On another but even more important topic, these two threads are interesting on how the longer-range precision artillery that Ukraine is now receiving from the West is going to allow them to badly disrupt Russian logistics:
My concern is that Ukraine aren't getting that many and can't these be easily knocked out. Talking from ignorance so would love to be corrected.
The Russians have managed to knock out some of the M777 howitzers, so they might eventually spot a HIMARS with a UAV and be able to drop an artillery shell on it. However they seem to be quite mobile, so if Ukraine are cautious they can pull them out of range during the day and then drive them up to firing positions during the night.
There have been some interesting videos of them reloading the things in fields, etc, which is suggestive of them being careful to distribute the reloads, so that Russia can't as easily catch one at an ammo dump.
I think Ukraine will gradually receive more of these over time.
One thing is utterly clear from this war - the days of lining up your artillery in a neat row, setting up a tea tent etc are gone. Shoot and scoot from individual positions (with high level coordination) is the only credible way forward.
A question is whether armouring artillery systems is worth it vs mobility.
There's another aspect to this that surprised me (it should not have): to accurately fire artillery, you need not only to know where the target is; you need to know where you are, and that means your position has to be surveyed. Hence all the piccies of squaddies looking through theodolites (I think - I doubt they are levels).
A lot of positions can be pre-surveyed and marked. Does military-grade GPS signals give enough accuracy for this purpose?
There was some brilliant footage last week of the Ukranians using a drone as a ‘spotter’ for long-range artillery, which was brilliantly innovative.
The MLRS really need to use each firing location once, and then move on quickly, keeping them well hidden during the day. The Russians will quickly work backwards to get the firing locations.
On another but even more important topic, these two threads are interesting on how the longer-range precision artillery that Ukraine is now receiving from the West is going to allow them to badly disrupt Russian logistics:
My concern is that Ukraine aren't getting that many and can't these be easily knocked out. Talking from ignorance so would love to be corrected.
The Russians have managed to knock out some of the M777 howitzers, so they might eventually spot a HIMARS with a UAV and be able to drop an artillery shell on it. However they seem to be quite mobile, so if Ukraine are cautious they can pull them out of range during the day and then drive them up to firing positions during the night.
There have been some interesting videos of them reloading the things in fields, etc, which is suggestive of them being careful to distribute the reloads, so that Russia can't as easily catch one at an ammo dump.
I think Ukraine will gradually receive more of these over time.
One thing is utterly clear from this war - the days of lining up your artillery in a neat row, setting up a tea tent etc are gone. Shoot and scoot from individual positions (with high level coordination) is the only credible way forward.
A question is whether armouring artillery systems is worth it vs mobility.
There's another aspect to this that surprised me (it should not have): to accurately fire artillery, you need not only to know where the target is; you need to know where you are, and that means your position has to be surveyed. Hence all the piccies of squaddies looking through theodolites (I think - I doubt they are levels).
A lot of positions can be pre-surveyed and marked. Does military-grade GPS signals give enough accuracy for this purpose?
The US has the ability to degrade the L1C (consumer) GPS signal while maintaining accuracy by using the high gain antennae on the GPS Block III satellites to provide localised high precision GNSS via M-Code. This is the so-called Blue on Blue Electronic Attack which has been the holy grail of GPS ever since Clinton removed the 'wobble' from the civvie signal.
However, I don't know if they are doing the L1C degradation over Ukraine and they certainly won't be sharing any M-Code compatible. hardware with them in case it falls into Russian hands.
If FPN'd, it won't be Starmer but more immediately, if Boris is ousted, the next Prime Minister will be a Conservative. That's why all the Labour candidates are out a bit.
I'm almost tempted to go back in on Starmer - he's been out and back in more times than I can count, largely due people thinking Johnson is toast and the next PM will therefore be Tory. Has been profitable to back him long and sell him short (last round was buy at ~12 and sell at ~6, IRRC). But this time I'm thinking Johnson really might be toast in the short term before Starmer comes back in, so I'm sitting on my hands.
I'm green on Johnson exit markets and on next PM, but there's more for me if Johnson's still in place on 1 January and Starmer next PM. But I'll sacrifice that for the greater good of an end to the Johnson premiership
(Also have some old positions on Con most seats and a little on Con majority, which would both benefit from a competent Johnson replacement. Perhaps there will be a dead cat bounce even if the next PM is an idiot...)
(((Dan Hodges))) @DPJHodges · 5m We’re obviously not there yet. But it’s not inconceivable we’ll soon be at the point where the number of no confidence letters themselves exceed the number of votes required to remove Boris.
Be remiss of Brady not to share that with the PM.
Although Boris would no doubt go full-on Downfall mode "Ah, but those numbers include many letters from my loyalist MPs, to again thwart the forces ranged against me...."
What is our Penny up to?
I’d be pissed at not being promoted to the cabinet.
Perhaps she was offered a job. And turned it down!
It’s weird how DALL-E hasn’t *quite* mastered eyes and noses - yet. I suspect this is a lagging glitch from its former constraint - don’t use faces
Hair colours totally wrong for Betty and Barny. Fail.
The actual Flintstone's content there is very low. Outfits, jewellery, hairstyles on the ladies are totally different. I'm not 'knocking' the technology; it's interesting. But I think I'm missing something. It just runs through hundreds of thousands of stock images on a database and finesses them into a single image using the command - is that right? That's impressive but not an unexpected or radical development surely?
It does seem to be a key element of how it works. Where they talk about 'training data', that's all the world's art and photography libraries (one assumes) that has been run through it. Don't get me wrong; it's very clever, just not imo extremely radical.
