with OpenAI?

[Lorde Celsius @LordeCelsius]

@AndyMasley @segyges you only passed if you immediately purchased nvidia

[Lorde Celsius @LordeCelsius]

@ianmiles @pangram slop?