"Q: How many votes did John Major receive in the Huntingdon constituency at the 1987 general election?
A: In the 1987 general election, John Major, representing the Conservative Party, received 38,904 votes in the Huntingdon constituency."
Wrong, as usual. It's like someone who tries their best to estimate it and hopes no-one notices the mistake because the person they're answering to is too busy doing something else.
Its why Claude is far better...
"While I aim to be precise with numbers, for this specific detail about John Major's exact vote count in the Huntingdon constituency during the 1987 general election, I should be transparent that I could hallucinate specific numbers. While I know Major represented Huntingdon and won the seat comfortably in 1987, I don't feel confident providing the exact vote count without being able to verify it."
"To get the exact verified count, I'd recommend checking official UK electoral records or parliamentary archives."
Yup. An LLM is an LLM: nothing more. It’s a good party trick but it’s not “artificial intelligence” in any meaningful sense.
I would say its more than a party trick, they are lots of genuinely useful things that they aid with e.g. coding, but it is a just another tool. What I wouldn't want to be doing is some sort of job that is text or coding heavily that is overwhelmingly a very repetitive task with little innovation, because tools can certainly be produced that will then only require 1 of you rather than current 5 or 10.
Even for coding - they can help you finish something off (but my word, you better check the output) - but they cannot innovate or problem solve.
Have you tried the full-fat o1 model? If you give it the context it's really quite good on the problem solving. Sadly. I wrote it off originally while giving it similar prompts to gpt-*, but once you realise it needs cajoled in a different way it's really quite good. Again, sadly.
It's worth a quick play with the 'QwQ-32B-Preview' model to see how even an open-source 'free' version is working through a problem using a similar paradigm.
"Q: How many votes did John Major receive in the Huntingdon constituency at the 1987 general election?
A: In the 1987 general election, John Major, representing the Conservative Party, received 38,904 votes in the Huntingdon constituency."
Wrong, as usual. It's like someone who tries their best to estimate it and hopes no-one notices the mistake because the person they're answering to is too busy doing something else.
Its why Claude is far better...
"While I aim to be precise with numbers, for this specific detail about John Major's exact vote count in the Huntingdon constituency during the 1987 general election, I should be transparent that I could hallucinate specific numbers. While I know Major represented Huntingdon and won the seat comfortably in 1987, I don't feel confident providing the exact vote count without being able to verify it."
"To get the exact verified count, I'd recommend checking official UK electoral records or parliamentary archives."
"I don't know" would be clearer. AI seems to major in TLDR quite a lot of the time ?
It is negative with the current Claude model, that it is very verbose. They have introduced "modes" which are obviously some extra tokens that are appended to direct for instance to be more concise, but I have found they can lead to the above guard rails breaking.
AI is fun, but it has a long way to go before it can be trusted or relied upon. Andy’s example above is a good example of its failings. Sadly, it does appear to have a fatal flaw insofar as it uses other AIs’ outputs as its source, thus amplifying errors. The fear is that this flaw might not be easily solvable.
"In this profound keynote, Vector co-founder Geoffrey Hinton explores the philosophical implications of artificial intelligence and its potential to surpass human intelligence. Drawing from decades of expertise, Hinton shares his growing concerns about AI's existential risks while examining fundamental questions about consciousness, understanding, and the nature of intelligence itself."
(He makes some quite good (academic) jokes. So it's not all doom'n'gloom)
"Q: How many votes did John Major receive in the Huntingdon constituency at the 1987 general election?
A: In the 1987 general election, John Major, representing the Conservative Party, received 38,904 votes in the Huntingdon constituency."
Wrong, as usual. It's like someone who tries their best to estimate it and hopes no-one notices the mistake because the person they're answering to is too busy doing something else.
Its why Claude is far better...
"While I aim to be precise with numbers, for this specific detail about John Major's exact vote count in the Huntingdon constituency during the 1987 general election, I should be transparent that I could hallucinate specific numbers. While I know Major represented Huntingdon and won the seat comfortably in 1987, I don't feel confident providing the exact vote count without being able to verify it."
