I have no applications in my life where I need AI.
Still less, one that I would pay for.
I suspect that I am far, far more reflective of AI than the tech bros who own it.
I've had several friends and family (who are generally not technical people) mention they've resorted to using ChatGPT for one specific reason - medical advice. They either can't get a GP appointment in a reasonable timescale or the GP has brushed off their concerns, so they consult AI.
If my limited sample is in any way representative this is possibly quite widespread.
There are an extraordinary number of areas where LLMs are incredibly useful in "real life" - medical advice, legal advice, comparing contracts and documents, summarizing key points or issues to be aware of.
I use them all the time to build disposable, throw away, apps that I host that do one specific thing for me.
Are they "conscious"? Well, given their limited context window, and inability to learn skills outside said context window (see attempts to get LLMs to build things in Z80 assembly language), then I'd go with "no".
But if you're not using them today, you are making your life much harder than it need be.
The Telegraph has revealed that children, including ones with learning difficulties, have been promised up to £500 in vouchers, which can be redeemed at retailers like XBOX & Uber Eats, if they take part in an NHS 'transgender' trial to block their puberty
Not necessarily. Actually *reading* the article it says...
It can now be revealed that the children taking part in the trial will be incentivised with the promise of up to £500 in vouchers for completing psychometric tests. These will measure the effect of the puberty blockers on their brains, including effects on impulse control and memory...Participants, aged under 16, will receive £30 vouchers for each of the 15 cognitive assessments they complete, as well as £15 vouchers for each of the three MRI scans they undertake, over the two years of the trial...Children in another arm of the trial, not taking the drugs, will be given £15 vouchers for each bone density scan and blood test they have.
Points to note
The children are already in the trial and have been randomised into one of the two arms, so they've already been consented.
Those in the intervention arm (on the drugs) are paid a nominal amount (£15 or £30) for each of the tests (cogntive assessments or MRI scans) they undertake
Those in the control arm (not on the drugs) are paid a nominal amount (£15) for each of the tests (bone density scan and blood test) they undertake
In short the participants are paid nominal sums to take part in tests they have already agreed to prior to assignment, and that this is an incentive to complete the trial, not an incentive to take part in the trial. Since a criticism of many studies in the trans sphere is dropout rate, this is a good idea.
There's a concept in RCT called "equipoise". The surgeons/doctors in a study must be genuinely undecided/unbiased as to which of the two arms is better, and the patients must be equally noncommittal: this enables randomisation to take place believably. In trans studies this is difficult since if the patients believe the study is biased or bent towards one outcome they will refuse to take part or simply bugger off to Europe and get the treatment they want there (Baroness Cass pointed this out as a reason to undertake the PATHWAY studies). The PATHWAY designers seem to have created a study that enables equipoise. The fact that these sums are paid to make sure they stay in the study prevents dropout
(Incidentally @isam, you said last night you wanted to know how tests of drugs in humans take place. You might want to google the term "CTIMP")
The Bar Council has decided to have a Commissioner for Conduct following a report by Harriet Harman into sexism, misogyny and other bad behaviour. It has decided to appoint former Tory MP Maria Miller to this role. She is the Culture Secretary in Cameron's government who had to resign following a report showing that she had claimed ten of thousands of pounds of expenses to which she was not entitled. She then gave a 32 second apology to the Commons, was widely criticised for this and had to apologise again, not just for the expenses claim but also for her failure to co-operate with the investigation.
And this was the best candidate the Bar Council could come up with?
There has been no transparency to practising subscriber barristers about the process or terms of appointment of Miller or, come to that, Harman.
Structural dishonesty and tolerance of it are big problems in our country.
I have no applications in my life where I need AI.
Still less, one that I would pay for.
I suspect that I am far, far more reflective of AI than the tech bros who own it.
Rare that we agree and I'm not saying I've never seen AI but it seems a case of using 200 words when 20 will do just fine.
It's a constipation of verbage designed literally to fill a space with nothing, words which I was once told didn't come easy but now do it would seem.
As for the world of work, weren't computers supposedly to allow us to have a life of leisure? Yes, it didn't for me either - indeed, the volume of information created by computer systems is analogous to the amount of verbage created by some AI products.
I warned my successors when I retired - you either manage the information or the information will manage you.
That's the end of the world of work, not through replacement but asphyxiation - we will drown in our information and suffocate on our words simultaenously.
This is a total misunderstanding of what AI can do. It’s much much more than just generating text for publication. I would argue that that’s it’s least useful function.
I’m a solicitor and I use it every day. Not to write anything for me and not to summarise anything, but to discuss concepts, look for strengths and weaknesses in commercial positions, and to sense check ideas. It’s use in this manner is massively under appreciated, by my colleagues also.
The Dutch has passed a law, as is being reported, that will tax unrealised capital gains at 36%. Due to come in 2028. Although there is some discussion as to changes.
Starmer's inability to keep staff is quite interesting. His whole pitch was he knew how to run the office of a big public sector position. The turn over in less than 2 years of people who are thought of pros is becoming a thing.
I have no applications in my life where I need AI.
Still less, one that I would pay for.
I suspect that I am far, far more reflective of AI than the tech bros who own it.
