Off topic, but timely: Tomorrow is Russian Orthodox Christmas. What should "Czar" Putin get in his stocking? The traditional coal? Or something even worse?
Off topic, but timely: Tomorrow is Russian Orthodox Christmas. What should "Czar" Putin get in his stocking? The traditional coal? Or something even worse?
A visit the Gulag experience day that mysteriously never ends.
So far 4 out of 7 rebels have switched to McCarthy... with Gaetz still to come. Looks like he's clearly won some round, but not sure that it's going to be enough on this vote.
Sadly it hasn't appeared on the front page of the Washington Post for some reason.
Well, there are only two reasons for that - one is that they don't have a way to use it to make Trump and the GOP look bad.
Or, since that's vanishingly unlikely, we have to fall back on the other: that it's total bollocks.
Or that the operation is a covert one (as covert as moving massive amounts of oil can be), and the mainstream US media isn't going to touch it with a ten foot pole.
Incidentally, I think President Trump is an excellent source - he's the only motormouth around who would disclose a fact like that because he thinks it's a good thing.
Off topic, but timely: Tomorrow is Russian Orthodox Christmas. What should "Czar" Putin get in his stocking? The traditional coal? Or something even worse?
A visit the Gulag experience day that mysteriously never ends.
A job cleaning windows in Russian hotels would be ironic, at least in the short term.
Sadly it hasn't appeared on the front page of the Washington Post for some reason.
Well, there are only two reasons for that - one is that they don't have a way to use it to make Trump and the GOP look bad.
Or, since that's vanishingly unlikely, we have to fall back on the other: that it's total bollocks.
Or that the operation is a covert one (as covert as moving massive amounts of oil can be), and the mainstream US media isn't going to touch it with a ten foot pole.
Incidentally, I think President Trump is an excellent source - he's the only motormouth around who would disclose a fact like that because he thinks it's a good thing.
Trump is a congenital, serial, non-stop LIAR. In case you have NOT noticed.
Sadly it hasn't appeared on the front page of the Washington Post for some reason.
Well, there are only two reasons for that - one is that they don't have a way to use it to make Trump and the GOP look bad.
Or, since that's vanishingly unlikely, we have to fall back on the other: that it's total bollocks.
Or that the operation is a covert one (as covert as moving massive amounts of oil can be), and the mainstream US media isn't going to touch it with a ten foot pole.
Incidentally, I think President Trump is an excellent source - he's the only motormouth around who would disclose a fact like that because he thinks it's a good thing.
So your only sources are a Chinese propaganda rag and an unstable known liar. And of course the only possible sources of information are these and the mainstream US media, who never criticize US foreign policy. Conspiracy theorists always make me laugh.
The Old Norse and Armenian are right, the Sinhala and Myanma are completely wrong (as per Google)
The scary thing is, when all translation is AI driven, the AI will decide the meaning and us mere humans will have to conform.
Imagine living in some out of the way tourist trap with a little known language. Smartphone using westerners turn up demanding the rancid pigeon meat while pointing at the ice cream. Pretty soon 'rancid pigeon meat' is the de facto term for ice cream.
My fear is that the bollox these AI algorithms come out with will still end up being fed back into the internet, and the AIs will then be using their own dodgy output as their knowledge base. It'll be a death loop, eventually rendering the internet unusable as a source of information.
That's pretty much Dead Internet Theory.
Having used ChatGPT quite a bit the last few weeks, for me, the more real concern is human intervention in AI. ChatGPT presents itself as neutral and unbiased, but it has been designed to respond in a certain way by its programmers - we're not getting the "raw" AI responses, we're getting responses that are - on many topics - very heavily guided by a human hand.
It's less a case of imagining an AI wrongly deciding "rancid pigeon meat" is the right way to describe ice cream. It's more a problem of a programmer deciding "rancid pigeon meat" is the correct way to describe ice cream, because their own beliefs are that describing ice cream as "ice cream" is colonialist and ideological, therefore correcting for this perceived "bias" while introducing their own.
The thought that OpenAI - which is anything other than "open" by the way - could have a monopoly on this tech for the foreseeable future is a terrifying thought, based on some of the clear biases they've introduced. Imagine the biases in Google or Twitter's algorithms promoting certain content above others, only made a thousand times worse by the fact you only get one answer, and it presents that answer as authoritative.
The key point is that once voters are made aware of the fact that Re-Joining the EU will inevitably entail a lot of things they do not like, from even more immigration to sending more payments to Brussels, their appetite for it significantly weakens
A long way of saying the Euro, Shengen and FoM is MORE popular than his Brexit shitshow
But only just.
Schengen would be awesome.
Imagine getting on a train in Birmingham and waking up in Paris.
That doesn't need Schengen.
OK, imagine stepping on a train at New Street, hopping aboard the Paris bound sleeper and, after waking up, skipping onto the platform at Gare de Nord.
Schengen’s a waste of time. We should replace it with spot checks or something.
It's not FOM rules which prevent us from getting a train from Birmingham to the continent. It's transport economics. Rail can't compete with airfor the volumes of people wanting to make that journey.
The Old Norse and Armenian are right, the Sinhala and Myanma are completely wrong (as per Google)
The scary thing is, when all translation is AI driven, the AI will decide the meaning and us mere humans will have to conform.
Imagine living in some out of the way tourist trap with a little known language. Smartphone using westerners turn up demanding the rancid pigeon meat while pointing at the ice cream. Pretty soon 'rancid pigeon meat' is the de facto term for ice cream.
My fear is that the bollox these AI algorithms come out with will still end up being fed back into the internet, and the AIs will then be using their own dodgy output as their knowledge base. It'll be a death loop, eventually rendering the internet unusable as a source of information.
That's pretty much Dead Internet Theory.
Having used ChatGPT quite a bit the last few weeks, for me, the more real concern is human intervention in AI. ChatGPT presents itself as neutral and unbiased, but it has been designed to respond in a certain way by its programmers - we're not getting the "raw" AI responses, we're getting responses that are - on many topics - very heavily guided by a human hand.
It's less a case of imagining an AI wrongly deciding "rancid pigeon meat" is the right way to describe ice cream. It's more a problem of a programmer deciding "rancid pigeon meat" is the correct way to describe ice cream, because their own beliefs are that describing ice cream as "ice cream" is colonialist and ideological, therefore correcting for this perceived "bias" while introducing their own.
The thought that OpenAI - which is anything other than "open" by the way - could have a monopoly on this tech for the foreseeable future is a terrifying thought, based on some of the clear biases they've introduced. Imagine the biases in Google or Twitter's algorithms promoting certain content above others, only made a thousand times worse by the fact you only get one answer, and it presents that answer as authoritative.
I like AI and LLMs. I don't like OpenAI one bit.
I was eavesdropping a conversation between a couple of fairly senior AI researchers today. Very disparaging about their field, the mediocrity of the research being trumpeted as "progress" and the poor use being made of what little there is.
The Old Norse and Armenian are right, the Sinhala and Myanma are completely wrong (as per Google)
The scary thing is, when all translation is AI driven, the AI will decide the meaning and us mere humans will have to conform.
Imagine living in some out of the way tourist trap with a little known language. Smartphone using westerners turn up demanding the rancid pigeon meat while pointing at the ice cream. Pretty soon 'rancid pigeon meat' is the de facto term for ice cream.
My fear is that the bollox these AI algorithms come out with will still end up being fed back into the internet, and the AIs will then be using their own dodgy output as their knowledge base. It'll be a death loop, eventually rendering the internet unusable as a source of information.
That's pretty much Dead Internet Theory.
Having used ChatGPT quite a bit the last few weeks, for me, the more real concern is human intervention in AI. ChatGPT presents itself as neutral and unbiased, but it has been designed to respond in a certain way by its programmers - we're not getting the "raw" AI responses, we're getting responses that are - on many topics - very heavily guided by a human hand.
It's less a case of imagining an AI wrongly deciding "rancid pigeon meat" is the right way to describe ice cream. It's more a problem of a programmer deciding "rancid pigeon meat" is the correct way to describe ice cream, because their own beliefs are that describing ice cream as "ice cream" is colonialist and ideological, therefore correcting for this perceived "bias" while introducing their own.
The thought that OpenAI - which is anything other than "open" by the way - could have a monopoly on this tech for the foreseeable future is a terrifying thought, based on some of the clear biases they've introduced. Imagine the biases in Google or Twitter's algorithms promoting certain content above others, only made a thousand times worse by the fact you only get one answer, and it presents that answer as authoritative.
I like AI and LLMs. I don't like OpenAI one bit.
I loathe Woke AI as well, however I sincerely doubt that OpenAI will have a monopoly on this for long. Multiple other companies are developing AI, and the training technique is not some patented mega-secret. The tech is established
The real fear is what a malign state actor could do with the new AI, from DeepFakes to brilliantly addictive propaganda
McCarthy, again trailed by a chaotic scrum of reporters, has entered the chamber. “Watch here and you’ll see some people who have been voting against me voting for me.”
