eynsham

Gödel’s second incompleteness theorem shows that you can prove anything, without that proof being meaningful is a lens into that.

What has Gödel incompleteness to do with that? We can just take any sentence φ as an axiom, and we’ve a trivial proof thereof.

Most people at e.g. Oxford and Cambridge (and, I suspect, many other universities) use their university emails for a fairly wide variety of extramural correspondence, and are stuck with M$ as the provider, alas.

To get in, you need 4 A* from an independent schools

This is not, generally, true. Cambridge tends to have higher offers, and I had a friend offered 5A*+S1 (STEP), but that’s mostly because he buggered up his interview (and he was taking 5 A-levels). The standard offer at Oxford for mathematics is A*A*A, which is exactly what those applying from any public school will be asked to achieve.

or just 3 As from state schools

Not for maths. For easy courses like PPE, someone who misses a standard offer of A*AA might be let off by a few grades. People at public schools who miss their offers are much likelier to be a bit thick.

What do you think fully homomorphic encryption is, then?

I don’t see how free software licences ‘sound too good to be true’, in that they don’t even sound relevantly good: they don’t suggest in the slightest that support will be available.

Gates is one of the best placed people to claim that the advantages of meeting people in person &c in disbursing his wealth (rather than just ‘giv[ing] talks’) outweigh the emissions. There’re obviously diminishing returns (do COP delegations need to be so large?) but it would be surprising if the optimum were zero flights.

The European Investment Bank provided €1.5b of funding.¹ EIB decisions don’t generally attract lots of attention from member states other than those concerned, since it is generally understood that the EIB is funding lots of projects in member states simultaneously. Similarly, the budget of the Commission and similar bodies will generally be set in advance, usually with a formal or informal understanding as to the broad distribution of funds between member states that will follow.

In any case, this project seems to me to be no more extraordinary than the redistributive effects of e.g. Medicaid or Pentagon spending, or the construction of Interstates. The Interstates, in the present US political environment, might indeed seem extraordinary; but the question is then not how one convinces people from state A to spend on state B, but how to convince people to make large long-term investments in the first place.

1: https://web.archive.org/web/20130111042126/http://archiviost...

It’s rather odd that the ICO is saying this rather than industry lobbyists. No 10 is full of middle-aged technically unsophisticated types who want to look otherwise, and substitute credulity for wordliness, and I think Quangoland’s denizens are following the party line. (The shadow cabinet, but they aren’t that focused on growth because they have the luxury of not having to run the country.) The ensuing disasters may well be far worse than Horizon, if HMG ever gets anywhere with its AI plans.

Most important legislation, including the BNA, is government legislation (indeed, see the white paper: https://www.uniset.ca/naty/maternity/wpaper.pdf). It is therefore drafted by parliamentary counsel, whose advice remains available when amendments are proposed. Most governments also command sufficient majorities to push this kind of legislation through, or at least to come to consensus on amendments. The relevant passage seems clear enough that parliamentary counsel could have drafted it and so I doubt there were ‘too many chefs’ as you put it (although I haven’t checked Hansard).

It is also hard to see what these drafting habits have to do with the common law system. Points 1–5 could be true of a legislature in pretty much any legal tradition.

I think that it would help if you were to suggest a term people who don’t want to ‘shut down discussion about related topics all together’ should use. Otherwise, the effect (although perhaps not the intention) of deprecating the term ‘attacks on trans people’ is that the sort of discussion you admit is possible theoretically will be impossible for want of a suitable term to designate the sorts of attacks it concerns.

Is it not possible that ‘attacking trans people’ is both (sometimes) a euphemism for criticism of maximalist positions and (at other times) a perfectly normal term that designates approximately what ‘attacking x’ generally means? There is such a thing as an unsubstantive and utterly unpleasant insult explicitly motivated by the fact that its target is trans. Many trans people say that there are many such, and one does not need to believe everything that trans people say (surely with the result of inconsistency!) to think that the evidence they present is not wholly concocted.

Others may misidentify respectable, good, or correct arguments as ‘attacks’ in narrower senses, but that no more makes the underlying categories meaningless than the misapplication of such descriptions as ‘true’, ‘valid’, ‘scientifically established’, or ‘by definition’. I have no general pithy answer to what one should do about the sorts of attack I have described, but I venture that it is reasonable to talk or attempt to do something about them. What term would you prefer?