Rare that we agree and I'm not saying I've never seen AI but it seems a case of using 200 words when 20 will do just fine.
It's a constipation of verbage designed literally to fill a space with nothing, words which I was once told didn't come easy but now do it would seem.
As for the world of work, weren't computers supposedly to allow us to have a life of leisure? Yes, it didn't for me either - indeed, the volume of information created by computer systems is analogous to the amount of verbage created by some AI products.
I warned my successors when I retired - you either manage the information or the information will manage you.
That's the end of the world of work, not through replacement but asphyxiation - we will drown in our information and suffocate on our words simultaenously.
“We will drown.” Maybe. Or maybe tooling shifts toward compression and synthesis rather than raw generation. Historically we oscillate — first we overproduce, then we build better filters.
I do think you’re right about the prolixity though. The cost of producing words has collapsed, and when production becomes trivial, volume explodes. That absolutely risks more noise, not more clarity.
The question isn’t whether more text gets generated — it will. The question is whether we get better at constraining, filtering and compressing it. Used lazily, these tools bloat everything. Used well, they actually reduce cognitive load.
I hope you're right and we can advance through the overproduction stage relatively swiftly though I'm not optimistic. Just as we keep boxes of paperwork for decades after they have ceased to have irrelevance, it may well be we will be retaining the data and verbal information long after it is really required.
I have no applications in my life where I need AI.
Still less, one that I would pay for.
I suspect that I am far, far more reflective of AI than the tech bros who own it.
Rare that we agree and I'm not saying I've never seen AI but it seems a case of using 200 words when 20 will do just fine.
It's a constipation of verbage designed literally to fill a space with nothing, words which I was once told didn't come easy but now do it would seem.
As for the world of work, weren't computers supposedly to allow us to have a life of leisure? Yes, it didn't for me either - indeed, the volume of information created by computer systems is analogous to the amount of verbage created by some AI products.
I warned my successors when I retired - you either manage the information or the information will manage you.
That's the end of the world of work, not through replacement but asphyxiation - we will drown in our information and suffocate on our words simultaenously.
We've already seen that in recruitment.
It's more efficient to apply for a job with AI, it's more efficient to screen applications by AI. Combine the two and the system collapses.
I have no applications in my life where I need AI.
Still less, one that I would pay for.
I suspect that I am far, far more reflective of AI than the tech bros who own it.
Rare that we agree and I'm not saying I've never seen AI but it seems a case of using 200 words when 20 will do just fine.
It's a constipation of verbage designed literally to fill a space with nothing, words which I was once told didn't come easy but now do it would seem.
As for the world of work, weren't computers supposedly to allow us to have a life of leisure? Yes, it didn't for me either - indeed, the volume of information created by computer systems is analogous to the amount of verbage created by some AI products.
I warned my successors when I retired - you either manage the information or the information will manage you.
That's the end of the world of work, not through replacement but asphyxiation - we will drown in our information and suffocate on our words simultaenously.
This is a total misunderstanding of what AI can do. It’s much much more than just generating text for publication. I would argue that that’s it’s least useful function.
I’m a solicitor and I use it every day. Not to write anything for me and not to summarise anything, but to discuss concepts, look for strengths and weaknesses in commercial positions, and to sense check ideas. It’s use in this manner is massively under appreciated, by my colleagues also.
Deep thinking / planning mode I also find really useful. You do have to stand your ground, be confident to say no, and not give into the sycophant nature of many LLMs. So its a little bit chicken and egg in that respect, if you don't know what you are doing, you will accept far too much of what it says, if you do have a good knowledge you can challenge and debate.
I have no applications in my life where I need AI.
Still less, one that I would pay for.
I suspect that I am far, far more reflective of AI than the tech bros who own it.
Rare that we agree and I'm not saying I've never seen AI but it seems a case of using 200 words when 20 will do just fine.
It's a constipation of verbage designed literally to fill a space with nothing, words which I was once told didn't come easy but now do it would seem.
As for the world of work, weren't computers supposedly to allow us to have a life of leisure? Yes, it didn't for me either - indeed, the volume of information created by computer systems is analogous to the amount of verbage created by some AI products.
I warned my successors when I retired - you either manage the information or the information will manage you.
That's the end of the world of work, not through replacement but asphyxiation - we will drown in our information and suffocate on our words simultaenously.
This is a total misunderstanding of what AI can do. It’s much much more than just generating text for publication. I would argue that that’s it’s least useful function.
I’m a solicitor and I use it every day. Not to write anything for me and not to summarise anything, but to discuss concepts, look for strengths and weaknesses in commercial positions, and to sense check ideas. It’s use in this manner is massively under appreciated, by my colleagues also.
Deep thinking / planning mode I also find really useful. You do have to stand your ground, be confident to say no, and not give into the sycophant nature of many LLMs. So its a little bit chicken and egg in that respect, if you don't know what you are doing, you will accept far too much of what it says, if you do have a good knowledge you can challenge and debate.
“Explain X from first principles” is a prompt everyone should use. Absolutely life changing from an educational perspective
I have no applications in my life where I need AI.
Still less, one that I would pay for.
I suspect that I am far, far more reflective of AI than the tech bros who own it.
Rare that we agree and I'm not saying I've never seen AI but it seems a case of using 200 words when 20 will do just fine.
