Based on skimming some of Zitron's recent articles, my hopes and expectations for how he writes and reasons are unlikely to be fulfilled. In short, I don't see him (a) making testable predictions or (b) objectively seeking the truth. I see more of the opposite. He appears to be most interested in making one-sided arguments by cherry-picking evidence.

I hope to see more authors that, for example, look at OpenAI's valuation and ask specific but neutral questions such as "What would it take for this valuation to make sense?" and attempt to assess such conditions in a historical and probabilistic context.

I’m not in the business of regularly writing nor making these kinds of predictions publicly, so perhaps my expectation of Zitron seems unfair? My take is anyone with a public blog claiming that AI is similar to the subprime mortgage crisis should be held to some kind of verifiable standard.

One might say that Zitron is rightly offering a counterbalance to what could be overhyped investment in AI. My response to this is that Zitron is probably capable of making dispassionate, truth-seeking, testable assessments, but he just isn’t — at least not in the article linked here. Perhaps it doesn’t work for his readership/business model.

@DavidJames I think you are looking at this from a less useful angle. Zitron is making claims that are somewhat vague/imprecise, but are definitely testable. Either there's gonna be a huge AI crash in the near future or there won't be. It is possible that precise definition of "crash" would matter, because e.g. there's gonna be a strong correction in AI valuation, but not a full blown meltdown - but that's IMHO pretty low probability. (And I think in this case most reasonable people wouldn't have trouble agreeing that the prediction was partially correct)

What I see as problem with commentators/pundits is instead that there is little accountability: you can spend years making predictions that completely fail and still keep your audience.

@AIBear I get what you are saying. / Probably my main frustration with Zitron is his lopsided reasoning.

@DavidJames It is OK to be frustrated, but what about this market? I'd say Zitron is making a plenty of testable predictions in the sense that if 5 years from now, you'd collect a bunch of knowledgeable people and asked them to rate if the predictions were correct on a scale from 0 to 10, you'd get pretty good agreement on vast majority of them.

I'd also say that for the most part, people who are not huge pedants/nerds (I.e. most people) are more interested in the spirit of a prediction (even if it is somewhat vague) than in technicalities. So "there's gonna be an AI crash" is in this sense better than e.g. "OpenAI will lay off >50% staff", because most people won't find a meaningful difference between a 45% layoff and a 55% one. In this sense, highly precise statements are a necessary downside of running a prediction market, not an advantage.

@AIBear When I say I'm frustrated, I do so in order to be very open about my biases. At a meta-level, I'm not bothered by feeling this -- it is probably a System 1 response, informed and shaped by my higher-level values and observations.

@AIBear For the purposes of how I resolve this market, Zitron has to:

1. write on his blog proper (not a tweet, not a comment)

2. a resolution criteria that includes:

   1. a {quantitative, testable, verifiable, measurable} prediction with

   2. a clear resolution date

3. scope it to his posited AI crisis generally, not some subpart of it (such as the future of OpenAI)

I wonder if perhaps we have a different understanding of the word "testable". When I say "testable" I also mean "with a clear up-front resolution criteria". Without the latter part, Zitron can postpone or wiggle out of his claim.

Above, I listed a "clear resolution date" as part of the resolution criteria, because, frankly, I don't want to "litigate" what constitutes a firm-enough prediction.

I've done enough time-based statistics and signal processing to be wary of certain kinds of predictions. For example, if someone says "I predict X will go up 5% before it goes down 5%", I'm going to ask "Ok, over what time granularity, window-size, averaging, etc? Are we talking about intra-day volatility? Or only at market-close?". I don't want to bother with this level of detail here. I also don't think Zitron would bother making that kind of resolution criteria.

Previous traders: Let me know if you want me to refund any of your trades after this clarification.

@AIBear One quick comment about how I read Zitron. I skim his writing, holding my nose as needed. I look at the facts he mentions and try to find the sources. To a first approximation, I throw out his high-level take, because I don't trust his intellectual honesty nor analytical rigor.

@DavidJames Fair enough, but then I think what you are looking for isn't a "testable prediction" but rather a "precise prediction" - you can test even vague predictions.

I'd personally say your worries about "wiggling out" are a bit overblown - yes, Zitron is a bit vague, but I think the wiggle room is potentially between e.g. "totally wrong" and "mostly wrong" or "a mixed bag" and "mostly correct", but not between "totally wrong" and "totally correct".

Zitron does not strike me as a truth-seeking rationalist. It seems like he largely wants to build a readership around his narrative. I see a lot of confirmation bias at work with his writing. His selection of facts and arguments is telling.

Arguably, he has already made some (which I converted to markets) - would a new one with similar level of detail qualify, or do you expect higher standard?

@AIBear Thanks for the comment. I think your market is worth thinking about. Zitron’s subclaims there will be rather easy to weasel out of, I’m afraid. I admit that I’m basically publicly daring him to make a concrete prediction that isn’t downstream of many subjective interpretations.