Sudden hunch that the Final will be France V Argentina
Mbappe V Messi, the two best players in the world. That would be quite a final
Sudden hunch? That was my tip before it began.
Argentina did not look good vs Australia.
Who looks better in that half? Brazil have been underwhelming. Holland uninspiring from the softest of groups. Japan? An ageing Croatia? Korea? That's it.
I've spent the last 48 hours using the gpt chatbot in a variety of instances. It has written 5000 word spec scripts for me, analysed business problems, I've even fed it enough information about myself to psychoanalyse me (by creating a psychiatrist "character" and a "me" character), and had it write a stream of consciousness about what I might be thinking. It was accurate. I told it the project I am working on at the moment and asked it to create a slide deck outline. I then asked it to fill each individual slide with text - it was accurate. I asked it to use UK law to create a will for me. Done. I asked it to negotiate a more complex corporate contract I was involved in - a few errors, but it was largely correct.
I cannot even begin to comprehend the number of jobs this either erases, or shakes to their very core. A lot of white collar jobs won't exist within a couple of years.
As with Stable Diffusion, promptcraft is everything. But the difference between stable diffusion and this language model is that SD only creates pretty images. With the right prompt, GPT3.5 can be you psychatrist, your lover, your confidant, or even you. It can analyse you your darkest desires, explain to you in an academic essay the origins of your fears and phobias,or have a dialogue with itself about whether or not the "AI you" you have created is a real person. It can write a novel from your perspective as a child, about your childhood, with only a little prompting.
This tech is terrifying, and is light years beyond stable diffusion making pretty pictures. It literally changes everything, and I think as more and more people play with the tech, they are going to realise what an earthquake this is.
When we last spoke, I was using the chatbot as a google search, evaluating its answers to individual questions. This is useless - the way to use the chatbot is to interact with it enough to build a world it can learn from and respond to. Once you do that - as I say, it learns fast and effectively becomes "human". While it may have no inner sense of self, the question is, if you can't tell the difference (and I as a human can't tell the difference between its output and a real human, once it is trained well enough), does it matter?
So it can copy-paste things you can Google, and it can do a bad job at law. It sounds like as happened with computers, it will create lots of jobs about interacting with it, but would take a bit longer to fully replace analytical jobs where the world isn't pre-defined as you put it. As for the book trade it is already about selling personalities from traditional and social media.
https://twitter.com/GLandsbergis/status/1599441986288656385 Why aren’t we sending Ukraine all the tools needed to end the war? Why are we avoiding specifics during the debate on Ukraine‘s membership of NATO? Because the belief is still alive that after the war we can return to business as usual, as if February 24th was just a glitch…
… Our strategy needs to be rethought and reflected in a new security architecture that will ensure the safety of the continent for decades to come. And we should start creating this new system with Ukraine, not with Putin’s Russia.
I DO remember the Winter of Discontent which was also a very snowy and cold winter. It was pretty awful. But TSE @TheScreamingEagles says this:
"What makes this period of industrial strife feel different is the sheer number of professions on strike, including the truly best of society, such as nurses and lawyers, without who society would struggle to function."
I think that's absolutely spot on. In 1978/9 there was a militancy, almost a wrecking ball, by the unions. THIS time it really does feel much more as if it's largely decent people wanting a wage that keeps step with inflation.
And the tories who encouraged the nation to clap our NHS workers have got to be extremely careful here ...
Apparently they talk of little else at the Kremlin other than real terms paycuts for our nurses.
I've spent the last 48 hours using the gpt chatbot in a variety of instances. It has written 5000 word spec scripts for me, analysed business problems, I've even fed it enough information about myself to psychoanalyse me (by creating a psychiatrist "character" and a "me" character), and had it write a stream of consciousness about what I might be thinking. It was accurate. I told it the project I am working on at the moment and asked it to create a slide deck outline. I then asked it to fill each individual slide with text - it was accurate. I asked it to use UK law to create a will for me. Done. I asked it to negotiate a more complex corporate contract I was involved in - a few errors, but it was largely correct.