It's a constipation of verbage designed literally to fill a space with nothing, words which I was once told didn't come easy but now do it would seem.
As for the world of work, weren't computers supposedly to allow us to have a life of leisure? Yes, it didn't for me either - indeed, the volume of information created by computer systems is analogous to the amount of verbage created by some AI products.
I warned my successors when I retired - you either manage the information or the information will manage you.
That's the end of the world of work, not through replacement but asphyxiation - we will drown in our information and suffocate on our words simultaenously.
This is a total misunderstanding of what AI can do. It’s much much more than just generating text for publication. I would argue that that’s it’s least useful function.
I’m a solicitor and I use it every day. Not to write anything for me and not to summarise anything, but to discuss concepts, look for strengths and weaknesses in commercial positions, and to sense check ideas. It’s use in this manner is massively under appreciated, by my colleagues also.
Deep thinking / planning mode I also find really useful. You do have to stand your ground, be confident to say no, and not give into the sycophant nature of many LLMs. So its a little bit chicken and egg in that respect, if you don't know what you are doing, you will accept far too much of what it says, if you do have a good knowledge you can challenge and debate.
“Explain X from first principles” is a prompt everyone should use. Absolutely life changing from an educational perspective
Its very important to put them in role play mode by including what role they are to play in the prompt. And then its also useful to rather than say explain x, say you are such and such a person if somebody was to challenge you on this topic ....
I have no applications in my life where I need AI.
Still less, one that I would pay for.
I suspect that I am far, far more reflective of AI than the tech bros who own it.
Rare that we agree and I'm not saying I've never seen AI but it seems a case of using 200 words when 20 will do just fine.
It's a constipation of verbage designed literally to fill a space with nothing, words which I was once told didn't come easy but now do it would seem.
As for the world of work, weren't computers supposedly to allow us to have a life of leisure? Yes, it didn't for me either - indeed, the volume of information created by computer systems is analogous to the amount of verbage created by some AI products.
I warned my successors when I retired - you either manage the information or the information will manage you.
That's the end of the world of work, not through replacement but asphyxiation - we will drown in our information and suffocate on our words simultaenously.
This is a total misunderstanding of what AI can do. It’s much much more than just generating text for publication. I would argue that that’s it’s least useful function.
I’m a solicitor and I use it every day. Not to write anything for me and not to summarise anything, but to discuss concepts, look for strengths and weaknesses in commercial positions, and to sense check ideas. It’s use in this manner is massively under appreciated, by my colleagues also.
Deep thinking / planning mode I also find really useful. You do have to stand your ground, be confident to say no, and not give into the sycophant nature of many LLMs. So its a little bit chicken and egg in that respect, if you don't know what you are doing, you will accept far too much of what it says, if you do have a good knowledge you can challenge and debate.
“Explain X from first principles” is a prompt everyone should use. Absolutely life changing from an educational perspective
Its very important to put them in role play mode by including what role they are to play in the prompt. And then its also useful to rather than say explain x, say you are such and such a person if somebody was to challenge you on this topic ....
I find that’s not required if you use Claude. Especially when you’re only trying to understand a concept. I agree with you though that you do have to know enough about something to guide it and to challenge it. However every model upgrade I find I have to point it back in the right direction less.
I have no applications in my life where I need AI.
Still less, one that I would pay for.
I suspect that I am far, far more reflective of AI than the tech bros who own it.
Rare that we agree and I'm not saying I've never seen AI but it seems a case of using 200 words when 20 will do just fine.
It's a constipation of verbage designed literally to fill a space with nothing, words which I was once told didn't come easy but now do it would seem.
As for the world of work, weren't computers supposedly to allow us to have a life of leisure? Yes, it didn't for me either - indeed, the volume of information created by computer systems is analogous to the amount of verbage created by some AI products.
I warned my successors when I retired - you either manage the information or the information will manage you.
That's the end of the world of work, not through replacement but asphyxiation - we will drown in our information and suffocate on our words simultaenously.
“We will drown.” Maybe. Or maybe tooling shifts toward compression and synthesis rather than raw generation. Historically we oscillate — first we overproduce, then we build better filters.
I do think you’re right about the prolixity though. The cost of producing words has collapsed, and when production becomes trivial, volume explodes. That absolutely risks more noise, not more clarity.
The question isn’t whether more text gets generated — it will. The question is whether we get better at constraining, filtering and compressing it. Used lazily, these tools bloat everything. Used well, they actually reduce cognitive load.
I hope you're right and we can advance through the overproduction stage relatively swiftly though I'm not optimistic. Just as we keep boxes of paperwork for decades after they have ceased to have irrelevance, it may well be we will be retaining the data and verbal information long after it is really required.
I think you’re right that institutions are terrible at deleting things. Storage gets cheaper, risk departments get louder, and nothing ever dies. But that’s slightly orthogonal to whether generation itself is the problem. We’ve been over-retaining since long before LLMs. If anything, the pressure from volume might finally force better lifecycle management rather than endless accumulation.
I have no applications in my life where I need AI.
Still less, one that I would pay for.
I suspect that I am far, far more reflective of AI than the tech bros who own it.
Rare that we agree and I'm not saying I've never seen AI but it seems a case of using 200 words when 20 will do just fine.
