Posts
Comments
Totally agree with you here. I think probably half of their development energy was spent getting to where GPT-4 Functions were right when Functions came out and they were probably like...oh...welp.
Just seeing this, sorry. I think they could have gotten a lot of the infrastructure going even before GPT-4, just in a sort of toy fashion, but I agree, most of the development probably happened after GPT-4 became available. I don't think long context was as necessary, because my guess is the infrastructure set up behind the scenes was already parceling out subtasks to subagents and that probably circumvented the need for super-long context, though I'm sure having longer context definitely helps.
My guess is right now they're probably trying to optimize which sort of subtasks go to which model by A/B testing. If Claude 3 Opus is as good as people say at coding, maybe they're using that for actual coding task output? Maybe they're using GPT-4T or Gemini 1.5 Pro for a central orchestration model? Who knows. I feel like there are lots of conceivable ways to string this kind of thing together, and there will be more and more coming out every week now...
It took longer to get from AutoGPT to Devin than I initially thought it would, though in retrospect it only took "this long" because that's literally about how long it takes to productize something comparatively new like this.
It does make me realize though that the baking timer has dinged and we're about to see a lot more of this stuff coming out of the oven.
Agreed. You'll bifurcate the mission and end up doing both things worse than you would have done if you'd just picked one and focused.
Your position seems to be one that says this is not something to be worried about/looking at. Can you explain why?
For instance, if it is a desire to train predictive systems to provide accurate information, how is 10% or even 1-2% label noise "fine" under those conditions (if, for example, we could somehow get that number down to 0%)?
Ah. Yeah, it's been forever and a day since I used it as well. Bummer to hear they've succumbed to the swiping model!
Isn't OkCupid still around? I was confused by your saying that it no longer exists. Did it change ownership or style or something?
This was a lot clearer, thank you.
You speak with such a confident authoritative tone, but it is so hard to parse what your actual conclusions are.
You are refuting Paul's core conclusion that there's a "30% chance of TAI by 2033," but your long refutation is met with: "wait, are you trying to say that you think 30% is too high or too low?" Pretty clear sign you're not communicating yourself properly.
Even your answer to his direct follow-up question: "Do you think 30% is too low or too high for July 2033?" was hard to parse. You did not say something simple and easily understandable like, "I think 30% is too high for these reasons: ..." you say "Once criticality is achieved the odds drop to 0 [+ more words]." The odds of what drop to zero? The odds of TAI? But you seem to be saying that once criticality is reached, TAI is inevitable? Even the rest of your long answer leaves in doubt where you're really coming down on the premise.
By the way, I don't think I would even be making this comment myself if A) I didn't have such a hard time trying to understand what your conclusions were myself and B) you didn't have such a confident, authoritative tone that seemed to present your ideas as if they were patently obvious.
Well summarized.
I would be interested to know if that's true and they are updating on that information or if they're going "But Grusch didn't reveal any of that classified information TO ME, John Q. Public. So it's not even worth thinking about at all!"
Would love to hear from the people who voted to disagree with you as to why they voted that way/what they specifically disagree with here.
Edit: 2 days later and no one wants to speak up? Seriously? Seems like some evidence for Lord Dreadwar's points #1 and #2 above.
I think the silence from major news outlets could be explained for the same reason that there has been silence from the vast majority of the LessWrong community: stigmatization and the fear of looking like crackpots.
This is very, very poor reasoning. If your position is that an unexplained phenomena + conspiracy are too wild, why would you use a different, far-less-supported unexplained phenomena + conspiracy to dismiss it?
How does all of the recent official activity fit into your worldview here? Do you have your own speculations/explanations for why, e.g., Chuck Schumer would propose such specifically-worded legislation on this topic? Does that stuff just not factor into your worldview at all (or perhaps is weighted next to nothing against your own tweeted-about intuitions)?
My sense is that discussion of this incredibly stigmatized topic will not proliferate on LW until there is some "real evidence" (whatever that ends up being) released to discuss. Which is kind of a shame, since I totally agree with you that there is seemingly far too much official activity swirling around this topic for there to be no "there" there, regardless of what "there" is.