I cannot even begin to comprehend the number of jobs this either erases, or shakes to their very core. A lot of white collar jobs won't exist within a couple of years.
As with Stable Diffusion, promptcraft is everything. But the difference between stable diffusion and this language model is that SD only creates pretty images. With the right prompt, GPT3.5 can be you psychatrist, your lover, your confidant, or even you. It can analyse you your darkest desires, explain to you in an academic essay the origins of your fears and phobias,or have a dialogue with itself about whether or not the "AI you" you have created is a real person. It can write a novel from your perspective as a child, about your childhood, with only a little prompting.
This tech is terrifying, and is light years beyond stable diffusion making pretty pictures. It literally changes everything, and I think as more and more people play with the tech, they are going to realise what an earthquake this is.
When we last spoke, I was using the chatbot as a google search, evaluating its answers to individual questions. This is useless - the way to use the chatbot is to interact with it enough to build a world it can learn from and respond to. Once you do that - as I say, it learns fast and effectively becomes "human". While it may have no inner sense of self, the question is, if you can't tell the difference (and I as a human can't tell the difference between its output and a real human, once it is trained well enough), does it matter?
So it can copy-paste things you can Google, and it can do a bad job at law. It sounds like as happened with computers, it will create lots of jobs about interacting with it, but would take a bit longer to fully replace analytical jobs where the world isn't pre-defined as you put it. As for the book trade it is already about selling personalities from traditional and social media.
It does not copy paste. I used it to create a therapist, who asked me questions and then delivered a shockingly accurate insight about my life, using things I hadn't told it.
I then asked it to deliver a stream of consciousness about what I might be thinking, and it was also shockingly accurate. Neither my state of mind, nor my case history, are searchable via google. So it was using deductive reasoning, at the very least. It is not simply regurgitating answers from a search engine.
I advise you to spend a day with the AI and understand its strengths and limitations - After nearly three days of interacting with it, I don't believe it is conscious, but it does have the power to deduce things and reason its own arguments.
The car is not merely a better horse - it represented a sea change in technology, the end of one era and the beginning of the next. After spending a few days with the AI I am convinced we are standing at a similar sea change.
(Continuing my conversation with Leon from Friday night where I was sceptical about whether or not this AI would actually change anything - as you can see, it's taken just a couple of days before I'm converted).
Aaron has confirmed that he will contest the next general election.
His inevitable loss will be a disappointment.
He gets a pay off if he loses, a seat in the Opposition benches in the unlikely event of a win, and young enough to help reshape the party away from the Johnson omnishambles.
I've spent the last 48 hours using the gpt chatbot in a variety of instances. It has written 5000 word spec scripts for me, analysed business problems, I've even fed it enough information about myself to psychoanalyse me (by creating a psychiatrist "character" and a "me" character), and had it write a stream of consciousness about what I might be thinking. It was accurate. I told it the project I am working on at the moment and asked it to create a slide deck outline. I then asked it to fill each individual slide with text - it was accurate. I asked it to use UK law to create a will for me. Done. I asked it to negotiate a more complex corporate contract I was involved in - a few errors, but it was largely correct.
I cannot even begin to comprehend the number of jobs this either erases, or shakes to their very core. A lot of white collar jobs won't exist within a couple of years.
As with Stable Diffusion, promptcraft is everything. But the difference between stable diffusion and this language model is that SD only creates pretty images. With the right prompt, GPT3.5 can be you psychatrist, your lover, your confidant, or even you. It can analyse you your darkest desires, explain to you in an academic essay the origins of your fears and phobias,or have a dialogue with itself about whether or not the "AI you" you have created is a real person. It can write a novel from your perspective as a child, about your childhood, with only a little prompting.
This tech is terrifying, and is light years beyond stable diffusion making pretty pictures. It literally changes everything, and I think as more and more people play with the tech, they are going to realise what an earthquake this is.
When we last spoke, I was using the chatbot as a google search, evaluating its answers to individual questions. This is useless - the way to use the chatbot is to interact with it enough to build a world it can learn from and respond to. Once you do that - as I say, it learns fast and effectively becomes "human". While it may have no inner sense of self, the question is, if you can't tell the difference (and I as a human can't tell the difference between its output and a real human, once it is trained well enough), does it matter?