It's a constipation of verbage designed literally to fill a space with nothing, words which I was once told didn't come easy but now do it would seem.
As for the world of work, weren't computers supposedly to allow us to have a life of leisure? Yes, it didn't for me either - indeed, the volume of information created by computer systems is analogous to the amount of verbage created by some AI products.
I warned my successors when I retired - you either manage the information or the information will manage you.
That's the end of the world of work, not through replacement but asphyxiation - we will drown in our information and suffocate on our words simultaenously.
This is a total misunderstanding of what AI can do. It’s much much more than just generating text for publication. I would argue that that’s it’s least useful function.
I’m a solicitor and I use it every day. Not to write anything for me and not to summarise anything, but to discuss concepts, look for strengths and weaknesses in commercial positions, and to sense check ideas. It’s use in this manner is massively under appreciated, by my colleagues also.
Deep thinking / planning mode I also find really useful. You do have to stand your ground, be confident to say no, and not give into the sycophant nature of many LLMs. So its a little bit chicken and egg in that respect, if you don't know what you are doing, you will accept far too much of what it says, if you do have a good knowledge you can challenge and debate.
“Explain X from first principles” is a prompt everyone should use. Absolutely life changing from an educational perspective
Its very important to put them in role play mode by including what role they are to play in the prompt. And then its also useful to rather than say explain x, say you are such and such a person if somebody was to challenge you on this topic ....
I find that’s not required if you use Claude. Especially when you’re only trying to understand a concept. I agree with you though that you do have to know enough about something to guide it and to challenge it. However every model upgrade I find I have to point it back in the right direction less.
Its is why I really like Claude Code, you can add in the boring stuff about role playing in md files and also it updates its own files. What I also do is have a running doc of summary of decisions / plans we have made and get it to auto update that on a regular basis. Not only do I have a doc ready to go to show other people but it seems to help Claude keep on track with the goal of what we are trying to achieve. This all seems to minimise the context window / compression issues.
"I mean, no small island like the UK could cope with vast numbers of people coming into the UK.
"I mean, it just overburdens the National Health Service, the traffic service, the police, everybody.
"The country was designed for 55 or 60 million people and we've got 70 million people and all the services break down as a consequence.
Is that why he fled to a country with a population density orders of magnitude higher than the UK?
Monaco was *designed* to accomodate the super rich at high density.
Ah yes, Monaco - famously designed in 1297 by the Grimaldi family with the explicit brief of "let's create a high-density enclave for future hedge fund managers and oligarchs."
I have no applications in my life where I need AI.
Still less, one that I would pay for.
I suspect that I am far, far more reflective of AI than the tech bros who own it.
Fine. Most people don’t “need” a dishwasher either. Or spellcheck. Or GPS. They just quietly use them because they’re convenient.
I'm not anti-technology. I'm anti THIS technology. I think it has been massively overhyped and will not have a tiny fraction of the paid applications that those pumping billions into it expect.
When that sinks in, the dot.com bubble will look like a blip.
Needs a senate vote and Trump to sign it into law so won’t happen but a hopeful sign the madness can be dealt with .
Bit by bit - like eating an Elephant!
I’m so witty sometimes. 🙂
The GOP elephant. Did I have to explain that on here?
Nevermind.
I’m now suffering from a cold shoulder. My mum said it’s an old persons complaint. So I took something for it and nodded off and dreamt I was a composer writing “the farmyard suite” with bits of music for all animals and vegetables, it was very exciting. But I’ve woken up now and find I have no composing ability at all.
The Telegraph has revealed that children, including ones with learning difficulties, have been promised up to £500 in vouchers, which can be redeemed at retailers like XBOX & Uber Eats, if they take part in an NHS 'transgender' trial to block their puberty
Not necessarily. Actually *reading* the article it says...
It can now be revealed that the children taking part in the trial will be incentivised with the promise of up to £500 in vouchers for completing psychometric tests. These will measure the effect of the puberty blockers on their brains, including effects on impulse control and memory...Participants, aged under 16, will receive £30 vouchers for each of the 15 cognitive assessments they complete, as well as £15 vouchers for each of the three MRI scans they undertake, over the two years of the trial...Children in another arm of the trial, not taking the drugs, will be given £15 vouchers for each bone density scan and blood test they have.
Points to note
The children are already in the trial and have been randomised into one of the two arms, so they've already been consented.
Those in the intervention arm (on the drugs) are paid a nominal amount (£15 or £30) for each of the tests (cogntive assessments or MRI scans) they undertake
Those in the control arm (not on the drugs) are paid a nominal amount (£15) for each of the tests (bone density scan and blood test) they undertake
In short the participants are paid nominal sums to take part in tests they have already agreed to prior to assignment, and that this is an incentive to complete the trial, not an incentive to take part in the trial. Since a criticism of many studies in the trans sphere is dropout rate, this is a good idea.