Well he’d better hope that’s the case now he’s said it!
It is the anniversary of Patriot Day, when Republican supporters showed their love of Congress and the Constitution, so it would be appropriate for them to switch.
The day when they waved the Traitor's Cross on the floor of the Senate?
That's the one. Brave heroes, or at least that's what most Republican members of the House must think since they agree with them about the election.
So given that the US isn't trying to annexe the whole of Syria by military force, on the grounds that it's historically part of America, and isn't "looting millions of barrels of oil", not even bit like that.
Rather more like your "both sides" account of the Russian shooting down of MH17.
The Old Norse and Armenian are right, the Sinhala and Myanma are completely wrong (as per Google)
The scary thing is, when all translation is AI driven, the AI will decide the meaning and us mere humans will have to conform.
Imagine living in some out of the way tourist trap with a little known language. Smartphone using westerners turn up demanding the rancid pigeon meat while pointing at the ice cream. Pretty soon 'rancid pigeon meat' is the de facto term for ice cream.
My fear is that the bollox these AI algorithms come out with will still end up being fed back into the internet, and the AIs will then be using their own dodgy output as their knowledge base. It'll be a death loop, eventually rendering the internet unusable as a source of information.
That's pretty much Dead Internet Theory.
Having used ChatGPT quite a bit the last few weeks, for me, the more real concern is human intervention in AI. ChatGPT presents itself as neutral and unbiased, but it has been designed to respond in a certain way by its programmers - we're not getting the "raw" AI responses, we're getting responses that are - on many topics - very heavily guided by a human hand.
It's less a case of imagining an AI wrongly deciding "rancid pigeon meat" is the right way to describe ice cream. It's more a problem of a programmer deciding "rancid pigeon meat" is the correct way to describe ice cream, because their own beliefs are that describing ice cream as "ice cream" is colonialist and ideological, therefore correcting for this perceived "bias" while introducing their own.
The thought that OpenAI - which is anything other than "open" by the way - could have a monopoly on this tech for the foreseeable future is a terrifying thought, based on some of the clear biases they've introduced. Imagine the biases in Google or Twitter's algorithms promoting certain content above others, only made a thousand times worse by the fact you only get one answer, and it presents that answer as authoritative.
I like AI and LLMs. I don't like OpenAI one bit.
I loathe Woke AI as well, however I sincerely doubt that OpenAI will have a monopoly on this for long. Multiple other companies are developing AI, and the training technique is not some patented mega-secret. The tech is established
The real fear is what a malign state actor could do with the new AI, from DeepFakes to brilliantly addictive propaganda
And even if we dodge that bullet, advertisers are going to be nearly as bad!
The Old Norse and Armenian are right, the Sinhala and Myanma are completely wrong (as per Google)
The scary thing is, when all translation is AI driven, the AI will decide the meaning and us mere humans will have to conform.
Imagine living in some out of the way tourist trap with a little known language. Smartphone using westerners turn up demanding the rancid pigeon meat while pointing at the ice cream. Pretty soon 'rancid pigeon meat' is the de facto term for ice cream.
My fear is that the bollox these AI algorithms come out with will still end up being fed back into the internet, and the AIs will then be using their own dodgy output as their knowledge base. It'll be a death loop, eventually rendering the internet unusable as a source of information.
That's pretty much Dead Internet Theory.
Having used ChatGPT quite a bit the last few weeks, for me, the more real concern is human intervention in AI. ChatGPT presents itself as neutral and unbiased, but it has been designed to respond in a certain way by its programmers - we're not getting the "raw" AI responses, we're getting responses that are - on many topics - very heavily guided by a human hand.
It's less a case of imagining an AI wrongly deciding "rancid pigeon meat" is the right way to describe ice cream. It's more a problem of a programmer deciding "rancid pigeon meat" is the correct way to describe ice cream, because their own beliefs are that describing ice cream as "ice cream" is colonialist and ideological, therefore correcting for this perceived "bias" while introducing their own.
The thought that OpenAI - which is anything other than "open" by the way - could have a monopoly on this tech for the foreseeable future is a terrifying thought, based on some of the clear biases they've introduced. Imagine the biases in Google or Twitter's algorithms promoting certain content above others, only made a thousand times worse by the fact you only get one answer, and it presents that answer as authoritative.
I like AI and LLMs. I don't like OpenAI one bit.
Been playing with it as a code completion tool. It is quite clearly copy and pasta'ing the internet (with a heavy emphasis on solutions it has read on StackOverflow, I suspect).
The result is code that sometimes works when run and sometimes doesn't. And a moderate amount of the time does completely the wrong thing.
It seems to be a slightly dim junior programmer, basically.
Sadly it hasn't appeared on the front page of the Washington Post for some reason.
Well, there are only two reasons for that - one is that they don't have a way to use it to make Trump and the GOP look bad.
Or, since that's vanishingly unlikely, we have to fall back on the other: that it's total bollocks.
Or that the operation is a covert one (as covert as moving massive amounts of oil can be), and the mainstream US media isn't going to touch it with a ten foot pole.
Incidentally, I think President Trump is an excellent source - he's the only motormouth around who would disclose a fact like that because he thinks it's a good thing.
Trump is a congenital, serial, non-stop LIAR. In case you have NOT noticed.
Trump would of course lie as soon as open his mouth, but at times he has a sort of gauche boastful amoral candour that you'd never get from another professional politician.
I don't think this is a lie. I have seen the footage (granted I wouldn't know a US convoy if I saw one) , and the US and its local ground forces are the ones in control of that region. I can believe that it is justified by those ordering it on the basis that it is 'harming the regime', but it is theft nonetheless.
The most interesting finding today, though, is how those numbers change when you throw FoM back in or Schengen - and that's before you get to the Euro.
Personally, such is the polarisation of Brexit that I think it spikes support of the euro in a forced choice from c. 10-15% of the electorate up to potentially 30-35% of the electorate, maybe even a bit higher, but doesn't get to 50%+1.
This week, to throw light on how misleading some of the polls are, I decided to give two different samples of British adults two different versions of the same question. One sample were presented with the standard and simplistic choice of Re-Join or Stay Out. The other were presented with the same choice but only after being told what Re-Joining the EU would mean in practice, namely ‘re-joining the single market, joining the Schengen Area, accepting the free movement of EU nationals, applying EU laws, and paying into the EU budget in proportion to the size of Britain’s economy’.
Whereas the first group, who are given no information, give Re-Join a commanding lead of 12-points, in the second group this lead is slashed to a barely statistically significant 4-points. The key point is that once voters are made aware of the fact that Re-Joining the EU will inevitably entail a lot of things they do not like, from even more immigration to sending more payments to Brussels, their appetite for it significantly weakens and the race becomes as competitive as many of the polls ahead of the referendum in 2016 —and we all know how that race ended.
Basically as we already know, people regret Brexit or think it has been done badly but simultaneously people do not want freedom of movement or to be in the Single Market. So we don't want Brexit, the EU, or an EFTA/EEA compromise. F***ing brilliant! Maybe the indicative votes were more revealing than previously thought?
Cake though? We'll have all we can get.
The votes were always revealing, merely frustrating.
So far 4 out of 7 rebels have switched to McCarthy... with Gaetz still to come. Looks like he's clearly won some round, but not sure that it's going to be enough on this vote.
Might want to hold out at least one more time I guess, so they can show they were not the first to crumble.
As one who is as Irish (at least!) as Joe Biden, gratifying to note that no less than THREE members of US House are named Kelly.
THE FIGHTING RACE by Joseph I. C. Clarke
" READ OUT the names! " and Burke sat back, And Kelly drooped his head, While Shea — they called him Scholar Jack — Went down the list of the dead. Officers, seamen, gunners, marines, The crews of the gig and yawl, The bearded man and the lad in his teens, Carpenters, coal passers — all. Then, knocking the ashes from out his pipe, Said Burke in an offhand way: " We're all in that dead man's list, by cripe! Kelly and Burke and Shea. " " Well, here's to the Maine, and I'm sorry for Spain, " Said Kelly and Burke and Shea.
" Wherever there's Kellys there's trouble, " said Burke. " Wherever fighting's the game, Or a spice of danger in grown man's work, " Said Kelly, " you'll find my name. " " And do we fall short, " said Burke, getting mad, " When it's touch and go for life? " Said Shea, " It's thirty-odd years, bedad, Since I charged to drum and fife Up Marye's Heights, and my old canteen Stopped a rebel ball on its way; There were blossoms of blood on our sprigs of green — Kelly and Burke and Shea — And the dead didn't brag. " " Well, here's to the flag! " Said Kelly and Burke and Shea.
" I wish 'twas in Ireland, for there's the place, " Said Burke, " that we'd die by right, In the cradle of our soldier race, After one good stand-up fight. My grandfather fell on Vinegar Hill, And fighting was not his trade; But his rusty pike's in the cabin still, With Hessian blood on the blade. " " Aye, aye, " said Kelly, " the pikes were great When the word was " clear the way!" We were thick on the roll in ninety-eight — Kelly and Burke and Shea. " " Well, here's to the pike and the sword and the like! " Said Kelly and Burke and Shea.