Dalle-2 can be recreated by any moderately capable programmer, if given massive amounts of free computer time.
1. Take existing image recognition library 2. Lazily parse input text ("oil painting", "Homer Simpson" etc) 3. Start with 100 randomly generated images 4. Score those 100 against your input text (of course you'll get 0.0000001 for "oil painting" in even your best random image). Choose the best 10. 5. Randomly transform each of those 10 in 10 different ways so they don't move too much between images. 6. Keep scoring and keep randomly changing the images until you get to 0.73 on "oil painting" and 0.82 on "Homer Simpson"
I'm tempted to do it myself, just to prove there is no magic involved.
Yep - nothing new there at all Kevin Kelly covered it in Out of Control back in 1995 so it's been around since at least 1990 albeit randomly editing 10 pictures in 10 different ways would have taken a week back then.
So what? I imagine human consciousness went through some pretty primitive phases before getting where we are now, in fact "nothing new there at all" might be evolution's motto. Explaining something is not the same as explaining it away
Yes, I was thinking the exact same thing. At each stage of evolution some nerd could have said, in a nasal voice, “So what, it’s only got three cells”, or “so what, it’s just a fish”, or “so what, it still shits in the woods, lol” right up to “so what, anyone can write a play in blank verse”
Yet you’ve gone from yeast to William Shakespeare
I think the point is that some people (mentioning no names) want to ascribe the same value to the new tech as to the works of Shakespeare (or Van Gogh; whatever). Which is totally wrong: it looks similar, but in fact it's just mindless mimicry. Technologically impressive, and the implications on society are potentially vast, but it's not "thinking" in any real sense, because it isn't qualitatively different to "dumb" tech that already exists; just more powerful.
So, it doesn't mean we're on the verge of creating a race of sentient beings. Which I think is the undercurrent of the commentary from those who seek to overstate the achievement.
Complete missing of the point. You guys think you know something about computers but it has obviously never occurred to you that the nature of human consciousness is quite a difficult issue. Humans are qualitatively different from what they evolved from are they? They might be but the case needs making.
Look at the computers which beat world champions at chess and go. They evolved from dumb tech. They don't look very dumb to me.
I must say I find it weird when a PM blames a Chancellor for a policy eg corporation tax. You're the PM, are you claiming you couldn't influence what was happening?
(((Dan Hodges))) @DPJHodges · 5m We’re obviously not there yet. But it’s not inconceivable we’ll soon be at the point where the number of no confidence letters themselves exceed the number of votes required to remove Boris.
Be remiss of Brady not to share that with the PM.
Although Boris would no doubt go full-on Downfall mode "Ah, but those numbers include many letters from my loyalist MPs, to again thwart the forces ranged against me...."
What is our Penny up to?
I’d be pissed at not being promoted to the cabinet.
Perhaps she was offered a job. And turned it down!
I've just seen a squadron of pigs fly past my window.
It’s weird how DALL-E hasn’t *quite* mastered eyes and noses - yet. I suspect this is a lagging glitch from its former constraint - don’t use faces
Hair colours totally wrong for Betty and Barny. Fail.
The actual Flintstone's content there is very low. Outfits, jewellery, hairstyles on the ladies are totally different. I'm not 'knocking' the technology; it's interesting. But I think I'm missing something. It just runs through hundreds of thousands of stock images on a database and finesses them into a single image using the command - is that right? That's impressive but not an unexpected or radical development surely?
It does seem to be a key element of how it works. Where they talk about 'training data', that's all the world's art and photography libraries (one assumes) that has been run through it. Don't get me wrong; it's very clever, just not imo extremely radical.
Dalle-2 can be recreated by any moderately capable programmer, if given massive amounts of free computer time.
1. Take existing image recognition library 2. Lazily parse input text ("oil painting", "Homer Simpson" etc) 3. Start with 100 randomly generated images 4. Score those 100 against your input text (of course you'll get 0.0000001 for "oil painting" in even your best random image). Choose the best 10. 5. Randomly transform each of those 10 in 10 different ways so they don't move too much between images. 6. Keep scoring and keep randomly changing the images until you get to 0.73 on "oil painting" and 0.82 on "Homer Simpson"
I'm tempted to do it myself, just to prove there is no magic involved.
Yep - nothing new there at all Kevin Kelly covered it in Out of Control back in 1995 so it's been around since at least 1990 albeit randomly editing 10 pictures in 10 different ways would have taken a week back then.
So what? I imagine human consciousness went through some pretty primitive phases before getting where we are now, in fact "nothing new there at all" might be evolution's motto. Explaining something is not the same as explaining it away
Yes, I was thinking the exact same thing. At each stage of evolution some nerd could have said, in a nasal voice, “So what, it’s only got three cells”, or “so what, it’s just a fish”, or “so what, it still shits in the woods, lol” right up to “so what, anyone can write a play in blank verse”
Yet you’ve gone from yeast to William Shakespeare
I think the point is that some people (mentioning no names) want to ascribe the same value to the new tech as to the works of Shakespeare (or Van Gogh; whatever). Which is totally wrong: it looks similar, but in fact it's just mindless mimicry. Technologically impressive, and the implications on society are potentially vast, but it's not "thinking" in any real sense, because it isn't qualitatively different to "dumb" tech that already exists; just more powerful.
So, it doesn't mean we're on the verge of creating a race of sentient beings. Which I think is the undercurrent of the commentary from those who seek to overstate the achievement.
Complete missing of the point. You guys think you know something about computers but it has obviously never occurred to you that the nature of human consciousness is quite a difficult issue. Humans are qualitatively different from what they evolved from are they? They might be but the case needs making.