There's a concept in RCT called "equipoise". The surgeons/doctors in a study must be genuinely undecided/unbiased as to which of the two arms is better, and the patients must be equally noncommittal: this enables randomisation to take place believably. In trans studies this is difficult since if the patients believe the study is biased or bent towards one outcome they will refuse to take part or simply bugger off to Europe and get the treatment they want there (Baroness Cass pointed this out as a reason to undertake the PATHWAY studies). The PATHWAY designers seem to have created a study that enables equipoise. The fact that these sums are paid to make sure they stay in the study prevents dropout
(Incidentally @isam, you said last night you wanted to know how tests of drugs in humans take place. You might want to google the term "CTIMP")
Why are people dropping out? If there is a good reason why they are, the trial should be reporting that as part of the data.
They should not be remaining in a trial they would otherwise drop out of, solely due to fiscal considerations.
Dropout rates Many studies (most?) have dropout rates. Yes, dropouts are reported. Sample size calculations often incorporate a 10% or even 20% dropout rate to ensure the end amount are still big enough to provide sufficient statistical power to assess the (usually) two arms.
(Google "dropout rate" if you don't believe me).
Safety Studies are often monitored by one or more external committees, which may combine safety and procedural issues into one committee or separate them into two or more. Those external committees (ie from outside the University or organisation doing the study) will meet at least once a year and often more frequently. If the study is going seriously wrong or not recruiting fast enough then those committees may recommend that the study be terminated, and the people funding the study will also monitor it and may also recommend termination under certain circumstances.
(Google "adverse events" (AEs) and "serious adverse events" (SAEs) if you don't believe me. )
Why do people drop out? People move house, emigrate, simply want to stop, may develop conditions that prevent them from continuing, or simply die by accident or pre-existing conditions or (God forbid) from the intervention or control. They are there voluntarily, not by force.
Endpiece You seem to think that people are forced to take place in studies and studies are just thrown together by amateurs. This is Britain in the 2020s, and studies involving human subjects are regulated, registered and reported. They will be conducted and monitored by extremely skilled individuals and reported in a serious manner.
(Google "CTIMP" if you don't believe me, and that's just one example)
I have no applications in my life where I need AI.
Still less, one that I would pay for.
I suspect that I am far, far more reflective of AI than the tech bros who own it.
Fine. Most people don’t “need” a dishwasher either. Or spellcheck. Or GPS. They just quietly use them because they’re convenient.
I'm not anti-technology. I'm anti THIS technology. I think it has been massively overhyped and will not have a tiny fraction of the paid applications that those pumping billions into it expect.
When that sinks in, the dot.com bubble will look like a blip.
Bubbles and paradigm shifts aren’t mutually exclusive. The dot-com crash was real. So was the internet. Capital can be wildly misallocated and the underlying capability still be transformative. A correction wouldn’t prove AI is useless — it would just reprice it.
"I mean, no small island like the UK could cope with vast numbers of people coming into the UK.
"I mean, it just overburdens the National Health Service, the traffic service, the police, everybody.
"The country was designed for 55 or 60 million people and we've got 70 million people and all the services break down as a consequence.
Is that why he fled to a country with a population density orders of magnitude higher than the UK?
Monaco was *designed* to accomodate the super rich at high density.
Ah yes, Monaco - famously designed in 1297 by the Grimaldi family with the explicit brief of "let's create a high-density enclave for future hedge fund managers and oligarchs."
Being high-density enclaves was kind of the point of fortified medieval cities.
I have no applications in my life where I need AI.
Still less, one that I would pay for.
I suspect that I am far, far more reflective of AI than the tech bros who own it.
Rare that we agree and I'm not saying I've never seen AI but it seems a case of using 200 words when 20 will do just fine.
It's a constipation of verbage designed literally to fill a space with nothing, words which I was once told didn't come easy but now do it would seem.
As for the world of work, weren't computers supposedly to allow us to have a life of leisure? Yes, it didn't for me either - indeed, the volume of information created by computer systems is analogous to the amount of verbage created by some AI products.
I warned my successors when I retired - you either manage the information or the information will manage you.
That's the end of the world of work, not through replacement but asphyxiation - we will drown in our information and suffocate on our words simultaenously.
This is a total misunderstanding of what AI can do. It’s much much more than just generating text for publication. I would argue that that’s it’s least useful function.
I’m a solicitor and I use it every day. Not to write anything for me and not to summarise anything, but to discuss concepts, look for strengths and weaknesses in commercial positions, and to sense check ideas. It’s use in this manner is massively under appreciated, by my colleagues also.
Thanks for putting a contrary view. I'm no longer employed so I don't see AI the way you do.
It seems then AI can function as protagonist and sounding board as well as simple provider of words. In your world, you have case law, precedent, with which to work and from which conclusions can be drawn so it becomes a library as well but not just of facts but of commerical positions.
"I mean, no small island like the UK could cope with vast numbers of people coming into the UK.
"I mean, it just overburdens the National Health Service, the traffic service, the police, everybody.
"The country was designed for 55 or 60 million people and we've got 70 million people and all the services break down as a consequence.
Is that why he fled to a country with a population density orders of magnitude higher than the UK?
Monaco was *designed* to accomodate the super rich at high density.
Ah yes, Monaco - famously designed in 1297 by the Grimaldi family with the explicit brief of "let's create a high-density enclave for future hedge fund managers and oligarchs."
They were river pirates who stole it from ancestors of a friend of mine
Job number one is easing the cost of living pressure that many people still feel. Today’s GDP figures show our economy is growing. That means more money back in your pocket.