"To get the exact verified count, I'd recommend checking official UK electoral records or parliamentary archives."
I dislike the word hallucinate for what LLMs do with numbers in that situation. Hallucinate gives an air of imagination to what is more accurately termed bullshitting. It's to the credit of Claude's creators that they've blocked it off from bullshitting about some forms of factual information, it would be more impressive if it was able to self-correct.
I've no idea what the state of research on this question is. It feels like something conceptually very different to what LLMs are doing.
I was surprised to learn despite all the hype, OpenAI only has 10 million paying customers despite it currently being very cheap to subscribe. I presume Anthropic will be at an least order of magnitude less than that as nobody outside of particularly geeky people are aware of it as they don't do the PR hype thing.
I feel like "only" is doing some quite hard work there. I suspect they are also playing a line between acquiring paid customers and fulfilling their non-profit "for the benefit of all" claims. At least until the current spat with Musk is played out.
We've also no insight into how much their API via their own service or MS Azure is being used.
Wouldn't surprise me at all if they are burning cash like firelighters. But it's quite opaque as to how their finances are going.
They are losing $5bn a year at the moment.
The "non-profit" angle is basically dead now they have taken such a massive amount of Big Daddy Microsoft money (and between them and Nvidia, they are absolutely beholden to them to keep going) and large numbers of the high ups in AI safety team have left.
"Q: How many votes did John Major receive in the Huntingdon constituency at the 1987 general election?
A: In the 1987 general election, John Major, representing the Conservative Party, received 38,904 votes in the Huntingdon constituency."
Wrong, as usual. It's like someone who tries their best to estimate it and hopes no-one notices the mistake because the person they're answering to is too busy doing something else.
Its why Claude is far better...
"While I aim to be precise with numbers, for this specific detail about John Major's exact vote count in the Huntingdon constituency during the 1987 general election, I should be transparent that I could hallucinate specific numbers. While I know Major represented Huntingdon and won the seat comfortably in 1987, I don't feel confident providing the exact vote count without being able to verify it."
"To get the exact verified count, I'd recommend checking official UK electoral records or parliamentary archives."
Yup. An LLM is an LLM: nothing more. It’s a good party trick but it’s not “artificial intelligence” in any meaningful sense.
I'm a bit of a believer that we'll get there eventually (as a species, not the UK, obviously), or get to something which we cannot really distinguish from AI even if it is not really, but a surprising number of people have been a bit quick to believe the hype of the various companies desperate to rake in investment capital.
I get the investors doing so even with some doubts, if one of these companies hits gold it will be worth it, but a lot of people just believe the output of these LLMs uncritically despite the well known hallucination issues, never mind anything.
I have issues with even using the word “hallucinate”. I think it makes people imagine something is happening which is not.
Indeed. It’s just garden-variety error-returns. Nothing to do with hallucination - which implies some thinking is happening.
"Q: How many votes did John Major receive in the Huntingdon constituency at the 1987 general election?
A: In the 1987 general election, John Major, representing the Conservative Party, received 38,904 votes in the Huntingdon constituency."
Wrong, as usual. It's like someone who tries their best to estimate it and hopes no-one notices the mistake because the person they're answering to is too busy doing something else.
Its why Claude is far better...
"While I aim to be precise with numbers, for this specific detail about John Major's exact vote count in the Huntingdon constituency during the 1987 general election, I should be transparent that I could hallucinate specific numbers. While I know Major represented Huntingdon and won the seat comfortably in 1987, I don't feel confident providing the exact vote count without being able to verify it."
"To get the exact verified count, I'd recommend checking official UK electoral records or parliamentary archives."
I dislike the word hallucinate for what LLMs do with numbers in that situation. Hallucinate gives an air of imagination to what is more accurately termed bullshitting. It's to the credit of Claude's creators that they've blocked it off from bullshitting about some forms of factual information, it would be more impressive if it was able to self-correct.