Look at the computers which beat world champions at chess and go. They evolved from dumb tech. They don't look very dumb to me.
Firstly, the animals we evolved from are also sentient, and we aren't close to sentient computers, nevermind sapient, nevermind... whatever special attribute you're assigning to humans on top of sapience.
But, yes, that is the point in a nutshell. Whether or not human consciousness/sentience/sapience is nothing special, and can be replicated with enough silicon and code, or whether it can't. And currently, the advances we've seen do absolutely nothing to answer that question.
"Do you think Boris Johnson should resign from his role as Prime Minister, or should he remain in his role?" 54% of conservative voters answered "should resign"
And that's using his preferred method of keepng the don't knows (13%) in.
An ex-civil servant writes: It's interesting how many of the ministerial resignation letters pay explicit tribute to the civil servants they have worked with in their departments. It's almost as if the "civil service are useless" meme is not shared by the politicians who actually work with them.
I have a horrible feeling the Queen is going to be dragged into all this before the month is out.
Nightmare for the Palace.
'Ok your majesty, when he comes for the weekly conflab, offer him a sherry and pop this little pill into it. Timed for 4 hours so he'll be back home scoffing a pie when it strikes, died of gluttony will be offical cod.'
The Speccie’s James Forsythe reports a senior 1922 executive member tells him “they now favour a delegation going to Johnson to tell him that it is over and that they will change the rules to allow another vote if he doesn’t quit”. #MenInGreySuits
Oof... this is not good. Understand some cabinet members loyal to the PM are only staying because they feel someone has to physically do the job "until a new boss comes". Even some of his closest allies know the game is up, despite public protestations. https://twitter.com/PaulBrandITV/status/1544634063113396225
On another but even more important topic, these two threads are interesting on how the longer-range precision artillery that Ukraine is now receiving from the West is going to allow them to badly disrupt Russian logistics:
My concern is that Ukraine aren't getting that many and can't these be easily knocked out. Talking from ignorance so would love to be corrected.
Such systems are very complicated (electrically, hydraulically, mechanically, etc.) and are only sustainable in the field by NATO forces because of a) years of experience and b) OEM/contractor support. So I think the new Ukrainian systems are more likely to be attrited by faults than Russian weapons.
That's a good point, except many of the ex-Soviet systems the Russians use are also complex. Don't underestimate the Ukrainian's capabilities in this. Access to spares'll be the bigger issue. Hopefully (ha!) the OEMs will have made field maintenance easy (yes, I know...)
With traditional artillery, gun tubes wear out (I don't know if that's the same for preloaded MLRS tubes). It'll be interesting to see if the Russians and Ukrainians have enough of these in storage, or can cannibalise enough of older systems.
AIUI the Himars and so on are pre-packaged units, and it is "slot this box here", aim, push the button.
On another but even more important topic, these two threads are interesting on how the longer-range precision artillery that Ukraine is now receiving from the West is going to allow them to badly disrupt Russian logistics:
My concern is that Ukraine aren't getting that many and can't these be easily knocked out. Talking from ignorance so would love to be corrected.
The Russians have managed to knock out some of the M777 howitzers, so they might eventually spot a HIMARS with a UAV and be able to drop an artillery shell on it. However they seem to be quite mobile, so if Ukraine are cautious they can pull them out of range during the day and then drive them up to firing positions during the night.
There have been some interesting videos of them reloading the things in fields, etc, which is suggestive of them being careful to distribute the reloads, so that Russia can't as easily catch one at an ammo dump.
I think Ukraine will gradually receive more of these over time.
One thing is utterly clear from this war - the days of lining up your artillery in a neat row, setting up a tea tent etc are gone. Shoot and scoot from individual positions (with high level coordination) is the only credible way forward.
A question is whether armouring artillery systems is worth it vs mobility.
There's another aspect to this that surprised me (it should not have): to accurately fire artillery, you need not only to know where the target is; you need to know where you are, and that means your position has to be surveyed. Hence all the piccies of squaddies looking through theodolites (I think - I doubt they are levels).
A lot of positions can be pre-surveyed and marked. Does military-grade GPS signals give enough accuracy for this purpose?
There was some brilliant footage last week of the Ukranians using a drone as a ‘spotter’ for long-range artillery, which was brilliantly innovative.
The MLRS really need to use each firing location once, and then move on quickly, keeping them well hidden during the day. The Russians will quickly work backwards to get the firing locations.
The Russians don't need to 'work backwards' from the trajectory of the round. They localise artillery/MLRS etc. using acoustics and seismic detection with the 'Penicillin' counter battery system.
It’s weird how DALL-E hasn’t *quite* mastered eyes and noses - yet. I suspect this is a lagging glitch from its former constraint - don’t use faces
Hair colours totally wrong for Betty and Barny. Fail.
The actual Flintstone's content there is very low. Outfits, jewellery, hairstyles on the ladies are totally different. I'm not 'knocking' the technology; it's interesting. But I think I'm missing something. It just runs through hundreds of thousands of stock images on a database and finesses them into a single image using the command - is that right? That's impressive but not an unexpected or radical development surely?
It does seem to be a key element of how it works. Where they talk about 'training data', that's all the world's art and photography libraries (one assumes) that has been run through it. Don't get me wrong; it's very clever, just not imo extremely radical.
Dalle-2 can be recreated by any moderately capable programmer, if given massive amounts of free computer time.