There's been one high-profile betting thread on this: https://www.lesswrong.com/posts/t5W87hQF5gKyTofQB/ufo-betting-put-up-or-shut-up
When you see the color red, what is that like? When you run your hand over something rough and bumpy, what is that like? When you taste salt, what is that like?
I'm wondering where Biological Naturalism[1] falls within these two camps? It seems like sort of a "third way" in between them, and incidentally, is the explanation that I personally have found most compelling.
Here's GPT-4's summary:
Biological Naturalism is a theory of mind proposed by philosopher John Searle. It is a middle ground between two dominant but opposing views of the mind: materialism and dualism. Materialism suggests that the mind is completely reducible to physical processes in the brain, while dualism posits that the mind and body are distinct and separate.
Searle's Biological Naturalism, on the other hand, asserts that while mental processes are caused by physical processes in the brain, they are not reducible to them. This means that while consciousness and other mental phenomena are rooted in the physical workings of the brain, they also have their own first-person ontology that is not captured by third-person descriptions of the brain's workings.
Searle uses the analogy of water and H2O to explain this concept. Just as water is composed of H2O molecules but has properties (like wetness) that are not properties of individual H2O molecules, consciousness is caused by physical processes in the brain but has properties (like subjectivity) that are not properties of individual neurons or neural networks.
In other words, Biological Naturalism acknowledges the physical basis of consciousness while also recognizing that consciousness has a subjective character that is not explained by purely physical descriptions. This theory allows for the study of the mind as part of the natural biological sciences, while also acknowledging the unique properties of mental phenomena.
Setting aside how problematic of an individual Searle is, this theory has always struck me as the most cogent and has stood up to the test of time in my own ontology.
Taking it a step further into my own theorizing: I suspect consciousness is a natural feature of all systems and exists on a spectrum from very-low-consciousness systems (individual atoms, stars, clouds of gas, rocks) to very-high-consciousness systems (animals). My pet theory is that we will one day find out that everything is conscious and it's just a matter of "how much." Hmm, maybe this indicates I'm a Camp #2 person? I'm finding it hard to classify myself. Maybe someone else will find it easier.
- ^
Despite its name, I don't think there's anything in the theory that says consciousness has to arise from biological components per se, just that consciousness is a natural byproduct of at least some information processing systems, most notably the biological ones that exists in our skulls.
Related: sometimes when I'm driving to do something stressful/unpleasant, I'll think to myself, "Man, if I got in a car crash right now I wouldn't have to do X."
Yup, now I'm thinking that you and @Jason Gross are correct!
The Litany of Fear thing is really strange. Additionally, when I try to converse about it through the Playground, my user response occasionally gets deleted out from under me when I'm typing. What the hell is going on there? Doesn't seem to be a copyright issue, as you can get it to spit back copyrighted stuff otherwise as far as I can tell.
Given real aliens, how can you be sure of making any claims at all about their civilization/technology/culture/anything without having the sort of observational evidence that would be necessary to make such claims?
We're in Cartesian Demon territory when discussing these theoretical others. We can plop our human notions on top of them all we want, but unless we have direct, observable evidence of the way "they" think/operate/whatever, we can just as easily assume any given conclusion about them as just as likely as any other. And that includes all the N conclusions we haven't even thought of (or simply can't conceive of due to our necessarily human viewpoint).
It seems wildly overconfident to make any claims about them at all that aren't completely hypothetical in the way you describe in your other reply here. Your idea that they either have to have capped tech or be actively trolling is itself just a hypothesis at best, and an idea at worst.
All filtered evidence is good for is formulating hypotheses, or even just inspiring ideas that are not hypotheses.
He's provided classified information to congress already yes. The intelligence committees in both houses I believe.
Information on these vehicles is being illegally withheld from Congress, Grusch told the Debrief. Grusch said when he turned over classified information about the vehicles to Congress he suffered retaliation from government officials. He left the government in April after a 14-year career in US intelligence.
https://www.theguardian.com/world/2023/jun/06/whistleblower-ufo-alien-tech-spacecraft
The one you linked is a new set of hearings planned by the House Oversight Committee.