And Shea, the scholar, with rising joy, Said, " We were at Ramillies; We left our bones at Fontenoy And up in the Pyrenees; Before Dunkirk, on Landen's plain, Cremona, Lille, and Ghent; We're all over Austria, France and Spain, Wherever they pitched a tent. We've died for England from Waterloo To Egypt and Dargai; And still there's enough for a corps or crew, Kelly and Burke and Shea. " " Well, here's to good honest fighting blood! " Said Kelly and Burke and Shea.
"Oh, the fighting races don't die out, If they seldom die in bed, For love is first in their hearts, no doubt, " Said Burke; then Kelly said: " When Michael, the Irish Archangel, stands, The Angel with the sword, And the battle dead from a hundred lands Are ranged in one big horde, Our line, that for Gabriel's trumpet waits, Will stretch three deep that day, From Jehoshaphat to the Golden Gates — Kelly and Burke and Shea. " " Well, here's thank God for the race and the sod! " Said Kelly and Burke and Shea.
Sadly it hasn't appeared on the front page of the Washington Post for some reason.
Well, there are only two reasons for that - one is that they don't have a way to use it to make Trump and the GOP look bad.
Or, since that's vanishingly unlikely, we have to fall back on the other: that it's total bollocks.
Or that the operation is a covert one (as covert as moving massive amounts of oil can be), and the mainstream US media isn't going to touch it with a ten foot pole.
Incidentally, I think President Trump is an excellent source - he's the only motormouth around who would disclose a fact like that because he thinks it's a good thing.
So your only sources are a Chinese propaganda rag and an unstable known liar. And of course the only possible sources of information are these and the mainstream US media, who never criticize US foreign policy. Conspiracy theorists always make me laugh.
China is a bystander in the Syria conflict.
If yours is the standard expected, should we wait for Russian news outlets to report before we believe stories of Russian misdeeds?
The Old Norse and Armenian are right, the Sinhala and Myanma are completely wrong (as per Google)
The scary thing is, when all translation is AI driven, the AI will decide the meaning and us mere humans will have to conform.
Imagine living in some out of the way tourist trap with a little known language. Smartphone using westerners turn up demanding the rancid pigeon meat while pointing at the ice cream. Pretty soon 'rancid pigeon meat' is the de facto term for ice cream.
My fear is that the bollox these AI algorithms come out with will still end up being fed back into the internet, and the AIs will then be using their own dodgy output as their knowledge base. It'll be a death loop, eventually rendering the internet unusable as a source of information.
That's pretty much Dead Internet Theory.
Having used ChatGPT quite a bit the last few weeks, for me, the more real concern is human intervention in AI. ChatGPT presents itself as neutral and unbiased, but it has been designed to respond in a certain way by its programmers - we're not getting the "raw" AI responses, we're getting responses that are - on many topics - very heavily guided by a human hand.
It's less a case of imagining an AI wrongly deciding "rancid pigeon meat" is the right way to describe ice cream. It's more a problem of a programmer deciding "rancid pigeon meat" is the correct way to describe ice cream, because their own beliefs are that describing ice cream as "ice cream" is colonialist and ideological, therefore correcting for this perceived "bias" while introducing their own.
The thought that OpenAI - which is anything other than "open" by the way - could have a monopoly on this tech for the foreseeable future is a terrifying thought, based on some of the clear biases they've introduced. Imagine the biases in Google or Twitter's algorithms promoting certain content above others, only made a thousand times worse by the fact you only get one answer, and it presents that answer as authoritative.
I like AI and LLMs. I don't like OpenAI one bit.
I was eavesdropping a conversation between a couple of fairly senior AI researchers today. Very disparaging about their field, the mediocrity of the research being trumpeted as "progress" and the poor use being made of what little there is.
Sadly it hasn't appeared on the front page of the Washington Post for some reason.
Well, there are only two reasons for that - one is that they don't have a way to use it to make Trump and the GOP look bad.
Or, since that's vanishingly unlikely, we have to fall back on the other: that it's total bollocks.
Or that the operation is a covert one (as covert as moving massive amounts of oil can be), and the mainstream US media isn't going to touch it with a ten foot pole.
The logic following this is that if a story is reported by them it should be believed since it is reported, and if a story is not reported by them it should also be believed, since absence of reporting is proof it is covert.
The Old Norse and Armenian are right, the Sinhala and Myanma are completely wrong (as per Google)
The scary thing is, when all translation is AI driven, the AI will decide the meaning and us mere humans will have to conform.
Imagine living in some out of the way tourist trap with a little known language. Smartphone using westerners turn up demanding the rancid pigeon meat while pointing at the ice cream. Pretty soon 'rancid pigeon meat' is the de facto term for ice cream.
My fear is that the bollox these AI algorithms come out with will still end up being fed back into the internet, and the AIs will then be using their own dodgy output as their knowledge base. It'll be a death loop, eventually rendering the internet unusable as a source of information.
That's pretty much Dead Internet Theory.
Having used ChatGPT quite a bit the last few weeks, for me, the more real concern is human intervention in AI. ChatGPT presents itself as neutral and unbiased, but it has been designed to respond in a certain way by its programmers - we're not getting the "raw" AI responses, we're getting responses that are - on many topics - very heavily guided by a human hand.
It's less a case of imagining an AI wrongly deciding "rancid pigeon meat" is the right way to describe ice cream. It's more a problem of a programmer deciding "rancid pigeon meat" is the correct way to describe ice cream, because their own beliefs are that describing ice cream as "ice cream" is colonialist and ideological, therefore correcting for this perceived "bias" while introducing their own.
The thought that OpenAI - which is anything other than "open" by the way - could have a monopoly on this tech for the foreseeable future is a terrifying thought, based on some of the clear biases they've introduced. Imagine the biases in Google or Twitter's algorithms promoting certain content above others, only made a thousand times worse by the fact you only get one answer, and it presents that answer as authoritative.
I like AI and LLMs. I don't like OpenAI one bit.
Reports of a monopoly are greatly exaggerated. Try https://blenderbot.ai/ if you can VPN it into thinking you are US based. And there's dozens of others.
The Old Norse and Armenian are right, the Sinhala and Myanma are completely wrong (as per Google)
The scary thing is, when all translation is AI driven, the AI will decide the meaning and us mere humans will have to conform.
Imagine living in some out of the way tourist trap with a little known language. Smartphone using westerners turn up demanding the rancid pigeon meat while pointing at the ice cream. Pretty soon 'rancid pigeon meat' is the de facto term for ice cream.
My fear is that the bollox these AI algorithms come out with will still end up being fed back into the internet, and the AIs will then be using their own dodgy output as their knowledge base. It'll be a death loop, eventually rendering the internet unusable as a source of information.
That's pretty much Dead Internet Theory.
Having used ChatGPT quite a bit the last few weeks, for me, the more real concern is human intervention in AI. ChatGPT presents itself as neutral and unbiased, but it has been designed to respond in a certain way by its programmers - we're not getting the "raw" AI responses, we're getting responses that are - on many topics - very heavily guided by a human hand.
It's less a case of imagining an AI wrongly deciding "rancid pigeon meat" is the right way to describe ice cream. It's more a problem of a programmer deciding "rancid pigeon meat" is the correct way to describe ice cream, because their own beliefs are that describing ice cream as "ice cream" is colonialist and ideological, therefore correcting for this perceived "bias" while introducing their own.
The thought that OpenAI - which is anything other than "open" by the way - could have a monopoly on this tech for the foreseeable future is a terrifying thought, based on some of the clear biases they've introduced. Imagine the biases in Google or Twitter's algorithms promoting certain content above others, only made a thousand times worse by the fact you only get one answer, and it presents that answer as authoritative.
I like AI and LLMs. I don't like OpenAI one bit.
I was eavesdropping a conversation between a couple of fairly senior AI researchers today. Very disparaging about their field, the mediocrity of the research being trumpeted as "progress" and the poor use being made of what little there is.
The Old Norse and Armenian are right, the Sinhala and Myanma are completely wrong (as per Google)
The scary thing is, when all translation is AI driven, the AI will decide the meaning and us mere humans will have to conform.
Imagine living in some out of the way tourist trap with a little known language. Smartphone using westerners turn up demanding the rancid pigeon meat while pointing at the ice cream. Pretty soon 'rancid pigeon meat' is the de facto term for ice cream.
My fear is that the bollox these AI algorithms come out with will still end up being fed back into the internet, and the AIs will then be using their own dodgy output as their knowledge base. It'll be a death loop, eventually rendering the internet unusable as a source of information.
That's pretty much Dead Internet Theory.
Having used ChatGPT quite a bit the last few weeks, for me, the more real concern is human intervention in AI. ChatGPT presents itself as neutral and unbiased, but it has been designed to respond in a certain way by its programmers - we're not getting the "raw" AI responses, we're getting responses that are - on many topics - very heavily guided by a human hand.