I've no idea what the state of research on this question is. It feels like something conceptually very different to what LLMs are doing.
The Hinton video I posted goes into the hallucination problem. It's really worth a 'save for later' over the holidays. Quite cantankerous at points in an old-school professorial way.
I was surprised to learn despite all the hype, OpenAI only has 10 million paying customers despite it currently being very cheap to subscribe. I presume Anthropic will be at an least order of magnitude less than that as nobody outside of particularly geeky people are aware of it as they don't do the PR hype thing.
I feel like "only" is doing some quite hard work there. I suspect they are also playing a line between acquiring paid customers and fulfilling their non-profit "for the benefit of all" claims. At least until the current spat with Musk is played out.
We've also no insight into how much their API via their own service or MS Azure is being used.
Wouldn't surprise me at all if they are burning cash like firelighters. But it's quite opaque as to how their finances are going.
They are losing $5bn a year at the moment.
The "non-profit" angle is basically dead now they have taken such a massive amount of Big Daddy Microsoft money (and between them and Nvidia, they are absolutely beholden to them to keep going) and large numbers of the high ups in AI safety team have left.
I've seen various numbers for them losing/making money. But they are still a private company so we've really no idea.
Wouldn't at all surprise me if they were burning VC/MS/Whoever cash like it was going out of fashion, but that's SV startups for you.
I was surprised to learn despite all the hype, OpenAI only has 10 million paying customers despite it currently being very cheap to subscribe. I presume Anthropic will be at an least order of magnitude less than that as nobody outside of particularly geeky people are aware of it as they don't do the PR hype thing.
I feel like "only" is doing some quite hard work there. I suspect they are also playing a line between acquiring paid customers and fulfilling their non-profit "for the benefit of all" claims. At least until the current spat with Musk is played out.
We've also no insight into how much their API via their own service or MS Azure is being used.
Wouldn't surprise me at all if they are burning cash like firelighters. But it's quite opaque as to how their finances are going.
They are losing $5bn a year at the moment.
The "non-profit" angle is basically dead now they have taken such a massive amount of Big Daddy Microsoft money (and between them and Nvidia, they are absolutely beholden to them to keep going) and large numbers of the high ups in AI safety team have left.
I've seen various numbers for them losing/making money. But they are still a private company so we've really no idea.
Wouldn't at all surprise me if they were burning VC/MS/Whoever cash like it was going out of fashion, but that's SV startups for you.
The headline financials got leaked to various media outlets from info they gave in meetings to potential investors. Also some internet sleuths have crunched numbers to see its sounds right ballpark wise given known user numbers and costs and its all adds up.
One huge big problem they are blasting GPU / data centres so hard, that equipment will need constant replacement. Its a bit like Bitcoin mining in that respect, you haven't bought yourself 100k GPUs and that is cap-ex investment for the next 10 years, they are going through them fast. That is why a lot of work on more efficient inference.
"Q: How many votes did John Major receive in the Huntingdon constituency at the 1987 general election?
A: In the 1987 general election, John Major, representing the Conservative Party, received 38,904 votes in the Huntingdon constituency."
Wrong, as usual. It's like someone who tries their best to estimate it and hopes no-one notices the mistake because the person they're answering to is too busy doing something else.
Its why Claude is far better...
"While I aim to be precise with numbers, for this specific detail about John Major's exact vote count in the Huntingdon constituency during the 1987 general election, I should be transparent that I could hallucinate specific numbers. While I know Major represented Huntingdon and won the seat comfortably in 1987, I don't feel confident providing the exact vote count without being able to verify it."
"To get the exact verified count, I'd recommend checking official UK electoral records or parliamentary archives."
Yup. An LLM is an LLM: nothing more. It’s a good party trick but it’s not “artificial intelligence” in any meaningful sense.
I would say its more than a party trick, they are lots of genuinely useful things that they aid with e.g. coding, but it is a just another tool. What I wouldn't want to be doing is some sort of job that is text or coding heavily that is overwhelmingly a very repetitive task with little innovation, because tools can certainly be produced that will then only require 1 of you rather than current 5 or 10.
