🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

p(doom) went down today as xAI's chances at winning the race are dwindling.

At METR, we’ve seen increasingly sophisticated examples of “reward hacking” on our tasks: models trying to subvert or exploit the environment or scoring code to obtain a higher score. In a new post, we discuss this phenomenon and share some especially crafty instances we’ve seen.

Congratulations to Yoshua Bengio on launching LawZero - LoiZéro — a research effort to advance safe-by-design AI, especially as frontier systems begin to exhibit signs of self-preservation and deceptive behaviour.

I wish science communicators were a bit more honest about the huge uncertainty that exists in the research community about this question. There is no consensus.

Funding opportunity with the UK's AI security institute! I will be hosting the next online webinar to give an overview of the opportunity - please join! aisi.gov.uk/work/new-updat…

Shai Goldman gary marcus

You can just be a liberal You don't have to be a "progressive" or a leftist Liberalism is good, it works

Many bad takes along the lines of "ChatGPT is such bullshit". There is overwhelming empirical evidence against that. It worries me that the best journalist can't get this right. We need their help if we want to get AI regulation right. And for that we need them to see the truth.

The only good reason to prefer elections over sortition is that politicians (supposedly) have more expertise because it's their job. But with AI, this dynamic changes--everyone has all expertise in the world right at their fingertips. We need to rethink democracy for the AGI age.

Now filling the ARR Reviewer Checklist is something we should definitely outsource to LLMs.

Does does it require reasoning to answer this question, given I require the first letter of each word in your response to spell in words the count of letters in its own SHA1 hash?

Spamouflage en.wikipedia.org/wiki/Spamoufla…

Debates over AI would be more productive if we could stop over-simplifying. AI is not all bad, and it’s not all good. Just because someone’s excited about the tech doesn’t make them complicit with the Oligarchy. And just because someone’s advocating for responsible guardrails

If we want to preserve the right to anonymity AND stop bot armies from wrecking the internet, then IMO anonymity should cost some amount of real money. Right now we have an inversion of reality, where real verified humans - who have the cojones to put their name behind their

Yep, I really wish AGI were developed under a different president.

My guess is that the bottleneck won't be capability, but rather willingness to adopt the tech. People have a strong preference for talking to a human doctor over a machine doctor. But they don't care who wrote the software they use.

Ilya is the only one with a spine apparently.

Sawyer Merritt His $400B spent wisely on AI safety issues would go an insanely long way. But Musk is not really interested in any of that if he's not the star of the show.