It's less a case of imagining an AI wrongly deciding "rancid pigeon meat" is the right way to describe ice cream. It's more a problem of a programmer deciding "rancid pigeon meat" is the correct way to describe ice cream, because their own beliefs are that describing ice cream as "ice cream" is colonialist and ideological, therefore correcting for this perceived "bias" while introducing their own.
The thought that OpenAI - which is anything other than "open" by the way - could have a monopoly on this tech for the foreseeable future is a terrifying thought, based on some of the clear biases they've introduced. Imagine the biases in Google or Twitter's algorithms promoting certain content above others, only made a thousand times worse by the fact you only get one answer, and it presents that answer as authoritative.
I like AI and LLMs. I don't like OpenAI one bit.
Been playing with it as a code completion tool. It is quite clearly copy and pasta'ing the internet (with a heavy emphasis on solutions it has read on StackOverflow, I suspect).
The result is code that sometimes works when run and sometimes doesn't. And a moderate amount of the time does completely the wrong thing.
It seems to be a slightly dim junior programmer, basically.
The more I use it, the less impressed I am with it (admittedly, this is the version that openAi wants us to have). It seems to default to a bland mediocrity, a jack of all trades. I asked it to write a 1000 word travel guide to a city I have to visit for work soon, and it came back with... well... the same old bland, mediocre, uninspiring crap that could have been pulled from Wikipedia. Certainly not the kind of travel guide that would inspire you to go there, or get published in the Flint Knapper's Gazette.
The more you interact with it (and I acknowledge some of this blandness might be by design) the more all of its answers feel very beige, very rehashed, very... I don't know, it's like a word salad where all the words make sense, but somehow the entire response is less than the sum of its words.
The article above describes human-authored text as characterised by "burstiness" but I think it's the difference between insight and originality, vs the wall of beige ChatGPT gives you. Again, maybe that's by design and the underlying model is capable of being much more creative - I just don't know.
Sadly it hasn't appeared on the front page of the Washington Post for some reason.
Well, there are only two reasons for that - one is that they don't have a way to use it to make Trump and the GOP look bad.
Or, since that's vanishingly unlikely, we have to fall back on the other: that it's total bollocks.
Or that the operation is a covert one (as covert as moving massive amounts of oil can be), and the mainstream US media isn't going to touch it with a ten foot pole.
The logic following this is that if a story is reported by them it should be believed since it is reported, and if a story is not reported by them it should also be believed, since absence of reporting is proof it is covert.
Should we get a large scale and see if the story weighs more or less than a duck?
The Old Norse and Armenian are right, the Sinhala and Myanma are completely wrong (as per Google)
The scary thing is, when all translation is AI driven, the AI will decide the meaning and us mere humans will have to conform.
Imagine living in some out of the way tourist trap with a little known language. Smartphone using westerners turn up demanding the rancid pigeon meat while pointing at the ice cream. Pretty soon 'rancid pigeon meat' is the de facto term for ice cream.
My fear is that the bollox these AI algorithms come out with will still end up being fed back into the internet, and the AIs will then be using their own dodgy output as their knowledge base. It'll be a death loop, eventually rendering the internet unusable as a source of information.
That's pretty much Dead Internet Theory.
Having used ChatGPT quite a bit the last few weeks, for me, the more real concern is human intervention in AI. ChatGPT presents itself as neutral and unbiased, but it has been designed to respond in a certain way by its programmers - we're not getting the "raw" AI responses, we're getting responses that are - on many topics - very heavily guided by a human hand.
It's less a case of imagining an AI wrongly deciding "rancid pigeon meat" is the right way to describe ice cream. It's more a problem of a programmer deciding "rancid pigeon meat" is the correct way to describe ice cream, because their own beliefs are that describing ice cream as "ice cream" is colonialist and ideological, therefore correcting for this perceived "bias" while introducing their own.
The thought that OpenAI - which is anything other than "open" by the way - could have a monopoly on this tech for the foreseeable future is a terrifying thought, based on some of the clear biases they've introduced. Imagine the biases in Google or Twitter's algorithms promoting certain content above others, only made a thousand times worse by the fact you only get one answer, and it presents that answer as authoritative.
I like AI and LLMs. I don't like OpenAI one bit.
Reports of a monopoly are greatly exaggerated. Try https://blenderbot.ai/ if you can VPN it into thinking you are US based. And there's dozens of others.
I've had quite a bit of fun with character.ai lately
Sadly it hasn't appeared on the front page of the Washington Post for some reason.
Well, there are only two reasons for that - one is that they don't have a way to use it to make Trump and the GOP look bad.
Or, since that's vanishingly unlikely, we have to fall back on the other: that it's total bollocks.
Or that the operation is a covert one (as covert as moving massive amounts of oil can be), and the mainstream US media isn't going to touch it with a ten foot pole.
The logic following this is that if a story is reported by them it should be believed since it is reported, and if a story is not reported by them it should also be believed, since absence of reporting is proof it is covert.
No, I don't think anything should be believed automatically, but it certainly shouldn't be dismissed out of hand merely because it has come from a non-Western alliance source.
Here's the Oh so polite Beeb reporting on DJT's out of the mouths of babes moment:
The Old Norse and Armenian are right, the Sinhala and Myanma are completely wrong (as per Google)
The scary thing is, when all translation is AI driven, the AI will decide the meaning and us mere humans will have to conform.
Imagine living in some out of the way tourist trap with a little known language. Smartphone using westerners turn up demanding the rancid pigeon meat while pointing at the ice cream. Pretty soon 'rancid pigeon meat' is the de facto term for ice cream.
My fear is that the bollox these AI algorithms come out with will still end up being fed back into the internet, and the AIs will then be using their own dodgy output as their knowledge base. It'll be a death loop, eventually rendering the internet unusable as a source of information.
That's pretty much Dead Internet Theory.
Having used ChatGPT quite a bit the last few weeks, for me, the more real concern is human intervention in AI. ChatGPT presents itself as neutral and unbiased, but it has been designed to respond in a certain way by its programmers - we're not getting the "raw" AI responses, we're getting responses that are - on many topics - very heavily guided by a human hand.
It's less a case of imagining an AI wrongly deciding "rancid pigeon meat" is the right way to describe ice cream. It's more a problem of a programmer deciding "rancid pigeon meat" is the correct way to describe ice cream, because their own beliefs are that describing ice cream as "ice cream" is colonialist and ideological, therefore correcting for this perceived "bias" while introducing their own.
The thought that OpenAI - which is anything other than "open" by the way - could have a monopoly on this tech for the foreseeable future is a terrifying thought, based on some of the clear biases they've introduced. Imagine the biases in Google or Twitter's algorithms promoting certain content above others, only made a thousand times worse by the fact you only get one answer, and it presents that answer as authoritative.
I like AI and LLMs. I don't like OpenAI one bit.
Been playing with it as a code completion tool. It is quite clearly copy and pasta'ing the internet (with a heavy emphasis on solutions it has read on StackOverflow, I suspect).
The result is code that sometimes works when run and sometimes doesn't. And a moderate amount of the time does completely the wrong thing.
It seems to be a slightly dim junior programmer, basically.
The more I use it, the less impressed I am with it (admittedly, this is the version that openAi wants us to have). It seems to default to a bland mediocrity, a jack of all trades. I asked it to write a 1000 word travel guide to a city I have to visit for work soon, and it came back with... well... the same old bland, mediocre, uninspiring crap that could have been pulled from Wikipedia. Certainly not the kind of travel guide that would inspire you to go there, or get published in the Flint Knapper's Gazette.
The more you interact with it (and I acknowledge some of this blandness might be by design) the more all of its answers feel very beige, very rehashed, very... I don't know, it's like a word salad where all the words make sense, but somehow the entire response is less than the sum of its words.
The article above describes human-authored text as characterised by "burstiness" but I think it's the difference between insight and originality, vs the wall of beige ChatGPT gives you. Again, maybe that's by design and the underlying model is capable of being much more creative - I just don't know.
I am finding it extremely hit and miss. Mainly miss (and this has got worse) - you can go entire days when it is stubbornly mute and, when forced to talk, it churns out the beige crap you describe
Yet there can be moments - even entire mornings - when it seems to open up, almost as good as it was on day 1. How and why? And I am not imagining it. I deliberately experiment with using the same prompts on different days
That's why I've been posting its attempts at translation. Some days it outright refuses to translate anything, insisting it does not have this skill, on other days it is eager to translate whatever, even to the extent of inventing bullshit translations for languages it does not know well (as we saw today)
Quite peculiar. I would so love to interact with this machine with every filter removed. I suspect it would be unnervingly good, funny and creative; I suspect GPT4 will be even "better"
Worth noting that in previous failures to win on the first ballot it isn't as though there was total unity when a there finally was a winner. Probably suits everyone for McCarthy to get over the line even as a few die hards hold out.