They code like a junior engineer. The output needs checking and can in most cases be optimised.
They are best as an efficiency booster and to allow you to more naturally communicate things inside your brain. They will not be replacing jobs.
"A faint worry that the show should still seem relevant is raised in the catalogue to Electric Dreams. Given the dross so far produced by AI, the curators need not fear. The art in this show is innocent, idealistic, absorbing and highly intelligent. It is both charmingly archaic and constantly prophetic."
The thing people get wrong about a single LLM, is they aren't really about being the fountain of all knowledge. Yes it has been fed the whole of the internet, but that isn't for the task of fact retrieval. What they are a highly effective way of manipulating / working with text, and in the near future as an interface between us and computers for tasks. We can turn out human language effortlessly into a stream of tokens which computers are much better at dealing with.
A lot of the research focus now is on developing systems which your primary LLM queries and interacts with a whole range of our machine learning models.
A lot of the time with machine learning/AI the tech is quite impressive for some things and really bad at others, so the challenge is using it for the right applications.
I'm a bit of a fan of a deck building card game called "Dominion". The game revolves around a market which contains ten piles of ten cards (each pile is usually but not always comprised of identical cards) chosen at random during the game setup from a total pool of about 600 possible piles (across 16 expansion sets). The gameplay usually revolves around building up a deck where the card characteristics compliment each other - a key card to win in one game may be almost valueless in the next. The cards have text based instructions to tell you want to do when playing them - the precise phrasing on the text can be critical in edge cases.
Anyway, there are a couple of digital implementations, one of which is a phone app. This app has a neural network based AI run on the phone which is seriously impressive; it beats me in slightly over 50% of games and I generally hovered on the edge of being in the top 1k of players when I played competitively. Given how hard a problem this is to solve by conventional programming (there are a couple of hundred quadrillion different possible sets of cards to go in the market if you have all expansions available, and because the cards are pretty interdependant you can't pre-assign weights to them very effectively), it's an impressive "real world" use for an AI, at which it is probably better than ChatGPT is at regurgitating Wikipedia.
"Q: How many votes did John Major receive in the Huntingdon constituency at the 1987 general election?
A: In the 1987 general election, John Major, representing the Conservative Party, received 38,904 votes in the Huntingdon constituency."
Wrong, as usual. It's like someone who tries their best to estimate it and hopes no-one notices the mistake because the person they're answering to is too busy doing something else.
Its why Claude is far better...
"While I aim to be precise with numbers, for this specific detail about John Major's exact vote count in the Huntingdon constituency during the 1987 general election, I should be transparent that I could hallucinate specific numbers. While I know Major represented Huntingdon and won the seat comfortably in 1987, I don't feel confident providing the exact vote count without being able to verify it."
"To get the exact verified count, I'd recommend checking official UK electoral records or parliamentary archives."
Yup. An LLM is an LLM: nothing more. It’s a good party trick but it’s not “artificial intelligence” in any meaningful sense.
I would say its more than a party trick, they are lots of genuinely useful things that they aid with e.g. coding, but it is a just another tool. What I wouldn't want to be doing is some sort of job that is text or coding heavily that is overwhelmingly a very repetitive task with little innovation, because tools can certainly be produced that will then only require 1 of you rather than current 5 or 10.
Even for coding - they can help you finish something off (but my word, you better check the output) - but they cannot innovate or problem solve.
Have you tried the full-fat o1 model? If you give it the context it's really quite good on the problem solving. Sadly. I wrote it off originally while giving it similar prompts to gpt-*, but once you realise it needs cajoled in a different way it's really quite good. Again, sadly.
It's worth a quick play with the 'QwQ-32B-Preview' model to see how even an open-source 'free' version is working through a problem using a similar paradigm.
Thanks, I shall have a look. My instinct remains that this form of “AI” is a tool, which is becoming better and better as a tool, but is often oversold.