Have you ever

Yes, often.

‘an historical occasion’ - this is wrong

There is nothing wrong with the use of ‘an’ before ‘history’ and the forms of that lemma. It is unusual, and perhaps pretentious or otherwise silly, but only in the same way as ‘connexion’ is an unusual (&c) but perfectly correct spelling.

No, ‘all’ is not subjective. My threshold for ‘any’ is one, and threshold for all is, well, all of them. My threshold for a majority is the usual one, as is my threshold for eighty or sixty per cent. I am deriving the worldview that it would be absurd to deny that there is at least one dissident. It doesn’t matter whether I derive this based on an understanding of a sample that is representative of China as a whole so long as it is representative of a subset, since that is such a modest claim.

If you want to make a different point about the majority, or a very big majority, or two people, you should make that point instead. Obviously the Chinese with whom I interact are of a certain class and intellectual formation of which I must remain aware; in particular, the closeness of some to the party breeds a unique type of contempt. That is why I am not trying to generalise; seemingly, you are.

There are obviously good reasons to doubt the cruder ‘fantastical views’ that circulate in the West, but I cannot see why those reasons would lead one to doubt that there is a single ‘true Chinese dissident’. Am I to believe that every Chinese overseas I have met who freely discussed politics in a régime-sceptical way was paid off by the CIA or just trying to please me? Is every overseas dissident really just an empty NED shell? (I am sure many take money, but that doesn’t mean they also don’t have their own beliefs.) Does survey evidence that most Chinese are satisfied really mean that /all/ are? Are living standards so improved that nobody could possibly oppose the central people’s government? Is 润学 completely fabricated or reducible to economic factors?

Coetzee asked otherwise, although most South Africans pronouncing his surname would say /eə/ or /ɪə/, yes.

https://www.bbc.co.uk/blogs/magazinemonitor/2009/09/how_to_s...

Doxxing is when someone aggregates publicly available information, I see.

This and many other suggestions have been explored, and usually found wanting (see e.g. https://plato.stanford.edu/entries/knowledge-analysis/#NoFal...).

Why wouldn’t Britons also have brain fog? See chart 1 https://www.gov.uk/government/statistics/reported-road-casua....

My bike, with a Sturmey Archer three-speed, cost £160 (from a very reputable shop that services pretty much everything for free or only charges for parts). It is perfectly good for commuting and hobby rides up to at least 50 miles—and I say this as someone rather out of shape.

First, thanks for the conatructive tone. Every day I am astonished that people directly impacted—whose friends died and might have been sexually abused on 7.10 and people whose families have been blown to pieces by aerial bombardment—somehow still manage to be more reasonable (on occasion) than many of their supporters in the West.

Second, I am not sure that this is a ‘small’ minority. Even ignoring the Channel 14 polls (which do seem to be made up) the current coalition is on about 50 seats (looking at https://en.wikipedia.org/wiki/Opinion_polling_for_the_next_I...). And then I am not really convinced that Sa’ar or Lieberman would really be opposed to this specific action, which would make a majority. For that matter, I saw one poll saying a Sa’ar+Cohen+Bennett tie-up would hit a whole thirty-two seats and reduce the Likud to twelve (that's the one labelled ‘Merger of Gideon Sa'ar, Naftali Bennett and Yossi Cohen & Labor and Meretz under Yair Golan’…) Admittedly maybe Lieberman would say that it’s stupid but that’s different from saying it’s disgraceful.

I don’t see how the move out of Cairo really indicates a loss of control anywhere else. It’s not like all the army bases in the Sinai or Alexandria or anywhere else are going anywhere. And there are good reasons to leave Cairo otherwise—for example, the severe mismanagement of the Cairene urban environment, traffic, &c. As for the second point, there have always been Islamists more radical than the Muslim Brotherhood, and Daësh if anything has less power now than it did at its peak.

What change do you see in the institutional power of the army that would allow that to happen? And why do you think the Muslim Brotherhood would be so much more radical than it was under Morsi?

This still doesn’t address the claim that ‘first’ is wrong, and that sort of hyperbole, if wrong, is quite off-putting. If this is really ‘quite promising compared to existing…solutions’, there’s no need for inflated claims.