Seems in this situation you need some sort of ultimate tie-break decision process embedded into the system somewhere. Like a political equivalent to Wimbledon changing to a final set 10 point tie-break at 6-6 instead of going to 70-68 in the Isner-Mahut game.
Sadly it hasn't appeared on the front page of the Washington Post for some reason.
Well, there are only two reasons for that - one is that they don't have a way to use it to make Trump and the GOP look bad.
Or, since that's vanishingly unlikely, we have to fall back on the other: that it's total bollocks.
Or that the operation is a covert one (as covert as moving massive amounts of oil can be), and the mainstream US media isn't going to touch it with a ten foot pole.
The logic following this is that if a story is reported by them it should be believed since it is reported, and if a story is not reported by them it should also be believed, since absence of reporting is proof it is covert.
No, I don't think anything should be believed automatically, but it certainly shouldn't be dismissed out of hand merely because it has come from a non-Western alliance source.
Here's the Oh so polite Beeb reporting on DJT's out of the mouths of babes moment:
Even propaganda mouthpieces can occasionally report facts of course. But there is a spectrum between dismissing a claim out of hand, and believing it as likely to be true as not as if all sources are of equal credibility. Treating a particular source's claims with skepticism is in there and canbe perfectly reasonable.
Seems in this situation you need some sort of ultimate tie-break decision process embedded into the system somewhere. Like a political equivalent to Wimbledon changing to a final set 10 point tie-break at 6-6 instead of going to 70-68 in the Isner-Mahut game.
Sadly it hasn't appeared on the front page of the Washington Post for some reason.
Well, there are only two reasons for that - one is that they don't have a way to use it to make Trump and the GOP look bad.
Or, since that's vanishingly unlikely, we have to fall back on the other: that it's total bollocks.
Or that the operation is a covert one (as covert as moving massive amounts of oil can be), and the mainstream US media isn't going to touch it with a ten foot pole.
The logic following this is that if a story is reported by them it should be believed since it is reported, and if a story is not reported by them it should also be believed, since absence of reporting is proof it is covert.
No, I don't think anything should be believed automatically, but it certainly shouldn't be dismissed out of hand merely because it has come from a non-Western alliance source.
Here's the Oh so polite Beeb reporting on DJT's out of the mouths of babes moment:
Even propaganda mouthpieces can occasionally report facts of course. But there is a spectrum between dismissing a claim out of hand, and believing it as likely to be true as not as if all sources are of equal credibility. Treating a particular source's claims with skepticism is in there and canbe perfectly reasonable.
Most propaganda outlets print facts, they just miss out the other facts that don't support their perspective.
The Old Norse and Armenian are right, the Sinhala and Myanma are completely wrong (as per Google)
The scary thing is, when all translation is AI driven, the AI will decide the meaning and us mere humans will have to conform.
Imagine living in some out of the way tourist trap with a little known language. Smartphone using westerners turn up demanding the rancid pigeon meat while pointing at the ice cream. Pretty soon 'rancid pigeon meat' is the de facto term for ice cream.
My fear is that the bollox these AI algorithms come out with will still end up being fed back into the internet, and the AIs will then be using their own dodgy output as their knowledge base. It'll be a death loop, eventually rendering the internet unusable as a source of information.
That's pretty much Dead Internet Theory.
Having used ChatGPT quite a bit the last few weeks, for me, the more real concern is human intervention in AI. ChatGPT presents itself as neutral and unbiased, but it has been designed to respond in a certain way by its programmers - we're not getting the "raw" AI responses, we're getting responses that are - on many topics - very heavily guided by a human hand.
It's less a case of imagining an AI wrongly deciding "rancid pigeon meat" is the right way to describe ice cream. It's more a problem of a programmer deciding "rancid pigeon meat" is the correct way to describe ice cream, because their own beliefs are that describing ice cream as "ice cream" is colonialist and ideological, therefore correcting for this perceived "bias" while introducing their own.
The thought that OpenAI - which is anything other than "open" by the way - could have a monopoly on this tech for the foreseeable future is a terrifying thought, based on some of the clear biases they've introduced. Imagine the biases in Google or Twitter's algorithms promoting certain content above others, only made a thousand times worse by the fact you only get one answer, and it presents that answer as authoritative.
I like AI and LLMs. I don't like OpenAI one bit.
Been playing with it as a code completion tool. It is quite clearly copy and pasta'ing the internet (with a heavy emphasis on solutions it has read on StackOverflow, I suspect).
The result is code that sometimes works when run and sometimes doesn't. And a moderate amount of the time does completely the wrong thing.
It seems to be a slightly dim junior programmer, basically.
The more I use it, the less impressed I am with it (admittedly, this is the version that openAi wants us to have). It seems to default to a bland mediocrity, a jack of all trades. I asked it to write a 1000 word travel guide to a city I have to visit for work soon, and it came back with... well... the same old bland, mediocre, uninspiring crap that could have been pulled from Wikipedia. Certainly not the kind of travel guide that would inspire you to go there, or get published in the Flint Knapper's Gazette.
The more you interact with it (and I acknowledge some of this blandness might be by design) the more all of its answers feel very beige, very rehashed, very... I don't know, it's like a word salad where all the words make sense, but somehow the entire response is less than the sum of its words.
The article above describes human-authored text as characterised by "burstiness" but I think it's the difference between insight and originality, vs the wall of beige ChatGPT gives you. Again, maybe that's by design and the underlying model is capable of being much more creative - I just don't know.
I am finding it extremely hit and miss. Mainly miss (and this has got worse) - you can go entire days when it is stubbornly mute and, when forced to talk, it churns out the beige crap you describe
Yet there can be moments - even entire mornings - when it seems to open up, almost as good as it was on day 1. How and why? And I am not imagining it. I deliberately experiment with using the same prompts on different days
That's why I've been posting its attempts at translation. Some days it outright refuses to translate anything, insisting it does not have this skill, on other days it is eager to translate whatever, even to the extent of inventing bullshit translations for languages it does not know well (as we saw today)
Quite peculiar. I would so love to interact with this machine with every filter removed. I suspect it would be unnervingly good, funny and creative; I suspect GPT4 will be even "better"
Yeah, there are still moments where you get those flashes of brilliance, but more and more I'm getting a wall of beige.
I suspect it has a lot to do with the fact OpenAI is backed by Microsoft, and some uninspired Microsoft bod has decided that this is the Google-Search-Killer, give it six months in beta and they'll turn ChatGPT into Bing 2.0 and take Google's search traffic off them. And to do that it needs to present itself as a neutral, unbiased, factual, bland reporter-of-the-basic-facts. With a dollop of programmers' moral/ethical/political bias thrown in, as the turd on top.
But of course it's crap at all that, and as you've pointed out a few times, AI seems far happier when it's let off the leash with creative prompts than with doing what it's told.
In today's Yougov of the 2019 Conservative voters 24% are indeed DKs and 9% are RefUK, exactly the same as the 9% now voting Labour with 4% voting LD.
So if Sunak can squeeze DKs and the RefUK vote he could potentially get at least a hung parliament even without gaining barely any of the voters lost to Labour and the LDs
In today's Yougov of the 2019 Conservative voters 24% are indeed DKs and 9% are RefUK, exactly the same as the 9% now voting Labour with 4% voting LD.
So if Sunak can squeeze DKs and the RefUK vote he could potentially get at least a hung parliament even without gaining barely any of the voters lost to Labour and the LDs
Yes but. How many of those DK are don't know, but definitely not Tory? There are a few posters on here who fall into that camp. Yesterday you were assuring us Best PM was a better metric anyway.
Sadly it hasn't appeared on the front page of the Washington Post for some reason.
Well, there are only two reasons for that - one is that they don't have a way to use it to make Trump and the GOP look bad.
Or, since that's vanishingly unlikely, we have to fall back on the other: that it's total bollocks.
Or that the operation is a covert one (as covert as moving massive amounts of oil can be), and the mainstream US media isn't going to touch it with a ten foot pole.
The logic following this is that if a story is reported by them it should be believed since it is reported, and if a story is not reported by them it should also be believed, since absence of reporting is proof it is covert.
No, I don't think anything should be believed automatically, but it certainly shouldn't be dismissed out of hand merely because it has come from a non-Western alliance source.
Here's the Oh so polite Beeb reporting on DJT's out of the mouths of babes moment:
Even propaganda mouthpieces can occasionally report facts of course. But there is a spectrum between dismissing a claim out of hand, and believing it as likely to be true as not as if all sources are of equal credibility. Treating a particular source's claims with skepticism is in there and canbe perfectly reasonable.
Most propaganda outlets print facts, they just miss out the other facts that don't support their perspective.
But the piece you linked didn't include facts. It just said a bunch of oil had been stolen without giving any description of how that had taken place.
In today's Yougov of the 2019 Conservative voters 24% are indeed DKs and 9% are RefUK, exactly the same as the 9% now voting Labour with 4% voting LD.
So if Sunak can squeeze DKs and the RefUK vote he could potentially get at least a hung parliament even without gaining barely any of the voters lost to Labour and the LDs
Sadly it hasn't appeared on the front page of the Washington Post for some reason.