1. Take existing image recognition library 2. Lazily parse input text ("oil painting", "Homer Simpson" etc) 3. Start with 100 randomly generated images 4. Score those 100 against your input text (of course you'll get 0.0000001 for "oil painting" in even your best random image). Choose the best 10. 5. Randomly transform each of those 10 in 10 different ways so they don't move too much between images. 6. Keep scoring and keep randomly changing the images until you get to 0.73 on "oil painting" and 0.82 on "Homer Simpson"
I'm tempted to do it myself, just to prove there is no magic involved.
Yep - nothing new there at all Kevin Kelly covered it in Out of Control back in 1995 so it's been around since at least 1990 albeit randomly editing 10 pictures in 10 different ways would have taken a week back then.
So what? I imagine human consciousness went through some pretty primitive phases before getting where we are now, in fact "nothing new there at all" might be evolution's motto. Explaining something is not the same as explaining it away
Yes, I was thinking the exact same thing. At each stage of evolution some nerd could have said, in a nasal voice, “So what, it’s only got three cells”, or “so what, it’s just a fish”, or “so what, it still shits in the woods, lol” right up to “so what, anyone can write a play in blank verse”
Yet you’ve gone from yeast to William Shakespeare
I think the point is that some people (mentioning no names) want to ascribe the same value to the new tech as to the works of Shakespeare (or Van Gogh; whatever). Which is totally wrong: it looks similar, but in fact it's just mindless mimicry. Technologically impressive, and the implications on society are potentially vast, but it's not "thinking" in any real sense, because it isn't qualitatively different to "dumb" tech that already exists; just more powerful.
So, it doesn't mean we're on the verge of creating a race of sentient beings. Which I think is the undercurrent of the commentary from those who seek to overstate the achievement.
Complete missing of the point. You guys think you know something about computers but it has obviously never occurred to you that the nature of human consciousness is quite a difficult issue. Humans are qualitatively different from what they evolved from are they? They might be but the case needs making.
Look at the computers which beat world champions at chess and go. They evolved from dumb tech. They don't look very dumb to me.
Ask one to play tic-tac-toe.
Here’s a practical experiment (I’m not making any claims either way, incidentally)
DALL-E 2 drew this cartoon and another AI wrote this joke
And it made me chuckle. It is funny. It might be stolen from somewhere, but I haven’t seen it before. But then all humour is derivative…
Anyway: there was human intervention. Someone wrote a few prompts and the prompter had to stitch together two jokes from the AI to get the full comic strip
(I’ll try and find the Twitter thread where the prompter explains what he did)
Now, if a computer had done ALL of that by itself, how would you react? If AI could create entire comic strips from the get-go, strips that are genuinely funny and maybe even insightful - and I think they are possibly quite close to this, as we see here - would that be signs of “intelligence”?
In short: if you could just type in “DALL-E be funny” and it was funny on demand, how would you feel?
It’s weird how DALL-E hasn’t *quite* mastered eyes and noses - yet. I suspect this is a lagging glitch from its former constraint - don’t use faces
Hair colours totally wrong for Betty and Barny. Fail.
The actual Flintstone's content there is very low. Outfits, jewellery, hairstyles on the ladies are totally different. I'm not 'knocking' the technology; it's interesting. But I think I'm missing something. It just runs through hundreds of thousands of stock images on a database and finesses them into a single image using the command - is that right? That's impressive but not an unexpected or radical development surely?
It does seem to be a key element of how it works. Where they talk about 'training data', that's all the world's art and photography libraries (one assumes) that has been run through it. Don't get me wrong; it's very clever, just not imo extremely radical.
Dalle-2 can be recreated by any moderately capable programmer, if given massive amounts of free computer time.
1. Take existing image recognition library 2. Lazily parse input text ("oil painting", "Homer Simpson" etc) 3. Start with 100 randomly generated images 4. Score those 100 against your input text (of course you'll get 0.0000001 for "oil painting" in even your best random image). Choose the best 10. 5. Randomly transform each of those 10 in 10 different ways so they don't move too much between images. 6. Keep scoring and keep randomly changing the images until you get to 0.73 on "oil painting" and 0.82 on "Homer Simpson"
I'm tempted to do it myself, just to prove there is no magic involved.
Yep - nothing new there at all Kevin Kelly covered it in Out of Control back in 1995 so it's been around since at least 1990 albeit randomly editing 10 pictures in 10 different ways would have taken a week back then.
So what? I imagine human consciousness went through some pretty primitive phases before getting where we are now, in fact "nothing new there at all" might be evolution's motto. Explaining something is not the same as explaining it away
Yes, I was thinking the exact same thing. At each stage of evolution some nerd could have said, in a nasal voice, “So what, it’s only got three cells”, or “so what, it’s just a fish”, or “so what, it still shits in the woods, lol” right up to “so what, anyone can write a play in blank verse”
Yet you’ve gone from yeast to William Shakespeare
I think the point is that some people (mentioning no names) want to ascribe the same value to the new tech as to the works of Shakespeare (or Van Gogh; whatever). Which is totally wrong: it looks similar, but in fact it's just mindless mimicry. Technologically impressive, and the implications on society are potentially vast, but it's not "thinking" in any real sense, because it isn't qualitatively different to "dumb" tech that already exists; just more powerful.
So, it doesn't mean we're on the verge of creating a race of sentient beings. Which I think is the undercurrent of the commentary from those who seek to overstate the achievement.
Complete missing of the point. You guys think you know something about computers but it has obviously never occurred to you that the nature of human consciousness is quite a difficult issue. Humans are qualitatively different from what they evolved from are they? They might be but the case needs making.
Look at the computers which beat world champions at chess and go. They evolved from dumb tech. They don't look very dumb to me.