I've been playing with it quite a lot this weekend. It is a very impressive tool. These past few months I've been changing an application from one underlying framework to another - tedious, time-consuming, fraught with little annoyances. Gave it a long-ish snippet and asked it 'Could you convert this code from framework-X to framework-Y?' and about 20 seconds later it was done. And all correct from a quick read-through.
Asked it to show me how to do a 'silent install' of Matlab. No problem. It also mentioned I could pass an input file with custom options. So I asked it for further details. Nice easy-to-read explanation with an example to illustrate it. Then asked it how I'd put that all together to make it available via Microsoft's SCCM software distribution system - 20 seconds later it was done.
Same story over and over for technical tasks.
It did make me wonder what the future is for new-starts in technical (or similar) roles. The value proposition will change. A big part of learning the ropes is doing the grunt work, figuring out what works, how to organise things, how to give things sensible names. Do we skip to 'level 2' with new hires rather than letting those skills develop over a few years? I'm not even sure how to do that.
Though at one point I asked it 'What was the allure of Liz Truss?' and it got quite shirty with me.
I DO remember the Winter of Discontent which was also a very snowy and cold winter. It was pretty awful. But TSE @TheScreamingEagles says this:
"What makes this period of industrial strife feel different is the sheer number of professions on strike, including the truly best of society, such as nurses and lawyers, without who society would struggle to function."
I think that's absolutely spot on. In 1978/9 there was a militancy, almost a wrecking ball, by the unions. THIS time it really does feel much more as if it's largely decent people wanting a wage that keeps step with inflation.
And the tories who encouraged the nation to clap our NHS workers have got to be extremely careful here ...
Apparently they talk of little else at the Kremlin other than real terms paycuts for our nurses.
Comments
(I think they might well do, actually. It's a seat that's been moving rightwards for a long time.)
This would be a whole other task with those two.
https://twitter.com/GLandsbergis/status/1599441986288656385
Why aren’t we sending Ukraine all the tools needed to end the war? Why are we avoiding specifics during the debate on Ukraine‘s membership of NATO? Because the belief is still alive that after the war we can return to business as usual, as if February 24th was just a glitch…
… Our strategy needs to be rethought and reflected in a new security architecture that will ensure the safety of the continent for decades to come. And we should start creating this new system with Ukraine, not with Putin’s Russia.
https://twitter.com/AdamBienkov/status/1599335366992875520?t=wo1_86a9dEZqz7rhVuW_Hw&s=19
I then asked it to deliver a stream of consciousness about what I might be thinking, and it was also shockingly accurate. Neither my state of mind, nor my case history, are searchable via google. So it was using deductive reasoning, at the very least. It is not simply regurgitating answers from a search engine.
I advise you to spend a day with the AI and understand its strengths and limitations - After nearly three days of interacting with it, I don't believe it is conscious, but it does have the power to deduce things and reason its own arguments.
The car is not merely a better horse - it represented a sea change in technology, the end of one era and the beginning of the next. After spending a few days with the AI I am convinced we are standing at a similar sea change.
(Continuing my conversation with Leon from Friday night where I was sceptical about whether or not this AI would actually change anything - as you can see, it's taken just a couple of days before I'm converted).
NEW THREAD
Asked it to show me how to do a 'silent install' of Matlab. No problem. It also mentioned I could pass an input file with custom options. So I asked it for further details. Nice easy-to-read explanation with an example to illustrate it. Then asked it how I'd put that all together to make it available via Microsoft's SCCM software distribution system - 20 seconds later it was done.
Same story over and over for technical tasks.
It did make me wonder what the future is for new-starts in technical (or similar) roles. The value proposition will change. A big part of learning the ropes is doing the grunt work, figuring out what works, how to organise things, how to give things sensible names. Do we skip to 'level 2' with new hires rather than letting those skills develop over a few years? I'm not even sure how to do that.
Though at one point I asked it 'What was the allure of Liz Truss?' and it got quite shirty with me.
Fuck off.
The last one somewhat less so...