CEO of Microsoft AI Mustafa Suleyman joins FT editor Roula Khalaf to explain why most of the tasks accountants, lawyers and other professionals currently undertake will be fully automated by AI within the next 12 to 18 months
I have no applications in my life where I need AI.
Still less, one that I would pay for.
I suspect that I am far, far more reflective of AI than the tech bros who own it.
Fine. Most people don’t “need” a dishwasher either. Or spellcheck. Or GPS. They just quietly use them because they’re convenient.
I'm not anti-technology. I'm anti THIS technology. I think it has been massively overhyped and will not have a tiny fraction of the paid applications that those pumping billions into it expect.
When that sinks in, the dot.com bubble will look like a blip.
This I agree with. LLMs have uses, they're quite effective at certain tasks. But to justify the gargantuan investments being made the scale of use (paid use, mind) would need to be orders of magnitude beyond what is the case today.
A significant section of the population of the developed world would need to subscribe to multiple AI services to ensure any kind of viable return on the amounts being invested, and there's no sign of that happening.
OpenAI probably has less than 18 months left before they run out of other people's money and when they do the crash will be brutal.
"I mean, no small island like the UK could cope with vast numbers of people coming into the UK.
"I mean, it just overburdens the National Health Service, the traffic service, the police, everybody.
"The country was designed for 55 or 60 million people and we've got 70 million people and all the services break down as a consequence.
Is that why he fled to a country with a population density orders of magnitude higher than the UK?
Monaco was *designed* to accomodate the super rich at high density.
Ah yes, Monaco - famously designed in 1297 by the Grimaldi family with the explicit brief of "let's create a high-density enclave for future hedge fund managers and oligarchs."
The Telegraph has revealed that children, including ones with learning difficulties, have been promised up to £500 in vouchers, which can be redeemed at retailers like XBOX & Uber Eats, if they take part in an NHS 'transgender' trial to block their puberty
Did somebody saaaaayyy Women Just Can't Have a Penis?
CEO of Microsoft AI Mustafa Suleyman joins FT editor Roula Khalaf to explain why most of the tasks accountants, lawyers and other professionals currently undertake will be fully automated by AI within the next 12 to 18 months
We're gonna have a super difficult time deciding if AI exhibits consciousness. Since we have no agreement as to what it is.
Consciousness isn't something exhibited; if it were it would be simple to spot. Actually I think we are pretty much agreed as to what it is, it being the one thing Descartes says we cannot be mistaken about. What we have no idea of is by what means it exists. Either it is a property of matter in certain configuration, in which case it is mystically mysterious, or it isn't a property of matter, in which case it is mystically mysterious.
As to AI and consciousness, Hume went for the view, unfashionable at the time, that it was just obvious from the conduct of at least some animals that they were conscious. Ask any dog lover. AI might test this thought to destruction.
The Bar Council has decided to have a Commissioner for Conduct following a report by Harriet Harman into sexism, misogyny and other bad behaviour. It has decided to appoint former Tory MP Maria Miller to this role. She is the Culture Secretary in Cameron's government who had to resign following a report showing that she had claimed ten of thousands of pounds of expenses to which she was not entitled. She then gave a 32 second apology to the Commons, was widely criticised for this and had to apologise again, not just for the expenses claim but also for her failure to co-operate with the investigation.
And this was the best candidate the Bar Council could come up with?
There has been no transparency to practising subscriber barristers about the process or terms of appointment of Miller or, come to that, Harman.
Structural dishonesty and tolerance of it are big problems in our country.
I mean, WTF? Is there no end to this failing upwards? This is shocking. What is the point in trying to play by the rules? You get no thanks and no recognition.
The Telegraph has revealed that children, including ones with learning difficulties, have been promised up to £500 in vouchers, which can be redeemed at retailers like XBOX & Uber Eats, if they take part in an NHS 'transgender' trial to block their puberty
Not necessarily. Actually *reading* the article it says...
It can now be revealed that the children taking part in the trial will be incentivised with the promise of up to £500 in vouchers for completing psychometric tests. These will measure the effect of the puberty blockers on their brains, including effects on impulse control and memory...Participants, aged under 16, will receive £30 vouchers for each of the 15 cognitive assessments they complete, as well as £15 vouchers for each of the three MRI scans they undertake, over the two years of the trial...Children in another arm of the trial, not taking the drugs, will be given £15 vouchers for each bone density scan and blood test they have.
Points to note
The children are already in the trial and have been randomised into one of the two arms, so they've already been consented.
Those in the intervention arm (on the drugs) are paid a nominal amount (£15 or £30) for each of the tests (cogntive assessments or MRI scans) they undertake
Those in the control arm (not on the drugs) are paid a nominal amount (£15) for each of the tests (bone density scan and blood test) they undertake
In short the participants are paid nominal sums to take part in tests they have already agreed to prior to assignment, and that this is an incentive to complete the trial, not an incentive to take part in the trial. Since a criticism of many studies in the trans sphere is dropout rate, this is a good idea.