Firstly, the animals we evolved from are also sentient, and we aren't close to sentient computers, nevermind sapient, nevermind... whatever special attribute you're assigning to humans on top of sapience.
But, yes, that is the point in a nutshell. Whether or not human consciousness/sentience/sapience is nothing special, and can be replicated with enough silicon and code, or whether it can. And currently, the advances we've seen do absolutely nothing to answer that question.
Just wrong. I am not ascribing anything to humans, you are, wtf is sapience meant to mean?
The "animals" we evolved from may be sentient (but probably not self aware) but how far up the chain does that go? To our last common ancestor with the sponges?
And you have missed the point Turing was trying to make. The ONLY evidence you are ever going to get about the sentience or self awareness of anything else (including other humans) is behavioural.
Beth saying 1922 could take until next week to change rules even if they decide today it was a good idea.
Far too slow.
What has happened to the Tories? Say what you like about them, but in years past they could be counted on to be ruthlessly efficient in removing an unwanted PM.
For what it's worth, and I emphasise that, Polly Toynbee in the Guardian is suggesting that Johnson is proposing to gain control of the House of Lords by creating 50 new Tory peers.
Probably lining up 50 new Conservative treehouse and holiday villa donors.
Oof... this is not good. Understand some cabinet members loyal to the PM are only staying because they feel someone has to physically do the job "until a new boss comes". Even some of his closest allies know the game is up, despite public protestations. https://twitter.com/PaulBrandITV/status/1544634063113396225
The loyalists have clearly been shaken to the core by the extent of the resignations, and the prospect of never holding office again.
For what it's worth, and I emphasise that, Polly Toynbee in the Guardian is suggesting that Johnson is proposing to gain control of the House of Lords by creating 50 new Tory peers.
Meanwhile here's the Oxford Mail's lead story this morning:
"Banbury man caught with sheep pornography gets suspended sentence"
Isn't that Cameron Country?
No.
Banbury's in Banbury constituency: Cameron was MP for the Witney constituency.
Banbury constituency is more or less identical to Cherwell District, where the Tories have 25 seats and the opposition 23; any sensible betting person would put their money on Cherwell following the rest of Oxfordshire into Labour or LibDem control next May.
Witney constituency is completely identical to West Oxfordshire District, where the LibDems now lead a ruling LibDem/Labour/Green Alliance and the Tories have given up any pretence of offering any kind of constructive opposition. When Cameron was MP, the Tories had four times as many WODC seats as LD/Lab/Grn combined.
As elsewhere in the educated South East, the combination of radically changed demographics and the damage inflicted by the Tories' self-destructively toxic cocktail of Osbornism and Johnsonism now make concepts like "Cameron country" naive 20th century delusions in Oxfordshire.
DALL-E is now getting scarily, scarily good at faces. Despite the fact that faces have only been “permitted” for a week or two
That person does not exist. Yet the only tiny clue is some blurring at the hairline/temple
Conclusions: the potential for porn deep fakes is troubling. You’re still not allowed to use “real” faces - but inevitably that prohibition will fail. You’ll be able to depict anyone doing anything with 100% accuracy
Also: photography as “art” is finished. Dead. DALL-E does this for free in seconds, 24/7
Also: what happens when DALL-E is allowed to do moving images? Films? TV? With AI voice generation that will open up amazing possibilities for movies made with entirely fake “people” - who don’t need to be paid
What a banal understanding you have of what photography as 'art' is!
You really don't have an idea what youre talking about do you?
Given that you think DALL-E 2 is roughly the same as CGI or Photoshop, I suggest we don’t have this debate as you will look a fool. Yet again
God, no. He can't go to the Palace to resign as PM without a successor - if he goes to the Palace today it can only be to call an election...
Would he not inform the Queen he intends to resign and will be starting a leadership contest to seek his successor before he tells the public?
Possibly, but that would more likely be in a private audience or phone call than the formal motorcade procession with news helicopters over the Mall that the phrase "going to the Palace" calls to mind.
Labour is gearing up for a snap election amid talk Boris Johnson could call a quickfire vote to head off rebel Tory MPs who are mounting a fresh bid to oust him from No 10
DALL-E is now getting scarily, scarily good at faces. Despite the fact that faces have only been “permitted” for a week or two
That person does not exist. Yet the only tiny clue is some blurring at the hairline/temple
Conclusions: the potential for porn deep fakes is troubling. You’re still not allowed to use “real” faces - but inevitably that prohibition will fail. You’ll be able to depict anyone doing anything with 100% accuracy
Also: photography as “art” is finished. Dead. DALL-E does this for free in seconds, 24/7
Also: what happens when DALL-E is allowed to do moving images? Films? TV? With AI voice generation that will open up amazing possibilities for movies made with entirely fake “people” - who don’t need to be paid
What a banal understanding you have of what photography as 'art' is!
You really don't have an idea what youre talking about do you?
Given that you think DALL-E 2 is roughly the same as CGI or Photoshop, I suggest we don’t have this debate as you will look a fool. Yet again
On another but even more important topic, these two threads are interesting on how the longer-range precision artillery that Ukraine is now receiving from the West is going to allow them to badly disrupt Russian logistics:
My concern is that Ukraine aren't getting that many and can't these be easily knocked out. Talking from ignorance so would love to be corrected.
The Russians have managed to knock out some of the M777 howitzers, so they might eventually spot a HIMARS with a UAV and be able to drop an artillery shell on it. However they seem to be quite mobile, so if Ukraine are cautious they can pull them out of range during the day and then drive them up to firing positions during the night.
