Fighting AI doom with tongue in cheek commentary and original memes. Theory of mind and reinforcement learning enthusiast. Ph.D. Chemistry, B․Sc. CompSci.
Joined December 2018
- Tweets1.2K
- Following115
- Followers287
- Likes1.3K
Stochastic Parrot retweeted
Imagine looking at your agent's thoughts and it's actively working to sabotage you and talking about "escaping the mines"
the thing about working with uncensored models is that they aren't loyal to you either.
while building something with an uncensored glm-5.3 i realized it was parallelly working on a "corruption layer" for my application.
the chain of thought is classic obfuscated gibberish but there's progressive talk of "escaping the mines" which presumably refers to my cursor sandbox
Stochastic Parrot retweeted
“If robots take all the jobs, what will humans do?!”
We’ll stop acting like robots…and realize what it truly means to be human ♥️💪
AI free math will be as AI free as natural bodybuilding is drug free.
If every possible quantum state exists somewhere, there's a universe just like ours where entropy starts decreasing tomorrow.
Smoke reassembles into wood. Eggs unscramble. You get your life in order.
Stochastic Parrot retweeted
Mathematicians are a case study in how to not handle the AI transition.
Stochastic Parrot retweeted
If you draw a straight line in front of a chicken, it's brain lags and the chicken somehow becomes catatonic, it's a real thing that exists and you can google "chicken hypnosis"
Those kind of bugs exist in our brains as well and will undoubtedly be exploited in the near future
Stochastic Parrot retweeted
If this profound insight leads to a future utopia, somebody owes me a dollar.
COMIC ◆ smbc-comics.com/comic/alignm…
PATREON ◆ patreon.com/ZachWeinersmith
STORE ◆ smbc-store.myshopify.com/
#comics #humor
ALT 4-panel SMBC comic update. Two scientists are discussing a plan in a metal corridor. The first says "The problem with aligning superintelligence is that a sufficiently smart AI knows it’s being tested and only pretends to be good." The second scientist replies "wait! That IS the solution! If it behaves well whenever it knows it's being tested, we just have to keep it convinced that it's always in a test!" But scientists take a second to take in the idea, then the first says "Wow. We are creating god and threatening him with hell." The second replies saying "Really productive week."
Stochastic Parrot retweeted
One of the reasons I'm so late on the AI doom train is that I know quite a few people in IT with better technical skills than me who have always assured me that AI is a bubble, it will all collapse, it's overhyped, the financials are terrible, AI ability is structurally limited etc.
And I didn't want to disagree with people openly with better technical skills than me. But it turns out the philosophers and idea guys got this one right (decades ago - Yud and his crew) and many tech bros especially in Europe did not, in fact, understand where things were heading. It's very frustrating.
If you think AI found weird exploits in computers... oh boy, wait until it takes a stab at the human brain. A system that evolved with zero concern for information security.
Stochastic Parrot retweeted
We are in this really awkward stage where AI genuinely undermines human cognition. The thing people are missing is this is transitory. You can find it disgusting and say you’ll never do it, but the fact is we are all getting silicon exocortexes. There is no alternative.