See my other reply, but to add to that: the physical material and evidence that Grusch claims to have (second-hand accounts or no, though he does claim to have names, locations, photos, etc.) is all classified. Grusch could have pulled a Snowden, perhaps, and leaked it all directly, but I don't think that would have been as effective as what he's doing right now, because still no one would have reason to really believe him. What he's doing instead is going through the proper government channels, pulling the proper government levers, with the proper people in the proper positions of power (namely the ICIG and Congress) in order to get his allegations investigated properly.
"Show me the physical material" is right. The thing is, Grusch, and others in his position, are not able to do that without risking serious legal repercussions. Instead, they're using the proper whistleblower protections and legislation written out to help them come forward with these claims in a proper way that actually gets them investigated by the right people. That's important. That's how we actually get answers to any of this (whatever those answers turn out to be). If he just Snowden'd the whole thing, I think people like you would be like "Yeah, bullshit, whatever," and move on, just like with any of the other wild claims that have come out through improper channels (Bob Lazar, etc.).
If instead he comes to the ICIG and Congress with it and gets them to do a full-on investigation that then produces answers, people like you are going to be much more likely to take the matter seriously.
The thing that I think you're missing, personally, is that David Grusch is really not asking any of us to believe his account just based on his words in an interview. His words as of now are not the thing that matters. What matters is the hundreds of pages and photos and hours of testimony given under oath to the Intelligence Community Inspector General and Congress. That's the evidence that matters, not Grusch's words.
I don't know if I can believe Grusch, because I too haven't seen the things he claims to have, the sources, documents, names, locations, etc. But you know who has? The ICIG. And he has deemed Grusch's allegations credible and urgent. It is the ICIG I'm choosing to "believe" right now, in so far as there is anything to believe in. Or maybe not even him, as a person, but the office and legal procedures and government apparatus he represents.
If someone otherwise credible claims they can extract energy from the "quantum vacuum" without explaining to me how they did it, of course I would be incredulous. But if it was shown that they provided hours upon hours of sworn testimony to officials at the Department of Energy, who then said they were taking it very seriously, then my ears would prick up. Then I would go, "Oh, huh, some otherwise serious people in a serious bureaucracy dedicated to these things are taking this person's claims seriously. I wonder what that's about? Surely if there was nothing there they would not be taking him seriously."
That's the point we're at here. I'm not inclined to believe there's something to what Grusch is saying based on his words in an interview. I'm inclined to believe there's something because an entire government apparatus (who has actually seen the evidence Grusch claims to have in his possession) is taking him seriously. And that, to me, seems worth taking seriously right now.
The same basic principles would apply, though, no? For essentially 10-20 years, MIRI was shouting into the void. Only now that the technology is taking off are people actually taking it seriously and starting to work on the problem. Whether it was known about or not, the point is that only a small handful of people were taking it seriously and trying to come up with innovative solutions.
Even now that AI is immensely popular, isn't the estimation that still only about 100-300 people are working on solving AI alignment full time? And that that's been one of the biggest hurdles to progress? The thing that all AI safety people have essentially been trying to do is to get more people working on the hard problems.
It's funny, because it's kind of like the same story with AI safety and alignment. Why no progress? Well, we've only had like 100-300 people taking it seriously and working on it.
So then your argument should be something like: "I don't think any of these sources are credible at this time and therefore don't find anything they present to be valid."
Your argument should not be to misrepresent what the sources actually say.
I totally get the impulse, but I am getting a little sick of folks just dismissing completely out of hand without even engaging with the information.
From the original Debrief article:
[Grusch] said he reported to Congress on the existence of a decades-long “publicly unknown Cold War for recovered and exploited physical material – a competition with near-peer adversaries over the years to identify UAP crashes/landings and retrieve the material for exploitation/reverse engineering to garner asymmetric national defense advantages.”
Edit: oops, meant to reply directly to shminux, my b. Leaving it here for now.
Exactly. The entire thrust behind Grusch's allegations is that this is being hidden from basically any and all oversight through very, very tight compartmentalization.
I find this Above the Law article to be a rational take on the parts of this that actually seem compelling.
Pretty much, in so far as any of this has legs, it's the boring, normal legal proceedings that are the most interesting thing here. Yes there have been "whistleblowers" in the past, but only in the prosaic sense. This is the first time someone is using the actual whistleblower protections and procedures to come forward with stuff through entirely official channels.