There have been some interesting videos of them reloading the things in fields, etc, which is suggestive of them being careful to distribute the reloads, so that Russia can't as easily catch one at an ammo dump.
I think Ukraine will gradually receive more of these over time.
One thing is utterly clear from this war - the days of lining up your artillery in a neat row, setting up a tea tent etc are gone. Shoot and scoot from individual positions (with high level coordination) is the only credible way forward.
A question is whether armouring artillery systems is worth it vs mobility.
There's another aspect to this that surprised me (it should not have): to accurately fire artillery, you need not only to know where the target is; you need to know where you are, and that means your position has to be surveyed. Hence all the piccies of squaddies looking through theodolites (I think - I doubt they are levels).
A lot of positions can be pre-surveyed and marked. Does military-grade GPS signals give enough accuracy for this purpose?
There was some brilliant footage last week of the Ukranians using a drone as a ‘spotter’ for long-range artillery, which was brilliantly innovative.
The MLRS really need to use each firing location once, and then move on quickly, keeping them well hidden during the day. The Russians will quickly work backwards to get the firing locations.
The Russians don't need to 'work backwards' from the trajectory of the round. They localise artillery/MLRS etc. using acoustics and seismic detection with the 'Penicillin' counter battery system.
The thermal detection part of the system is unlikely to work at HIMARS range, surely - and do rocket launches provide a sufficient seismic signal (acoustics wouldn't be much use for the claimed 'instantaneous' detection) ?
Also, how many such systems are deployed given it was developed only recently ?
On another but even more important topic, these two threads are interesting on how the longer-range precision artillery that Ukraine is now receiving from the West is going to allow them to badly disrupt Russian logistics:
My concern is that Ukraine aren't getting that many and can't these be easily knocked out. Talking from ignorance so would love to be corrected.
The Russians have managed to knock out some of the M777 howitzers, so they might eventually spot a HIMARS with a UAV and be able to drop an artillery shell on it. However they seem to be quite mobile, so if Ukraine are cautious they can pull them out of range during the day and then drive them up to firing positions during the night.
There have been some interesting videos of them reloading the things in fields, etc, which is suggestive of them being careful to distribute the reloads, so that Russia can't as easily catch one at an ammo dump.
I think Ukraine will gradually receive more of these over time.
One thing is utterly clear from this war - the days of lining up your artillery in a neat row, setting up a tea tent etc are gone. Shoot and scoot from individual positions (with high level coordination) is the only credible way forward.
A question is whether armouring artillery systems is worth it vs mobility.
There's another aspect to this that surprised me (it should not have): to accurately fire artillery, you need not only to know where the target is; you need to know where you are, and that means your position has to be surveyed. Hence all the piccies of squaddies looking through theodolites (I think - I doubt they are levels).
A lot of positions can be pre-surveyed and marked. Does military-grade GPS signals give enough accuracy for this purpose?
There was some brilliant footage last week of the Ukranians using a drone as a ‘spotter’ for long-range artillery, which was brilliantly innovative.
The MLRS really need to use each firing location once, and then move on quickly, keeping them well hidden during the day. The Russians will quickly work backwards to get the firing locations.
The Russians don't need to 'work backwards' from the trajectory of the round. They localise artillery/MLRS etc. using acoustics and seismic detection with the 'Penicillin' counter battery system.
Interesting. What range does that have, or does it feed into other systems? So it’s imperative that the Ukranians move the MLRS as quickly as possible after they’ve let off their rockets.
It’s weird how DALL-E hasn’t *quite* mastered eyes and noses - yet. I suspect this is a lagging glitch from its former constraint - don’t use faces
Hair colours totally wrong for Betty and Barny. Fail.
The actual Flintstone's content there is very low. Outfits, jewellery, hairstyles on the ladies are totally different. I'm not 'knocking' the technology; it's interesting. But I think I'm missing something. It just runs through hundreds of thousands of stock images on a database and finesses them into a single image using the command - is that right? That's impressive but not an unexpected or radical development surely?
It does seem to be a key element of how it works. Where they talk about 'training data', that's all the world's art and photography libraries (one assumes) that has been run through it. Don't get me wrong; it's very clever, just not imo extremely radical.
Dalle-2 can be recreated by any moderately capable programmer, if given massive amounts of free computer time.
1. Take existing image recognition library 2. Lazily parse input text ("oil painting", "Homer Simpson" etc) 3. Start with 100 randomly generated images 4. Score those 100 against your input text (of course you'll get 0.0000001 for "oil painting" in even your best random image). Choose the best 10. 5. Randomly transform each of those 10 in 10 different ways so they don't move too much between images. 6. Keep scoring and keep randomly changing the images until you get to 0.73 on "oil painting" and 0.82 on "Homer Simpson"
I'm tempted to do it myself, just to prove there is no magic involved.
Yep - nothing new there at all Kevin Kelly covered it in Out of Control back in 1995 so it's been around since at least 1990 albeit randomly editing 10 pictures in 10 different ways would have taken a week back then.
So what? I imagine human consciousness went through some pretty primitive phases before getting where we are now, in fact "nothing new there at all" might be evolution's motto. Explaining something is not the same as explaining it away
Yes, I was thinking the exact same thing. At each stage of evolution some nerd could have said, in a nasal voice, “So what, it’s only got three cells”, or “so what, it’s just a fish”, or “so what, it still shits in the woods, lol” right up to “so what, anyone can write a play in blank verse”
Yet you’ve gone from yeast to William Shakespeare
Yes but there is a but. Ability and consciousness are different things. Robots make cars. I have a book that knows the complete works of Shakespeare and an internet connection that knows how to access lots of stuff.