I am not sure that the ambiguity has to work that way. The suggestion I am making is that if we fix the distribution on questions and the training data, we might (a) know the value of X in this specific case, (b) be able to ensure that it is fairly high.

I’d say this is the murky case because on the fixed training data and query distribution X ≈ 1 and we know that even though we don’t know the value of X on other training data and other query distributions. I think that might be where the disagreement lies.

I think the thought experiment can be set up both ways. In one case Eric has a fixed probability of getting /any/ query right. (This might leave boosting open so we might want to gerrymander repeated queries out.) In the other this is relative to the distribution of queries.

I am glad that you think it an excellent point!

I think that this might be a nice way of getting round my objection, but there is one worry, which is that X is relative to a distribution on the questions we ask when we aren’t dealing with Encyclopedia Eric but with an LLM. I don’t actually use LLMs very much myself, partly out of arrogance and Luddite tendencies. But I suspect that the value of X for some sorts of questions (simple quiz questions, maybe?) and some LLMs (maybe not Google’s) will be high enough to end up in the murky case.

Of course, both you and I agree that there /is/ clearly a difference. I can see the attraction of appealing to the intuitive or pretheoretic notion of knowledge, since it’s a fairly straightforward way of stating the difference and it’s not obvious how else one might put it (I suppose ‘LLMs don’t explicitly think through the facts stored when deciding what to say’ is one way of putting it.)

I remember some time ago rather sleepily watching John Hawthorne talk about conditionals; my sole memory was of his banging on about ‘the little logician in the brain’ (I think the point was something like: some conditionals seem [in]felicitous in virtue of form because the little logician in the brain is reading them; others seem infelicitous because we examine them more closely and look e.g. at the referents involved, in which case e.g. Gricean considerations apply). One difference at least in the case of LLMs that makes sense to me is that there is no ‘little logician’ in LLMs.

I think the parent comment ultimately concerned the reliability of /beliefs derived from text in reference works v text output by LLMs/, and that seems to be what the replies by the commenter concern. If the point is merely that the text output by LLMs does not really reflect belief but the text in a dictionary reflects belief (of the person writing it), it is well-taken. Since it is fairly obvious and I think the original comment really was about the first question, I address the first rather than second question.

The point you make might be regarded as an argument about the first question. In each case, the ‘chain of custody’ (as the parent comment put it) is compared and some condition is proposed. The condition explicitly considered in the first question was reliability; it was suggested that reliability is not enough, because it isn’t justification (which we can understand pretheoretically, ignoring the post-Gettier literature). My point was that we can’t circumvent the post-Gettier literature because at least one seemingly plausible view of justification is just reliability, and so that needs to be rejected Gettier-style (see e.g. BonJour on clairvoyance). The condition one might read into your point here is something like: if in the ‘chain of custody’ some text is generated by something that is incapable of belief, the text at the end of the chain loses some sort of epistemic virtue (for example, beliefs acquired on reading it may not amount to knowledge). Thus,

text from an LLM is more suspect than text from a dictionary.

I am not sure that this is right. If I have a computer generate a proof of a proposition, I know the proposition thereby proved, even though ‘the text is entirely generated by an entity which is not capable of having justified true beliefs’ (or, arguably, beliefs at all). Or, even more prosaically, if I give a computer a list of capital cities, and then write a simple program to take the name of a country and output e.g. ‘[t]he capital of France is Paris’, the computer generates the text and is incapable of belief, but, in many circumstances, it is plausible to think that one thereby comes to know the fact output.

I don’t think that that is a reductio of the point about LLMs, because the output of LLMs is different from the output of, for example, an algorithm that searches for a formally verified proof, and the mechanisms by which it is generated also are.

That reading of the comment did occur to me, but I think neither dictionaries nor LLMs are capable of belief, and the comment was about the status of beliefs derived from them.

it’s not /justified/ belief

Beliefs derived from the output of LLMs that are ‘right most of the time’ pass one facially plausible precisification of ‘justification’ in that they are generated by a reliable belief-generation mechanism (see e.g. Goldman). To block this point one must engage with the post-Gettier literature at least to some extent. There is a clear difference between beliefs induced by reading the outputs of LLMs and those induced by the contents of a reference work, but it is inessential to the point and arguably muddies the water to present the distinction as difference in status as knowledge or non-knowledge.