There's a concept in RCT called "equipoise". The surgeons/doctors in a study must be genuinely undecided/unbiased as to which of the two arms is better, and the patients must be equally noncommittal: this enables randomisation to take place believably. In trans studies this is difficult since if the patients believe the study is biased or bent towards one outcome they will refuse to take part or simply bugger off to Europe and get the treatment they want there (Baroness Cass pointed this out as a reason to undertake the PATHWAY studies). The PATHWAY designers seem to have created a study that enables equipoise. The fact that these sums are paid to make sure they stay in the study prevents dropout
(Incidentally @isam, you said last night you wanted to know how tests of drugs in humans take place. You might want to google the term "CTIMP")
The “trial” is totally unethical by any conventional medical standard.
...and yet it passed the ethics committee.
On a wider note, there's an issue of legality here. Often people insist that studies on this subject should not be undertaken, especially on children. If this is society's settled view then the proper procedure is to make it illegal, as I have said to @AnneJGP more than once. But if it is legal (and the ethics committee thinks it's ethical) then somebody somewhere will do a study and that study will (if in the UK) be properly monitored.
Most people's objections have been expressed in sarcasm ("trial" in scare quotes) or inchoate fears ("why are they dropping out?") but that's not how medicine or statistics works. It may be how Britain works, and the full panoply of British elite disapproval (judicial reviews, sensationalist reporting, petitions, partisan experts) is being deployed to stop it. Given that LOTO would crawl over broken glass to stop it and she may be PM in 2029 I wouldn't be surprised if it is prematurely terminated, which is the worst of both worlds.
Angela Rayner is currently making some huge attacks on Rachel Reeves over business rates, energy bills and hospitality VAT.
Speaking at the nighttime economy summit in Liverpool, Rayner also appears to be doing some major pro-business positioning ahead of the inevitable leadership contest.
She says: "Confidence in politics matters. Businesses need to believe they will be treated fairly. That the rules won't shift without warning. That the long-standing structural issues will finally be addressed, not deferred again."
She says the government needs to drop ideology and be pragmatic to help businesses 👀
She is right though and sensible take on business
Meh. It's words. Does anyone seriously think Rayner's political instincts, or the interest groups she intends to serve, are pro-business ?
She pioneered the so-called workers rights bill at the behest of the Unions. A few token watering down of a couple of its elements doesn’t change that.
People might be missing the point. The most important part of what Rayner said is about stability: That the rules won't shift without warning. By and large, businesses (and people) can cope with or manage around policies they don't like, but they can't handle repeated, arbitrary lurches.
Labour politics-wise, this might be a repeat of the prawn cocktail offensive but more likely she has been talking to Andy Burnham.
She's more beholden to left wing ideology and has more populist instincts than Reeves (or Starmer, to the extent one can detect any coherent political philosophy or instincts in him at all). That is a recipe for more tax on business and more arbitrary changes, not less.
I like her, I'd want her fighting my corner if I was in her client group, but I'm not.
The focus on energy prices is an interesting one. Those are pretty much back at their 2021 prices in real terms this year, and the Govt will meet their manifesto reduction pledge.
If Starmer gets his f*cking comms sorted out (a very big if), it is potentially a winning issue.
(Forecast is 2026, not 2025. I looked over the numbers.)
prices have not dropped , they just increase for all Milliband's crap
Job number one is easing the cost of living pressure that many people still feel. Today’s GDP figures show our economy is growing. That means more money back in your pocket.
FFS. 0.1%. Has the man no shame at all? The only reason we have not had a recession over the last 6 months is because the government is spending money like it is going out of fashion. The fact that a boost to demand of £150bn a year can produce 0.1% for 2 quarters in a row is bordering on catastrophic. We simply cannot sustain a flat line on current policies. It is unaffordable.
The people who were purely “code monkeys” — i.e. implementing tickets with minimal understanding of the system — were already in a fragile position. AI just makes that fragility obvious. But that role was brittle long before LLMs showed up. It was vulnerable to offshoring, automation, better frameworks, you name it.
Indeed. The truth is a lot of people currently making a living writing code are not particularly skilled or valuable, the work they do is not hard. Some of them will either be replaced by LLMs or demoted to just bug-fixing the output.
I write mostly Verilog and assembly language and design much of the hardware that hosts my code. I do not expect to be replaced by an LLM any time in the foreseeable future. But if I spent my days writing Python or maintaining some shonky mobile app, I'd be worried.
I struck out on my own as a lawyer about 25 years ago, at a time when my bosses were still struggling with basic concepts like click and drag. I could see an opportunity to commoditise and automate a lot of my drafting work: essentially grab some data about parties and dates and so on from an existing document, match the document against a database of known document types which then told you what that document would say (and therefore what the new document I was providing should say) about issues x,y and z, make all of that just a row in an xls spreadsheet, feed that data into word templates with mergefields including lots of complex conditional fields that produce the right legal wording when fed the relevant row of data, then I learnt scrappy VBA and wrote short macros to automate the merging and saving and printing and the generating of related invoices and emails. As data was only being manually inputted once, and as the way it was recorded meant the most common typing errors could be global search and replaced or avoided by dropdown lists or whatever, I ended up doing the work maybe 5x faster than other lawyers with a lower error rate. What was my point ? Erm, yes, I think the real value, now as then, is in combining an understanding of the actual work to be done with an understanding of what can and can't be reliably and cleanly automated.