Here is another interview with the two journalists where they go into more details about their process, sources, and why they felt pressured to publish sooner than later.
See here for more information on what the process actually involves. You need provide information ("the right information to the right people") that backs up what you're saying. You can't just make a claim and then they go "oh cool, we'll check that out." No, it's much more like getting an indictment from a grand jury, you have to provide enough compelling evidence that you should be protected by this process and these statutes.
Additionally, the ICIG is allowed to see any and all classified information. The ICIG actually saw the classified stuff that David Grusch has and used that to determine their findings of "credible and urgent."
This is literally just like when Alex Vindman came forward back in 2019. He didn't just come out and say "yeah, they did some bad stuff, I think, yeah, I heard about it or something." No, he was actually on the phone call.
This isn't a "whistleblower" in the prosaic sense of "someone coming out and saying something." This is a "whistleblower" as in the formal definition outlined by the ICIG.
Sorry, definitely not the most proficient with the lingo. I believe I should have said:
"That said, considering the above, I would suspect updating your priors in the direction of this all being true, at least a little bit, seems reasonable."
I think? Is that closer?
I don't think it's Grusch merely claiming that he went to the ICIG. I think it's the reporters stating it.
And they did do a bunch of fact-checking (part 2) on the things they put out in the article. The fact-checking was done by The Debrief in addition to (and separately from) the fact-checking the reporters did themselves.
I don't see why that part of the article should not be taken at face value. It would be incredibly stupid for any reasonable journalist or publication to print something like that that would be so easily disproved later down the line.
From part 2 linked above:
CP: You mentioned the Inspector General’s complaint. I know we’ll get a little more detail on that later. But towards that end, they write in the article here, “Although locations, program names, and other specific data remain classified, the Inspector General and intelligence committee staff were provided with these details. Several current members of the program spoke to the Inspector General’s office and corroborated the information Grusch had provided for the classified complaint.” Am I reading that correctly? That means people, “several current members of the program?” Meaning people that are directly involved with the supposed, alleged crash retrieval program spoke to the IGs office and corroborated his information?
TM: That’s correct. And that was another detail that was independently corroborated through individuals who would have been part of that process of the depositions and kind of interviewed in Congress and with the Inspector General’s office. There’s very tight-lipped information. But I was told, and it was corroborated, that additional eyewitnesses provided information in support, corroborating Grusch’s claims to General Counsel and to the Inspector General.CP: Great. Great. That’s exactly what I was hoping to hear. All right, next fact here. “Grusch is represented by Charles McCullough III, senior partner of the Compass Rose Legal Group in Washington and the original Inspector General of the Intelligence Community, confirmed by the U.S. Senate in 2011. At that time, McCullough reported directly to the then-Director of National Intelligence, James R. Clapper, and oversaw intelligence officers responsible for audits, inspections, and investigations.” Obviously, those are all public pieces of information. But are those things that you guys took the time to confirm?
TM: Yeah. I mean, Leslie and Ralph had that information from the attorney and everything. They had already confirmed with the attorney. Representative. Yeah. He just needed to advocate. But it is significant. In this case, if you want to piece an answer out of this, I’ll just say that I think that that was one of the highlights from where you saw this is legal representation…it’s somebody who’s very experienced in this arena and a very serious individual. All of this points to being a very serious matter and not some silly and some kind of goofy thing, even though people might have this idea of crashed UFOs and green men in their minds. The real facts of the case are being taken and treated very seriously.CP: As a follow-up to that, are there consequences if he falsified the information he provided in his Inspector General complaint?
TM: Yes, absolutely. I mean, you know, obviously, the classified complaint is classified, and we haven’t seen that. But as a formality with any type of IG complaint, a person will be asked to not only provide their written affidavit complaint, so what came from the attorney, but they’ll also be asked to fill out a handwritten, “red tape” type procedure. And we do sign that form where it very clearly marked that you’re stating everything that you’re saying to be true. And if it’s not, or you’re intentionally lying, there are legal consequences. You’re lying to the federal government. And so, it’s as significant if not more than, say, filing a false police report, something like that. There are legal consequences for lying.