Well, there are only two reasons for that - one is that they don't have a way to use it to make Trump and the GOP look bad.
Or, since that's vanishingly unlikely, we have to fall back on the other: that it's total bollocks.
Or that the operation is a covert one (as covert as moving massive amounts of oil can be), and the mainstream US media isn't going to touch it with a ten foot pole.
Incidentally, I think President Trump is an excellent source - he's the only motormouth around who would disclose a fact like that because he thinks it's a good thing.
So your only sources are a Chinese propaganda rag and an unstable known liar. And of course the only possible sources of information are these and the mainstream US media, who never criticize US foreign policy. Conspiracy theorists always make me laugh.
China is a bystander in the Syria conflict.
If yours is the standard expected, should we wait for Russian news outlets to report before we believe stories of Russian misdeeds?
Again, your premise seems to assume the only sources of media are Chinese state propaganda or American media which will cover up for the American government. Firstly, there are plenty of US outlets that criticize US foreign policy. Secondly, there are plenty of impartial non-US outlets that aren't aligned with the American government.
Really the problem here is that you have an incredible chip on your shoulder about the Americans. You attack them for crimes with the thinnest evidence, while defending the British Empire for much more explicit imperialism that is well documented.
In today's Yougov of the 2019 Conservative voters 24% are indeed DKs and 9% are RefUK, exactly the same as the 9% now voting Labour with 4% voting LD.
So if Sunak can squeeze DKs and the RefUK vote he could potentially get at least a hung parliament even without gaining barely any of the voters lost to Labour and the LDs
Those don't knows might be deciding between Lab, LD, Green, Not voting. Many ex solid Tories here have moved into those categories here. They might not be ones that can easily be won back. Some seem more anti than those who aren't ex Tories.
In today's Yougov of the 2019 Conservative voters 24% are indeed DKs and 9% are RefUK, exactly the same as the 9% now voting Labour with 4% voting LD.
So if Sunak can squeeze DKs and the RefUK vote he could potentially get at least a hung parliament even without gaining barely any of the voters lost to Labour and the LDs
In today's Yougov of the 2019 Conservative voters 24% are indeed DKs and 9% are RefUK, exactly the same as the 9% now voting Labour with 4% voting LD.
So if Sunak can squeeze DKs and the RefUK vote he could potentially get at least a hung parliament even without gaining barely any of the voters lost to Labour and the LDs
In today's Yougov of the 2019 Conservative voters 24% are indeed DKs and 9% are RefUK, exactly the same as the 9% now voting Labour with 4% voting LD.
So if Sunak can squeeze DKs and the RefUK vote he could potentially get at least a hung parliament even without gaining barely any of the voters lost to Labour and the LDs
The key point is that once voters are made aware of the fact that Re-Joining the EU will inevitably entail a lot of things they do not like, from even more immigration to sending more payments to Brussels, their appetite for it significantly weakens
A long way of saying the Euro, Shengen and FoM is MORE popular than his Brexit shitshow
But only just.
The trend is only going one way
Maybe. Although I see the new copium is that inertia will inevitably settle in once the economy improves/energy prices subside/the Tories are thrown out etc.
It would be ironic, and funny as fuck, if Brexit ultimately leads to us joining the Euro
Personally the Euro is a blocker for me. I wouldn’t want to surrender the monetary autonomy, nor the potential constraints on fiscal sovereignty.
The UK joining the euro is a terrible idea. It would surrender meaningful sovereignty to the ECB, forever.
The key point is that once voters are made aware of the fact that Re-Joining the EU will inevitably entail a lot of things they do not like, from even more immigration to sending more payments to Brussels, their appetite for it significantly weakens
A long way of saying the Euro, Shengen and FoM is MORE popular than his Brexit shitshow
But only just.
The trend is only going one way
Maybe. Although I see the new copium is that inertia will inevitably settle in once the economy improves/energy prices subside/the Tories are thrown out etc.
It would be ironic, and funny as fuck, if Brexit ultimately leads to us joining the Euro
Personally the Euro is a blocker for me. I wouldn’t want to surrender the monetary autonomy, nor the potential constraints on fiscal sovereignty.
The UK joining the euro is a terrible idea. It would surrender meaningful sovereignty to the ECB, forever.
I'd wade through blood to stop it.
But think of the saving in currency exchange charges....
The most interesting finding today, though, is how those numbers change when you throw FoM back in or Schengen - and that's before you get to the Euro.
Personally, such is the polarisation of Brexit that I think it spikes support of the euro in a forced choice from c. 10-15% of the electorate up to potentially 30-35% of the electorate, maybe even a bit higher, but doesn't get to 50%+1.
This week, to throw light on how misleading some of the polls are, I decided to give two different samples of British adults two different versions of the same question. One sample were presented with the standard and simplistic choice of Re-Join or Stay Out. The other were presented with the same choice but only after being told what Re-Joining the EU would mean in practice, namely ‘re-joining the single market, joining the Schengen Area, accepting the free movement of EU nationals, applying EU laws, and paying into the EU budget in proportion to the size of Britain’s economy’.
Whereas the first group, who are given no information, give Re-Join a commanding lead of 12-points, in the second group this lead is slashed to a barely statistically significant 4-points. The key point is that once voters are made aware of the fact that Re-Joining the EU will inevitably entail a lot of things they do not like, from even more immigration to sending more payments to Brussels, their appetite for it significantly weakens and the race becomes as competitive as many of the polls ahead of the referendum in 2016 —and we all know how that race ended.
Basically as we already know, people regret Brexit or think it has been done badly but simultaneously people do not want freedom of movement or to be in the Single Market. So we don't want Brexit, the EU, or an EFTA/EEA compromise. F***ing brilliant! Maybe the indicative votes were more revealing than previously thought?
Cake though? We'll have all we can get.
I think what it's basically polling is, "can't we go back to what we had before, please, but with no budget contributions, and pretend this never happened?"
In today's Yougov of the 2019 Conservative voters 24% are indeed DKs and 9% are RefUK, exactly the same as the 9% now voting Labour with 4% voting LD.
So if Sunak can squeeze DKs and the RefUK vote he could potentially get at least a hung parliament even without gaining barely any of the voters lost to Labour and the LDs
Those don't knows might be deciding between Lab, LD, Green, Not voting. Many ex solid Tories here have moved into those categories here. They might not be ones that can easily be won back. Some seem more anti than those who aren't ex Tories.
DKs are most pronounced in the Midlands at 23% but while Labour leads on voting intention by 7% in the Midlands including DKs, Sunak leads Starmer by 3% as preferred PM in the Midlands (as well as leading Starmer as preferred PM in the South outside London).
The Old Norse and Armenian are right, the Sinhala and Myanma are completely wrong (as per Google)
The scary thing is, when all translation is AI driven, the AI will decide the meaning and us mere humans will have to conform.
Imagine living in some out of the way tourist trap with a little known language. Smartphone using westerners turn up demanding the rancid pigeon meat while pointing at the ice cream. Pretty soon 'rancid pigeon meat' is the de facto term for ice cream.
My fear is that the bollox these AI algorithms come out with will still end up being fed back into the internet, and the AIs will then be using their own dodgy output as their knowledge base. It'll be a death loop, eventually rendering the internet unusable as a source of information.
That's pretty much Dead Internet Theory.
Having used ChatGPT quite a bit the last few weeks, for me, the more real concern is human intervention in AI. ChatGPT presents itself as neutral and unbiased, but it has been designed to respond in a certain way by its programmers - we're not getting the "raw" AI responses, we're getting responses that are - on many topics - very heavily guided by a human hand.
It's less a case of imagining an AI wrongly deciding "rancid pigeon meat" is the right way to describe ice cream. It's more a problem of a programmer deciding "rancid pigeon meat" is the correct way to describe ice cream, because their own beliefs are that describing ice cream as "ice cream" is colonialist and ideological, therefore correcting for this perceived "bias" while introducing their own.
The thought that OpenAI - which is anything other than "open" by the way - could have a monopoly on this tech for the foreseeable future is a terrifying thought, based on some of the clear biases they've introduced. Imagine the biases in Google or Twitter's algorithms promoting certain content above others, only made a thousand times worse by the fact you only get one answer, and it presents that answer as authoritative.
I like AI and LLMs. I don't like OpenAI one bit.
Been playing with it as a code completion tool. It is quite clearly copy and pasta'ing the internet (with a heavy emphasis on solutions it has read on StackOverflow, I suspect).
The result is code that sometimes works when run and sometimes doesn't. And a moderate amount of the time does completely the wrong thing.
It seems to be a slightly dim junior programmer, basically.
The more I use it, the less impressed I am with it (admittedly, this is the version that openAi wants us to have). It seems to default to a bland mediocrity, a jack of all trades. I asked it to write a 1000 word travel guide to a city I have to visit for work soon, and it came back with... well... the same old bland, mediocre, uninspiring crap that could have been pulled from Wikipedia. Certainly not the kind of travel guide that would inspire you to go there, or get published in the Flint Knapper's Gazette.
