Mostly thoughts about programming. Maybe other stuff too.
It’s been a few months since I personally ran out of places to move the AI goalposts to. And the news keeps coming. An internal model from OpenAI just refuted a long-standing mathematical conjecture. For the first time, AI has produced an indisputably novel and significant piece of mathematical knowledge. Several years ago, a lot of people would have said, yes, this is definitely intelligence. But…
Does anybody care about “look what AI can do” posts anymore? We’re running out of places to move the goalposts to, but expectations are adjusting quickly. What would have been a huge breakthrough a year ago is a yawn today. Given that AI is solving open math problems, this little piece of news might not seem very exciting, but I figured I would share it anyway. Namely: ChatGPT is able to reproduce…
The history of artificial intelligence is full of moved goalposts. For example, chess was long thought to be a holy grail of intelligence. A solution to chess would be the solution to intelligence itself. But then chess got solved, and it didn’t seem like intelligence had been achieved. It turned out that chess wasn’t all that important after all, and intelligence was still out of reach. Goalposts…
I am pleased to announce that I was able to add a new feature to the Rust Clippy linter. Namely: the use_self lint will now notify that the Self keyword can be used in recursive type definitions. This feature is now officially available in the Nightly release. Hip hip hooray.
I have to admit it’s getting better A little better all the time – Lennon-McCartney
Recently I found an interesting Turing machine program. When started on a blank tape, it runs for more than 101565 steps before halting. Here it is1:
In the Busy Beaver game we ask: what is the longest that a Turing machine program of N states and K colors can run before halting when started on the blank tape? Answering this question requires enumerating every such program and checking whether or not it halts. But there is a big problem: there are just too many to check. The number of programs grows multiple-exponentially with N and K, O(nknk).…
In the classic Busy Beaver game we ask: what is the longest that a Turing machine program of N states and K colors can run before halting when started on the blank tape? The basic approach to solving this problem is to generate a list of candidate programs, then subject each program to a sequence of deciders, where a decider is a function that takes a program as input and returns a result of type…
I have some Rust code that does a bunch of pure computation. A lot of it. To solve basically an elaborate combinatorial problem.
The other day I found a cool new 3-state 3-color Turing machine program. Here it is: