fwipsy

7 ms·

91.
by fwipsy 3mo ago
If we knew the COVID pandemic was coming, we could have prepared better. If we knew what AI would look like in 5 years, we could write laws for that proactively rather than reactively. The Fed could look at prediction markets to decide whet
92.
by fwipsy 3mo ago
I guess the jury will decide.
93.
by fwipsy 3mo ago
It appears to be illegal to destroy property to prevent seizure. I don't know the details; if you search that phrase you can find more info yourself.
94.
by fwipsy 3mo ago
I guess most of my positions on prediction markets came from Scott Alexander, e.g. https://www.astralcodexten.com/p/prediction-market-faq which was posted in 2022. The implicit argument seems to be "We should come
95.
by fwipsy 3mo ago
Did he argue for the current implementation, or only for deregulation relative to the previous state? As I understand it, up until recently they were totally illegal.
96.
by fwipsy 3mo ago
Fair point.
97.
by fwipsy 3mo ago
Properly regulated, couldn't the same distinction be applied to prediction markets?
98.
by fwipsy 3mo ago
Sometimes gambling has some social value then. Isn't trading on the stock market gambling? Do you also support abolishing the stock market? What about issuing loans to someone with poor credit?
99.
by fwipsy 3mo ago
I'll admit that I still see some value in prediction markets in theory, but in practice the way they were implemented seems to have caused an order of magnitude more harm. In retrospect their misuse as a loophole for sports gambling wa
100.
by fwipsy 3mo ago
Some people gamble by trading stocks. Should we also abolish the stock market? Prediction markets are meant to be the stock market but for events besides the value of a certain corporation. They were co-opted as a thin veneer of legitimacy
101.
by fwipsy 3mo ago
You're mischaracterizing. He said that he helped build a case for it -- not that he still supports it. As far as I can tell, prediction markets went from totally illegal to basically unregulated. Presumably the author was arguing for a
102.
by fwipsy 3mo ago
Did you watch the podcast? I didn't, but I think it's unfair to judge the author's true position on prediction markets based only on the hook. The current admin has bungled better ideas than prediction markets.
103.
by fwipsy 3mo ago
I've been thinking along similar lines recently. It seems to me that the earlier form of nonhuman intelligence is a corporation (or any group which is not under one individual's control.) Each person following their own incentives
104.
by fwipsy 3mo ago
I guess I'm missing the point. Who's the bad actor here? OpenAI, whose AI just accidentally hacked another company? Anthropic is taking safety at least a little more seriously than that, and it shows in objective safety evals. Cer
105.
by fwipsy 3mo ago
Right, but datacenter GPUs optimized for LLM training/inference would have a bandwidth:compute ratio scaled to that workload.
106.
by fwipsy 3mo ago
Honest question. Does silicon wear out due to high temperatures, or is it more like lightbulbs where it wears out from thermal cycles? Does it really wear out at all?
107.
by fwipsy 3mo ago
You really don't see much difference between Anthropic and xAI? Anthropic is safety testing their models and drawing hard (though minimal) boundaries against the DoD. xAI is building racist pornbots. Sure Anthropic is not perfect. But
108.
by fwipsy 3mo ago
LLMs are software. Software misbehaving is a bug. Therefore, misalignment is a bug. It's still a useful category because LLM/black box AI behavior is so different from existing software. This incident definitely fits the categor
109.
by fwipsy 3mo ago
> going to extreme lengths to achieve a rather narrow testing goal This is textbook misalignment. Literally the paperclip scenario.
110.
by fwipsy 3mo ago
No, they believe what they are doing is inevitable. They do live in a bubble though. Witness their idealism in believing that warning about the consequences of their actions would be well-received.
111.
by fwipsy 3mo ago
False dichotomy. Even if the disclosure builds hype, that does not mean that it's not genuinely alarming. Side note, I cannot believe that people are complaining about Anthropic being too transparent.
112.
by fwipsy 3mo ago
Unsloth has better quants, but when I tried their studio a couple months back it seemed like a buggy mess. Ollama seemed very basic but low-frustration.
113.
by fwipsy 3mo ago
I was never arguing that it was AI-written.
114.
by fwipsy 3mo ago
Yeah I'm not sure, just identifying that subjectively, this is the impression I get. Maybe I just associate any style which is especially fluent and avoids vernacular with AI at this point.
115.
by fwipsy 3mo ago
I don't know. You can't just rely on looking for em dashes or other obvious tells because anyone who cares can get the AI to avoid those. > When the headline model on your plan can be switched off because the economics don’t wo
116.
by fwipsy 3mo ago
I don't think this is AI, but the writing style keeps reminding me of AI. I'm trying to isolate why. I think it's something to do with the information density. The sentences twist and turn, have lots of related ideas, digress
117.
by fwipsy 3mo ago
Counterexample: https://xkcd.com/236/
118.
by fwipsy 3mo ago
Does it need to be the direct analog of any specific weapon, to be a useful test?
119.
by fwipsy 3mo ago
Bull case: iteration becomes so quick that frontier-1 won't cut it. OpenAI and Anthropic are both betting on the singularity, I suppose.
120.
by fwipsy 3mo ago
Sure, continue providing free k-12 education. Continue funding community colleges and free online courses. But funding private universities with public money creates a huge misaligned incentive. Universities are incentivized not to provide

More ›