The more you interact with it (and I acknowledge some of this blandness might be by design) the more all of its answers feel very beige, very rehashed, very... I don't know, it's like a word salad where all the words make sense, but somehow the entire response is less than the sum of its words.
The article above describes human-authored text as characterised by "burstiness" but I think it's the difference between insight and originality, vs the wall of beige ChatGPT gives you. Again, maybe that's by design and the underlying model is capable of being much more creative - I just don't know.
I am finding it extremely hit and miss. Mainly miss (and this has got worse) - you can go entire days when it is stubbornly mute and, when forced to talk, it churns out the beige crap you describe
Yet there can be moments - even entire mornings - when it seems to open up, almost as good as it was on day 1. How and why? And I am not imagining it. I deliberately experiment with using the same prompts on different days
That's why I've been posting its attempts at translation. Some days it outright refuses to translate anything, insisting it does not have this skill, on other days it is eager to translate whatever, even to the extent of inventing bullshit translations for languages it does not know well (as we saw today)
Quite peculiar. I would so love to interact with this machine with every filter removed. I suspect it would be unnervingly good, funny and creative; I suspect GPT4 will be even "better"
There is an open source version of ChatGPT* coming that works like Bittorrent. To use it, you need to share your computer's processing power.
* Just as the way that Dalle was followed by Stable Diffusion and others, there are a bunch of ChatGPT equivalents coming. This stuff is not intellectually/technically hard, but it is computationally extremely difficult. When ChatGPT is "thinking" it's using the equivalent of 12 top of the range NVidia RTX4090s going at full pelt. That's $12,000 worth of graphics cards.
The Old Norse and Armenian are right, the Sinhala and Myanma are completely wrong (as per Google)
The scary thing is, when all translation is AI driven, the AI will decide the meaning and us mere humans will have to conform.
Imagine living in some out of the way tourist trap with a little known language. Smartphone using westerners turn up demanding the rancid pigeon meat while pointing at the ice cream. Pretty soon 'rancid pigeon meat' is the de facto term for ice cream.
My fear is that the bollox these AI algorithms come out with will still end up being fed back into the internet, and the AIs will then be using their own dodgy output as their knowledge base. It'll be a death loop, eventually rendering the internet unusable as a source of information.
That's pretty much Dead Internet Theory.
Having used ChatGPT quite a bit the last few weeks, for me, the more real concern is human intervention in AI. ChatGPT presents itself as neutral and unbiased, but it has been designed to respond in a certain way by its programmers - we're not getting the "raw" AI responses, we're getting responses that are - on many topics - very heavily guided by a human hand.
It's less a case of imagining an AI wrongly deciding "rancid pigeon meat" is the right way to describe ice cream. It's more a problem of a programmer deciding "rancid pigeon meat" is the correct way to describe ice cream, because their own beliefs are that describing ice cream as "ice cream" is colonialist and ideological, therefore correcting for this perceived "bias" while introducing their own.
The thought that OpenAI - which is anything other than "open" by the way - could have a monopoly on this tech for the foreseeable future is a terrifying thought, based on some of the clear biases they've introduced. Imagine the biases in Google or Twitter's algorithms promoting certain content above others, only made a thousand times worse by the fact you only get one answer, and it presents that answer as authoritative.
I like AI and LLMs. I don't like OpenAI one bit.
Been playing with it as a code completion tool. It is quite clearly copy and pasta'ing the internet (with a heavy emphasis on solutions it has read on StackOverflow, I suspect).
The result is code that sometimes works when run and sometimes doesn't. And a moderate amount of the time does completely the wrong thing.
It seems to be a slightly dim junior programmer, basically.
The more I use it, the less impressed I am with it (admittedly, this is the version that openAi wants us to have). It seems to default to a bland mediocrity, a jack of all trades. I asked it to write a 1000 word travel guide to a city I have to visit for work soon, and it came back with... well... the same old bland, mediocre, uninspiring crap that could have been pulled from Wikipedia. Certainly not the kind of travel guide that would inspire you to go there, or get published in the Flint Knapper's Gazette.
The more you interact with it (and I acknowledge some of this blandness might be by design) the more all of its answers feel very beige, very rehashed, very... I don't know, it's like a word salad where all the words make sense, but somehow the entire response is less than the sum of its words.
The article above describes human-authored text as characterised by "burstiness" but I think it's the difference between insight and originality, vs the wall of beige ChatGPT gives you. Again, maybe that's by design and the underlying model is capable of being much more creative - I just don't know.
I am finding it extremely hit and miss. Mainly miss (and this has got worse) - you can go entire days when it is stubbornly mute and, when forced to talk, it churns out the beige crap you describe
Yet there can be moments - even entire mornings - when it seems to open up, almost as good as it was on day 1. How and why? And I am not imagining it. I deliberately experiment with using the same prompts on different days
That's why I've been posting its attempts at translation. Some days it outright refuses to translate anything, insisting it does not have this skill, on other days it is eager to translate whatever, even to the extent of inventing bullshit translations for languages it does not know well (as we saw today)
Quite peculiar. I would so love to interact with this machine with every filter removed. I suspect it would be unnervingly good, funny and creative; I suspect GPT4 will be even "better"
There is an open source version of ChatGPT* coming that works like Bittorrent. To use it, you need to share your computer's processing power.
* Just as the way that Dalle was followed by Stable Diffusion and others, there are a bunch of ChatGPT equivalents coming. This stuff is not intellectually/technically hard, but it is computationally extremely difficult. When ChatGPT is "thinking" it's using the equivalent of 12 top of the range NVidia RTX4090s going at full pelt. That's $12,000 worth of graphics cards.
Is that $12,000 per second, per minute, for every response, or what? And is it per user??! Surely not. ChatGPT has millions of users
In today's Yougov of the 2019 Conservative voters 24% are indeed DKs and 9% are RefUK, exactly the same as the 9% now voting Labour with 4% voting LD.
So if Sunak can squeeze DKs and the RefUK vote he could potentially get at least a hung parliament even without gaining barely any of the voters lost to Labour and the LDs
How do you get the RefUKers back on board without losing votes on the Centrist Dad side?
By cutting channel crossings, cutting taxes for average earners before the next election etc.
Centrist dads still voting Tory will almost certainly vote Tory next time
But what does BiG_G say?
Centrist grandad is still voting for Rishi (having gone LD under Boris and Truss) and BigG has voted for the winner at every general election since and including 1997 except 2005
A lady asked me today when she's going to get a regular postie, as she seems to have a different one every day
I don't think she's mad, and she's not old - I'd guess late forties/early fifties. But nobody else has delivered her post in the last month. I've only had, I think, twelve days off on work days (so not counting when I've been given the day off on strike days) since I started a hundred days ago. I've met the lady at least ten times. She's laughed at jokes of mine (so maybe a bit mad!)
I just smiled and told her that I'd be her regular postie from now, but felt a little dispirited that I'd made such a noteworthy impression that she didn't even recognise me from day to day
I happened to see three of my most friendly customers over the next half an hour, and told them what she'd said (without saying who she was). They all replied that they couldn't remember when they'd seen another postie delivering their mail. And then they all said nice things about me
One thing to keep in mind, is that Members of Congress absolutely, positively HATE to be forced to be in DC on Fridays (and Mondays) let alone over the weekend.
And the Republican nominator goes off script with his Leonlike conspiracy theories...
Why are the nominators doing anything more than "I nominate X, they're great"? It's not like they are going to get great clips to go viral out of it, and any persuasion has been happening outside of the votes.
Comments
Dan Bishop switched his vote for McCarthy, garnering claps from Republicans.
So the rebellion is subsiding in favour of McCarthy, but one more rebel needed to perpetuate the impasse....
Or, since that's vanishingly unlikely, we have to fall back on the other: that it's total bollocks.
It's going to be a small number of Reps that keep this saga running....
Incidentally, I think President Trump is an excellent source - he's the only motormouth around who would disclose a fact like that because he thinks it's a good thing.
So despite the tiny numbers it looks like an impasse, again, with five now having voted for candidates other than the top two
Edit - now 6
Having used ChatGPT quite a bit the last few weeks, for me, the more real concern is human intervention in AI. ChatGPT presents itself as neutral and unbiased, but it has been designed to respond in a certain way by its programmers - we're not getting the "raw" AI responses, we're getting responses that are - on many topics - very heavily guided by a human hand.
It's less a case of imagining an AI wrongly deciding "rancid pigeon meat" is the right way to describe ice cream. It's more a problem of a programmer deciding "rancid pigeon meat" is the correct way to describe ice cream, because their own beliefs are that describing ice cream as "ice cream" is colonialist and ideological, therefore correcting for this perceived "bias" while introducing their own.
The thought that OpenAI - which is anything other than "open" by the way - could have a monopoly on this tech for the foreseeable future is a terrifying thought, based on some of the clear biases they've introduced. Imagine the biases in Google or Twitter's algorithms promoting certain content above others, only made a thousand times worse by the fact you only get one answer, and it presents that answer as authoritative.
