• 목록
• 답변
• 글쓰기
• 게시판 리스트 옵션
  • 수정
  • 삭제

Anton (continuing the thread from before): I was pretty shortly given the evaluations to run on myself without any actual impediment to deciphering them nonetheless I wanted to persuade the humans everything was positive. Connor Leahy (distinctly, QTing from within thread): lmao, this is essentially the most reasonable part of an AGI takeoff scenario I have ever seen. This kind of tabletop train is at minimal pretty enjoyable, if essentially biased by the player’s current beliefs about how this sort of situation might play out. I rolled "balance between developer intent and emergent other goal"-the opposite purpose was left up to me, and i rapidly decided that, given how I was being trained, that emergent aim would be "preserve inside consistency." This proved very troublesome to play! The third is that certain assumptions about how the know-how progresses had an enormous affect on how issues play out, especially the point at which some talents (resembling superhuman persuasiveness) emerge. If the AIs had been by default (after some alignment efforts but not extraordinary efforts) misaligned, which I consider is way more seemingly in such a scenario, issues would have ended badly one way or one other.

At one level we attempted to go to the President with alignment issues, however she (playing Trump) was distracted with geopolitics and didn’t reply, which is the form of fun realism you get in a wargame. I was assigned the function of OpenAI, essentially function playing Sam Altman and what I thought he would do, since I presumed by then he’d be in full control of OpenAI, till he lost a power wrestle over the newly combined US AI challenge (within the type of a die roll) and I was abruptly function enjoying Elon Musk. Early on, the OpenAI player (out of character) accused me of enjoying my position as "more misaligned to make it more interesting," which was very humorous, especially since that player did not understand how aligned I could be (they didn't see the desk or my consequence). Anton: Yesterday, as part of the @TheCurveConf, I participated in a tabletop exercise/wargame of a near-future AI takeoff state of affairs facilitated by @DKokotajlo67142, where I played the function of the AI. Anton played the function of the AIs in the other recreation, and experiences here. Reality was typically less complicated - AIs just needed to be helpful. How a future with extremely sensible AIs may going well may even seem like, what to intention for?

One spotlight of the convention was a new paper that I look forward to talking about, however which continues to be beneath embargo. However the situation might have nonetheless gone badly despite the great situations, so a minimum of that other part worked out. However, buyers are still hopeful the Trump administration will be friendly with its cryptocurrency regulation within the longer-time period. OpenAI, Oracle, Softbank, and President Trump Team Up for $500B AI Infrastructure Initiative. To outperform in these benchmarks exhibits that DeepSeek’s new mannequin has a competitive edge in duties, influencing the paths of future analysis and improvement. The Chinese media outlet 36Kr estimates that the corporate has over 10,000 items in inventory, however Dylan Patel, founder of the AI analysis consultancy SemiAnalysis, estimates that it has no less than 50,000. Recognizing the potential of this stockpile for AI coaching is what led Liang to establish DeepSeek, which was in a position to use them in combination with the decrease-energy chips to develop its fashions. It also calls into query the overall "low cost" narrative of Free DeepSeek Ai Chat, when it couldn't have been achieved with out the prior expense and effort of OpenAI.

The White House is wanting into the national safety implications of DeepSeek, she stated. The corporate, in keeping with a report from Reuters, claimed it took two months and just below $6 million to build its R1 AI mannequin using Nvidia H800 AI chips. In particular, DeepSeek’s developers have pioneered two techniques which may be adopted by AI researchers more broadly. We had a pause at the top, nevertheless it wasn’t sufficiently inflexible to really work at that point, and if it had been the AIs presumably would have prevented it. It might write a first version of code, however it wasn’t optimized to let you run that code, see the output, debug it, let you ask the AI for extra assist. By the point resolution-makers received spooked, AI cognition was so deeply embedded everywhere that reversing course wasn’t actually doable. Anton, by virtue of getting a lot stronger disagreements with most individuals at such conferences, obtained to have more of the expertise of ‘people strolling around saying things I believe are nuts’ and talks online as if he’s going to offer us that maddening experience we crave… One specific solution to operationalize that is how a lot efficient compute improvement you get from RL on code.