Comments
https://x.com/EdwardJDavey/status/2022005105336451194?s=20
I use them all the time to build disposable, throw away, apps that I host that do one specific thing for me.
Are they "conscious"? Well, given their limited context window, and inability to learn skills outside said context window (see attempts to get LLMs to build things in Z80 assembly language), then I'd go with "no".
But if you're not using them today, you are making your life much harder than it need be.
And this was the best candidate the Bar Council could come up with?
There has been no transparency to practising subscriber barristers about the process or terms of appointment of Miller or, come to that, Harman.
Structural dishonesty and tolerance of it are big problems in our country.
I’m a solicitor and I use it every day. Not to write anything for me and not to summarise anything, but to discuss concepts, look for strengths and weaknesses in commercial positions, and to sense check ideas. It’s use in this manner is massively under appreciated, by my colleagues also.
It's more efficient to apply for a job with AI, it's more efficient to screen applications by AI. Combine the two and the system collapses.
Since we have no agreement as to what it is.
I’m so witty sometimes. 🙂
BREAKING
A trade union leader has called for Keir Starmer to be replaced by Angela Rayner, saying it’s time Labour was led by a woman.
Maryam Eslamdoust, the gen sec of the TSSA, said Rayner speaks in a way that "resonates" with the public and would stand up to Donald Trump.
She also predicted the Prime Minister could be toppled in a fortnight’s time if Labour comes third in the Gorton and Denton by-election.
But that’s slightly orthogonal to whether generation itself is the problem. We’ve been over-retaining since long before LLMs. If anything, the pressure from volume might finally force better lifecycle management rather than endless accumulation.
When that sinks in, the dot.com bubble will look like a blip.
Nevermind.
I’m now suffering from a cold shoulder. My mum said it’s an old persons complaint. So I took something for it and nodded off and dreamt I was a composer writing “the farmyard suite” with bits of music for all animals and vegetables, it was very exciting. But I’ve woken up now and find I have no composing ability at all.
I never want to see him in his Y-fronts again.
Many studies (most?) have dropout rates. Yes, dropouts are reported. Sample size calculations often incorporate a 10% or even 20% dropout rate to ensure the end amount are still big enough to provide sufficient statistical power to assess the (usually) two arms.
(Google "dropout rate" if you don't believe me).
Safety
Studies are often monitored by one or more external committees, which may combine safety and procedural issues into one committee or separate them into two or more. Those external committees (ie from outside the University or organisation doing the study) will meet at least once a year and often more frequently. If the study is going seriously wrong or not recruiting fast enough then those committees may recommend that the study be terminated, and the people funding the study will also monitor it and may also recommend termination under certain circumstances.
(Google "adverse events" (AEs) and "serious adverse events" (SAEs) if you don't believe me. )
Why do people drop out?
People move house, emigrate, simply want to stop, may develop conditions that prevent them from continuing, or simply die by accident or pre-existing conditions or (God forbid) from the intervention or control. They are there voluntarily, not by force.
Endpiece
You seem to think that people are forced to take place in studies and studies are just thrown together by amateurs. This is Britain in the 2020s, and studies involving human subjects are regulated, registered and reported. They will be conducted and monitored by extremely skilled individuals and reported in a serious manner.
(Google "CTIMP" if you don't believe me, and that's just one example)
Capital can be wildly misallocated and the underlying capability still be transformative. A correction wouldn’t prove AI is useless — it would just reprice it.
It seems then AI can function as protagonist and sounding board as well as simple provider of words. In your world, you have case law, precedent, with which to work and from which conclusions can be drawn so it becomes a library as well but not just of facts but of commerical positions.
Interesting.
Job number one is easing the cost of living pressure that many people still feel. Today’s GDP figures show our economy is growing. That means more money back in your pocket.
https://x.com/Keir_Starmer/status/2021861357591916926?s=20
https://x.com/FT/status/2021913057065160828?s=20
NEW THREAD
A significant section of the population of the developed world would need to subscribe to multiple AI services to ensure any kind of viable return on the amounts being invested, and there's no sign of that happening.
OpenAI probably has less than 18 months left before they run out of other people's money and when they do the crash will be brutal.
About as likely as self-driving cars being widely adopted within the next 2 years back in 2015, I'd say.
As to AI and consciousness, Hume went for the view, unfashionable at the time, that it was just obvious from the conduct of at least some animals that they were conscious. Ask any dog lover. AI might test this thought to destruction.
On a wider note, there's an issue of legality here. Often people insist that studies on this subject should not be undertaken, especially on children. If this is society's settled view then the proper procedure is to make it illegal, as I have said to @AnneJGP more than once. But if it is legal (and the ethics committee thinks it's ethical) then somebody somewhere will do a study and that study will (if in the UK) be properly monitored.
Most people's objections have been expressed in sarcasm ("trial" in scare quotes) or inchoate fears ("why are they dropping out?") but that's not how medicine or statistics works. It may be how Britain works, and the full panoply of British elite disapproval (judicial reviews, sensationalist reporting, petitions, partisan experts) is being deployed to stop it. Given that LOTO would crawl over broken glass to stop it and she may be PM in 2029 I wouldn't be surprised if it is prematurely terminated, which is the worst of both worlds.