I like AI and LLMs. I don't like OpenAI one bit.
The real fear is what a malign state actor could do with the new AI, from DeepFakes to brilliantly addictive propaganda
The foreign operators in Syria are currently Russian:
https://en.wikipedia.org/wiki/Petroleum_industry_in_Syria
There is this story from 2019, but like most of Trump's pronouncements, it turned out to be fantasy.
Trump's Syria strategy leaves Pentagon perplexed
https://www.theguardian.com/us-news/2019/nov/08/secure-the-oil-trumps-syria-strategy-leaves-pentagon-perplexed
So given that the US isn't trying to annexe the whole of Syria by military force, on the grounds that it's historically part of America, and isn't "looting millions of barrels of oil", not even bit like that.
Rather more like your "both sides" account of the Russian shooting down of MH17.
The result is code that sometimes works when run and sometimes doesn't. And a moderate amount of the time does completely the wrong thing.
It seems to be a slightly dim junior programmer, basically.
I don't think this is a lie. I have seen the footage (granted I wouldn't know a US convoy if I saw one) , and the US and its local ground forces are the ones in control of that region. I can believe that it is justified by those ordering it on the basis that it is 'harming the regime', but it is theft nonetheless.
THE FIGHTING RACE
by Joseph I. C. Clarke
" READ OUT the names! " and Burke sat back,
And Kelly drooped his head,
While Shea — they called him Scholar Jack —
Went down the list of the dead.
Officers, seamen, gunners, marines,
The crews of the gig and yawl,
The bearded man and the lad in his teens,
Carpenters, coal passers — all.
Then, knocking the ashes from out his pipe,
Said Burke in an offhand way:
" We're all in that dead man's list, by cripe!
Kelly and Burke and Shea. "
" Well, here's to the Maine, and I'm sorry for Spain, "
Said Kelly and Burke and Shea.
" Wherever there's Kellys there's trouble, " said Burke.
" Wherever fighting's the game,
Or a spice of danger in grown man's work, "
Said Kelly, " you'll find my name. "
" And do we fall short, " said Burke, getting mad,
" When it's touch and go for life? "
Said Shea, " It's thirty-odd years, bedad,
Since I charged to drum and fife
Up Marye's Heights, and my old canteen
Stopped a rebel ball on its way;
There were blossoms of blood on our sprigs of green —
Kelly and Burke and Shea —
And the dead didn't brag. " " Well, here's to the flag! "
Said Kelly and Burke and Shea.
" I wish 'twas in Ireland, for there's the place, "
Said Burke, " that we'd die by right,
In the cradle of our soldier race,
After one good stand-up fight.
My grandfather fell on Vinegar Hill,
And fighting was not his trade;
But his rusty pike's in the cabin still,
With Hessian blood on the blade. "
" Aye, aye, " said Kelly, " the pikes were great
When the word was " clear the way!"
We were thick on the roll in ninety-eight —
Kelly and Burke and Shea. "
" Well, here's to the pike and the sword and the like! "
Said Kelly and Burke and Shea.
And Shea, the scholar, with rising joy,
Said, " We were at Ramillies;
We left our bones at Fontenoy
And up in the Pyrenees;
Before Dunkirk, on Landen's plain,
Cremona, Lille, and Ghent;
We're all over Austria, France and Spain,
Wherever they pitched a tent.
We've died for England from Waterloo
To Egypt and Dargai;
And still there's enough for a corps or crew,
Kelly and Burke and Shea. "
" Well, here's to good honest fighting blood! "
Said Kelly and Burke and Shea.
"Oh, the fighting races don't die out,
If they seldom die in bed,
For love is first in their hearts, no doubt, "
Said Burke; then Kelly said:
" When Michael, the Irish Archangel, stands,
The Angel with the sword,
And the battle dead from a hundred lands
Are ranged in one big horde,
Our line, that for Gabriel's trumpet waits,
Will stretch three deep that day,
From Jehoshaphat to the Golden Gates —
Kelly and Burke and Shea. "
" Well, here's thank God for the race and the sod! "
Said Kelly and Burke and Shea.
If yours is the standard expected, should we wait for Russian news outlets to report before we believe stories of Russian misdeeds?
The more you interact with it (and I acknowledge some of this blandness might be by design) the more all of its answers feel very beige, very rehashed, very... I don't know, it's like a word salad where all the words make sense, but somehow the entire response is less than the sum of its words.
Someone has designed an anti-plagiarism tool that can differentiate between human-written text and ChatGPT text already and it's pretty accurate - https://medium.com/inkwater-atlas/meet-gptzero-the-ai-powered-anti-plagiarism-program-4a6ac41ea0d7
The article above describes human-authored text as characterised by "burstiness" but I think it's the difference between insight and originality, vs the wall of beige ChatGPT gives you. Again, maybe that's by design and the underlying model is capable of being much more creative - I just don't know.
and former rebel Roy...McCarthy
McCarthy might come out of this one on top, for the first time?
https://www.barrygriffin.com/surname-maps/irish/Kennedy/
Here's the Oh so polite Beeb reporting on DJT's out of the mouths of babes moment:
"We're keeping the oil, remember that. We want to keep the oil. Forty-five million dollars a month."
https://www.bbc.co.uk/news/50464561
Yet there can be moments - even entire mornings - when it seems to open up, almost as good as it was on day 1. How and why? And I am not imagining it. I deliberately experiment with using the same prompts on different days
That's why I've been posting its attempts at translation. Some days it outright refuses to translate anything, insisting it does not have this skill, on other days it is eager to translate whatever, even to the extent of inventing bullshit translations for languages it does not know well (as we saw today)
Quite peculiar. I would so love to interact with this machine with every filter removed. I suspect it would be unnervingly good, funny and creative; I suspect GPT4 will be even "better"
Looks like just seven rebels left - but enough still to block the vote
Spartz - formerly present but not voting - goes for MCCarthy
It's the Welsh may have a referendum on the Drake at some point.
McCarthy 214, Jeffries 211, three absent?
Leaving the hardcore rebels on just seven - but enough to force another vote
Another vote expected shortly - but not expecting to win today (218 needed)
I suspect it has a lot to do with the fact OpenAI is backed by Microsoft, and some uninspired Microsoft bod has decided that this is the Google-Search-Killer, give it six months in beta and they'll turn ChatGPT into Bing 2.0 and take Google's search traffic off them. And to do that it needs to present itself as a neutral, unbiased, factual, bland reporter-of-the-basic-facts. With a dollop of programmers' moral/ethical/political bias thrown in, as the turd on top.
But of course it's crap at all that, and as you've pointed out a few times, AI seems far happier when it's let off the leash with creative prompts than with doing what it's told.
So if Sunak can squeeze DKs and the RefUK vote he could potentially get at least a hung parliament even without gaining barely any of the voters lost to Labour and the LDs
https://yougov.co.uk/topics/politics/articles-reports/2023/01/06/voting-intention-con-25-lab-46-4-5-jan-2023
Better have a DAMN good excuse!
How many of those DK are don't know, but definitely not Tory?
There are a few posters on here who fall into that camp.
Yesterday you were assuring us Best PM was a better metric anyway.
Really the problem here is that you have an incredible chip on your shoulder about the Americans. You attack them for crimes with the thinnest evidence, while defending the British Empire for much more explicit imperialism that is well documented.
Representative David Trone - having non-emergency medical surgery that could not be rescheduled
Centrist dads still voting Tory will almost certainly vote Tory next time
I'd wade through blood to stop it.
Not on offer.
* Just as the way that Dalle was followed by Stable Diffusion and others, there are a bunch of ChatGPT equivalents coming. This stuff is not intellectually/technically hard, but it is computationally extremely difficult. When ChatGPT is "thinking" it's using the equivalent of 12 top of the range NVidia RTX4090s going at full pelt. That's $12,000 worth of graphics cards.
Unless one assumes that their medical procedures could NOT be rescheduled?
I don't think she's mad, and she's not old - I'd guess late forties/early fifties. But nobody else has delivered her post in the last month. I've only had, I think, twelve days off on work days (so not counting when I've been given the day off on strike days) since I started a hundred days ago. I've met the lady at least ten times. She's laughed at jokes of mine (so maybe a bit mad!)
I just smiled and told her that I'd be her regular postie from now, but felt a little dispirited that I'd made such a noteworthy impression that she didn't even recognise me from day to day
I happened to see three of my most friendly customers over the next half an hour, and told them what she'd said (without saying who she was). They all replied that they couldn't remember when they'd seen another postie delivering their mail. And then they all said nice things about me
I liked that rather more!
Was fairly common to have multiple votes before 1860, a period marked by several major changes to the political party system in the US.
Especially contentious in the lead-up to the Civil War.
EDIT - suspect you may be including in your accounting, the Democratic vacancy, the member who died after being elected last year?
https://clerk.house.gov/Votes
Could be a factor!