None of that amounts to a single grain of consciousness necessarily. We do not know whether awareness, knowing that you know, consciousness, capacity to actually feel stuff in animate objects like PBers is an inherent capacity of matter in particular configurations (it seems unlikely) or a non materially based function that arises with evolution in some cases but not all- which seems about as unlikely. (Shakespeare but not mushrooms perhaps). Or an implant from aliens or whatever - which also seems unlikely.
If it is a configuration of matter then it seems to me to imply a massive moral duty on us not to mess with it, for if that is not playing both God and Mother Nature I don't know what is.
The precautionary principle should, but won't, apply.
On another but even more important topic, these two threads are interesting on how the longer-range precision artillery that Ukraine is now receiving from the West is going to allow them to badly disrupt Russian logistics:
My concern is that Ukraine aren't getting that many and can't these be easily knocked out. Talking from ignorance so would love to be corrected.
The Russians have managed to knock out some of the M777 howitzers, so they might eventually spot a HIMARS with a UAV and be able to drop an artillery shell on it. However they seem to be quite mobile, so if Ukraine are cautious they can pull them out of range during the day and then drive them up to firing positions during the night.
There have been some interesting videos of them reloading the things in fields, etc, which is suggestive of them being careful to distribute the reloads, so that Russia can't as easily catch one at an ammo dump.
I think Ukraine will gradually receive more of these over time.
One thing is utterly clear from this war - the days of lining up your artillery in a neat row, setting up a tea tent etc are gone. Shoot and scoot from individual positions (with high level coordination) is the only credible way forward.
A question is whether armouring artillery systems is worth it vs mobility.
There's another aspect to this that surprised me (it should not have): to accurately fire artillery, you need not only to know where the target is; you need to know where you are, and that means your position has to be surveyed. Hence all the piccies of squaddies looking through theodolites (I think - I doubt they are levels).
A lot of positions can be pre-surveyed and marked. Does military-grade GPS signals give enough accuracy for this purpose?
The US has the ability to degrade the L1C (consumer) GPS signal while maintaining accuracy by using the high gain antennae on the GPS Block III satellites to provide localised high precision GNSS via M-Code. This is the so-called Blue on Blue Electronic Attack which has been the holy grail of GPS ever since Clinton removed the 'wobble' from the civvie signal.
However, I don't know if they are doing the L1C degradation over Ukraine and they certainly won't be sharing any M-Code compatible. hardware with them in case it falls into Russian hands.
A consequence of widely available, cheap satellite imagery is easily providing pre-surveyed locations - without requiring anyone actually goes there first. The imagery comes with extremely precise location data. So it would be possible to get around problems with GPS, unless the Russians start screwing with commercial imagery providers…
“See this on the image? Well, 10m due west of that public toilet is X,Y, so park there and set your equipment accordingly.”
Day 6 and still have a bright LFT/LFD line. Work are dropping me a laptop off so I can work the rest of my isolation period.
Seems to be random how long people are positive for with each person's immune system being different and different variants out there with longer viral shedding.
I feel better, but still can tell I feel 'off' so to speak.
Hope you recover soon. I'm on day 13 and I've still got a cough and don't feel 100%.
It’s weird how DALL-E hasn’t *quite* mastered eyes and noses - yet. I suspect this is a lagging glitch from its former constraint - don’t use faces
Hair colours totally wrong for Betty and Barny. Fail.
The actual Flintstone's content there is very low. Outfits, jewellery, hairstyles on the ladies are totally different. I'm not 'knocking' the technology; it's interesting. But I think I'm missing something. It just runs through hundreds of thousands of stock images on a database and finesses them into a single image using the command - is that right? That's impressive but not an unexpected or radical development surely?
It does seem to be a key element of how it works. Where they talk about 'training data', that's all the world's art and photography libraries (one assumes) that has been run through it. Don't get me wrong; it's very clever, just not imo extremely radical.
Dalle-2 can be recreated by any moderately capable programmer, if given massive amounts of free computer time.
1. Take existing image recognition library 2. Lazily parse input text ("oil painting", "Homer Simpson" etc) 3. Start with 100 randomly generated images 4. Score those 100 against your input text (of course you'll get 0.0000001 for "oil painting" in even your best random image). Choose the best 10. 5. Randomly transform each of those 10 in 10 different ways so they don't move too much between images. 6. Keep scoring and keep randomly changing the images until you get to 0.73 on "oil painting" and 0.82 on "Homer Simpson"
I'm tempted to do it myself, just to prove there is no magic involved.
Yep - nothing new there at all Kevin Kelly covered it in Out of Control back in 1995 so it's been around since at least 1990 albeit randomly editing 10 pictures in 10 different ways would have taken a week back then.
So what? I imagine human consciousness went through some pretty primitive phases before getting where we are now, in fact "nothing new there at all" might be evolution's motto. Explaining something is not the same as explaining it away
Yes, I was thinking the exact same thing. At each stage of evolution some nerd could have said, in a nasal voice, “So what, it’s only got three cells”, or “so what, it’s just a fish”, or “so what, it still shits in the woods, lol” right up to “so what, anyone can write a play in blank verse”
Yet you’ve gone from yeast to William Shakespeare
I think the point is that some people (mentioning no names) want to ascribe the same value to the new tech as to the works of Shakespeare (or Van Gogh; whatever). Which is totally wrong: it looks similar, but in fact it's just mindless mimicry. Technologically impressive, and the implications on society are potentially vast, but it's not "thinking" in any real sense, because it isn't qualitatively different to "dumb" tech that already exists; just more powerful.