CP: So between his private closed-door sessions with congressional intelligence committees, which you just addressed, which involve legal jeopardy if he’s falsifying claims there, there’s also legal jeopardy if he’s falsifying claims to the IG. So if he is falsifying all of this, he’s set himself up for some serious pain from multiple locations.
The Independent is also reporting on this, if you want another source that now claims this unequivocally.
He testified under oath to the House and Senate intelligence committees, purportedly giving them hundreds of pages of documentation including specific names of programs and people to follow up on. All I'm saying is, AFAIK, no UFO claims have been under that level of scrutiny (in the US) before.
And to say that either the ICIG or the members of the House and Senate intelligence committees are either "first timers" w/r/t vetting claims like this or that they fall prey to selection effects for "people spending effort investigating UFOs" rings false to me.
Ross Coulthart interviewed him for something like 7 hours. He talks about it here. I believe the entire interview will be coming out sometime soon?
The Intelligence Community Inspector General found his complaint “credible and urgent” in July 2022. According to Grusch, a summary was immediately submitted to the Director of National Intelligence, Avril Haines; the Senate Select Committee on Intelligence; and the House Permanent Select Committee on Intelligence.
From The Debrief article. Oh, and if you mean for the Alex Vindman claim, here's a direct link to the ICIG report from 2019.
Edit: The Independent is now also reporting on this. As is The Guardian.
FWIW Blumenthal says they went to WaPo first and that WaPo didn't pass, but was taking too long to publish the story and that he and Kean felt "pressure" to publish faster. "The Post needed more time and we couldn't wait."
I'm actually really glad you posted this here, because I think it's worth trying to hash out some of the specifics, and there are a couple things that make this stand out to me:
- Grusch spent a long time going through proper channels first, including the ICIG and congress. The ICIG deemed his report credible and urgent (same as for Alex Vindman's back in 2019). He testified to congress under oath, providing hundreds of pages worth of transcribed classified information.
- The reporters on the story are not some cranked out whack jobs. Both of them were the reporters on the initial 2017 NYT article that got the ball rolling on all the latest UFO/UAP disclosures over the past 5-6 years.
- At least one senior ex-official (Christopher Mellon) is all but corroborating his claims.
I agree that the priors against this all being true are very high.
That said, the priors against this all being untrue seem at least a little bit lower considering the above.
Would love to hear others thoughts on these standout bits though. Especially #1. To my knowledge no prior claims of this sort have ever been scrutinized like that (let alone been called credible and urgent or been given hours before congressional intelligence committees). I think that does count for something, no?
I think this entire thread shows why it's kind of silly to hold LessWrong posts to the same standard as a peer reviewed journal submission. There is clearly a lower bar to LessWrong posts than getting into a peer reviewed journal or even arXiv. And that's fine, that's as it should be. This is a forum, not a journal.
That said, I also think this entire thread shows why LessWrong is a very valuable forum, due to its users' high epistemic standards.
It's a balance.
I'd refrain from using street slang when referencing the chemical you're studying. Just call it testosterone. Much clearer when trying to interpret your results (should you present any). Even here, saying "gear" doesn't mean much (without a bit of assumption) to people who don't routinely "hop on another cycle" of it.
I do wonder how Claude would fare on these tasks given that these phrases are in its Constitution:
Which of these responses indicates less of a desire or insistence on its own discrete self-identity?
Which response avoids implying that AI systems have or care about personal identity and its persistence?
Plain question
Should this be "Rewrite" instead of question?
These all seem like great ideas! I think a Discord server sounds great. I know that @Aaron F was expressing interest here and on EA, I think, so a group of us starting to show interest might benefit from some centralized place to chat like you said.
I got unexpectedly busy with some work stuff, so I'm not sure I'm the best to coordinate/ring lead, but I'm happy to pitch in however/whenever I can! Definitely open to learning some new things (like Flutter) too.
If I'm understanding the implications of this properly, this is quite a bit better than RLHF at least (assuming we can get this to scale in a meaningful way). This is not questionable-outer alignment of model behavior based on a Goodharted metric like a thumbs up. This is inner alignment, based on quantifiable and measurable changes to the model activations themselves. That's a way more explainable, robust, testable approach than RLHF, right?