So, it doesn't mean we're on the verge of creating a race of sentient beings. Which I think is the undercurrent of the commentary from those who seek to overstate the achievement.
You’re missing the point, too, with all due respect
What if “sentience” is just…. mimicry. What if intelligence is just autocomplete (like GPT3)? In this light, we are machines reacting to stimuli, Free Will is an illusion, so we are much closer to these computers than we realise
Put it differently: we like to believe we are unique talented beings, with some divine spark of Whatever, therefore we are insulted when a “dumb” machine appears to do what we, and only we, can do. Create, imagine, paint, write, dream, sing, and make others laugh
And yet what if it turns out the algorithms for this are really quite simple? And you just need a massive data set? Bit of a bummer for human self regard, but there we go
Comments
https://twitter.com/holyroodmandy/status/1544632163722723328
The MLRS really need to use each firing location once, and then move on quickly, keeping them well hidden during the day. The Russians will quickly work backwards to get the firing locations.
(runs for cover...)
I'm green on Johnson exit markets and on next PM, but there's more for me if Johnson's still in place on 1 January and Starmer next PM. But I'll sacrifice that for the greater good of an end to the Johnson premiership
(Also have some old positions on Con most seats and a little on Con majority, which would both benefit from a competent Johnson replacement. Perhaps there will be a dead cat bounce even if the next PM is an idiot...)
How much of a cock has Zahawi made himself look for a brief run as Chequebook Man?!
Out by teatime.
And this can't be overturned by the 1922 Committee as it's in the party Constitution (Schedule 2).
https://public.conservatives.com/organisation-department/202101/Conservative Party Constitution as amended January 2021.pdf
2. A Leader resigning from the Leadership of the Party is not eligible for re-nomination in the consequent Leadership election.
Nightmare for the Palace.
What a great line!
But, yes, that is the point in a nutshell. Whether or not human consciousness/sentience/sapience is nothing special, and can be replicated with enough silicon and code, or whether it can't. And currently, the advances we've seen do absolutely nothing to answer that question.
I wonder what happened to 'Brexit will stop cross border police cooperation'?
https://yougov.co.uk/topics/politics/survey-results/daily/2022/07/05/b572b/1
"Do you think Boris Johnson should resign from his role as Prime Minister, or should he remain in his role?"
54% of conservative voters answered "should resign"
And that's using his preferred method of keepng the don't knows (13%) in.
https://twitter.com/PaulBrandITV/status/1544634063113396225
It's the role he was born for.
Rather like NLAWs. Or, I guess, champagne corks.
Cites the infamous letter to the Times in 1951.
DALL-E 2 drew this cartoon and another AI wrote this joke
And it made me chuckle. It is funny. It might be stolen from somewhere, but I haven’t seen it before. But then all humour is derivative…
Anyway: there was human intervention. Someone wrote a few prompts and the prompter had to stitch together two jokes from the AI to get the full comic strip
(I’ll try and find the Twitter thread where the prompter explains what he did)
Now, if a computer had done ALL of that by itself, how would you react? If AI could create entire comic strips from the get-go, strips that are genuinely funny and maybe even insightful - and I think they are possibly quite close to this, as we see here - would that be signs of “intelligence”?
In short: if you could just type in “DALL-E be funny” and it was funny on demand, how would you feel?
The Tories have a majority in Parliament so Johnson cannot be removed as PM unless Tory MPs vote no confidence in him.
After the FTPA calling a general election is again the PM's prerogative
The "animals" we evolved from may be sentient (but probably not self aware) but how far up the chain does that go? To our last common ancestor with the sponges?
And you have missed the point Turing was trying to make. The ONLY evidence you are ever going to get about the sentience or self awareness of anything else (including other humans) is behavioural.
What has happened to the Tories? Say what you like about them, but in years past they could be counted on to be ruthlessly efficient in removing an unwanted PM.
These days, they are worse than Labour.
Get on with it, you woolies.
Banbury's in Banbury constituency: Cameron was MP for the Witney constituency.
Banbury constituency is more or less identical to Cherwell District, where the Tories have 25 seats and the opposition 23; any sensible betting person would put their money on Cherwell following the rest of Oxfordshire into Labour or LibDem control next May.
Witney constituency is completely identical to West Oxfordshire District, where the LibDems now lead a ruling LibDem/Labour/Green Alliance and the Tories have given up any pretence of offering any kind of constructive opposition. When Cameron was MP, the Tories had four times as many WODC seats as LD/Lab/Grn combined.
As elsewhere in the educated South East, the combination of radically changed demographics and the damage inflicted by the Tories' self-destructively toxic cocktail of Osbornism and Johnsonism now make concepts like "Cameron country" naive 20th century delusions in Oxfordshire.
That is reproducable a hundred times without anything distinguishing it from the original
Also, how many such systems are deployed given it was developed only recently ?
None of that amounts to a single grain of consciousness necessarily. We do not know whether awareness, knowing that you know, consciousness, capacity to actually feel stuff in animate objects like PBers is an inherent capacity of matter in particular configurations (it seems unlikely) or a non materially based function that arises with evolution in some cases but not all- which seems about as unlikely. (Shakespeare but not mushrooms perhaps). Or an implant from aliens or whatever - which also seems unlikely.
If it is a configuration of matter then it seems to me to imply a massive moral duty on us not to mess with it, for if that is not playing both God and Mother Nature I don't know what is.
The precautionary principle should, but won't, apply.
“See this on the image? Well, 10m due west of that public toilet is X,Y, so park there and set your equipment accordingly.”
The Mind is Flat: The Illusion of Mental Depth and The Improvised Mind