The thing people get wrong about a single LLM, is they aren't really about being the fountain of all knowledge. Yes it has been fed the whole of the internet, but that isn't for the task of fact retrieval. What they are a highly effective way of manipulating / working with text, and in the near future as an interface between us and computers for tasks. We can turn out human language effortlessly into a stream of tokens which computers are much better at dealing with.
A lot of the research focus now is on developing systems which your primary LLM queries and interacts with a whole range of our machine learning models.
A lot of the time with machine learning/AI the tech is quite impressive for some things and really bad at others, so the challenge is using it for the right applications.
I'm a bit of a fan of a deck building card game called "Dominion". The game revolves around a market which contains ten piles of ten cards (each pile is usually but not always comprised of identical cards) chosen at random during the game setup from a total pool of about 600 possible piles (across 16 expansion sets). The gameplay usually revolves around building up a deck where the card characteristics compliment each other - a key card to win in one game may be almost valueless in the next. The cards have text based instructions to tell you want to do when playing them - the precise phrasing on the text can be critical in edge cases.
Anyway, there are a couple of digital implementations, one of which is a phone app. This app has a neural network based AI run on the phone which is seriously impressive; it beats me in slightly over 50% of games and I generally hovered on the edge of being in the top 1k of players when I played competitively. Given how hard a problem this is to solve by conventional programming (there are a couple of hundred quadrillion different possible sets of cards to go in the market if you have all expansions available, and because the cards are pretty interdependant you can't pre-assign weights to them very effectively), it's an impressive "real world" use for an AI, at which it is probably better than ChatGPT is at regurgitating Wikipedia.
Near* optimal strategies for things like card games, chess, GO, through counterfactual regret minimization and self play. The ideas they are based upon aren't that new, its is the GPU grunt and some clever software engineering to enable running billions of games.
* when I say near, I mean a lot nearer than humans have got.
"Q: How many votes did John Major receive in the Huntingdon constituency at the 1987 general election?
A: In the 1987 general election, John Major, representing the Conservative Party, received 38,904 votes in the Huntingdon constituency."
Wrong, as usual. It's like someone who tries their best to estimate it and hopes no-one notices the mistake because the person they're answering to is too busy doing something else.
Its why Claude is far better...
"While I aim to be precise with numbers, for this specific detail about John Major's exact vote count in the Huntingdon constituency during the 1987 general election, I should be transparent that I could hallucinate specific numbers. While I know Major represented Huntingdon and won the seat comfortably in 1987, I don't feel confident providing the exact vote count without being able to verify it."
"To get the exact verified count, I'd recommend checking official UK electoral records or parliamentary archives."
Yup. An LLM is an LLM: nothing more. It’s a good party trick but it’s not “artificial intelligence” in any meaningful sense.
I would say its more than a party trick, they are lots of genuinely useful things that they aid with e.g. coding, but it is a just another tool. What I wouldn't want to be doing is some sort of job that is text or coding heavily that is overwhelmingly a very repetitive task with little innovation, because tools can certainly be produced that will then only require 1 of you rather than current 5 or 10.
They code like a junior engineer. The output needs checking and can in most cases be optimised.
They are best as an efficiency booster and to allow you to more naturally communicate things inside your brain. They will not be replacing jobs.
This exactly. “AI” is (or could be) a productive booster. You can use it to help ten people do the work of 50, but you struggle to directly replace any of the original ten.
"Q: How many votes did John Major receive in the Huntingdon constituency at the 1987 general election?
A: In the 1987 general election, John Major, representing the Conservative Party, received 38,904 votes in the Huntingdon constituency."
Wrong, as usual. It's like someone who tries their best to estimate it and hopes no-one notices the mistake because the person they're answering to is too busy doing something else.
Its why Claude is far better...
"While I aim to be precise with numbers, for this specific detail about John Major's exact vote count in the Huntingdon constituency during the 1987 general election, I should be transparent that I could hallucinate specific numbers. While I know Major represented Huntingdon and won the seat comfortably in 1987, I don't feel confident providing the exact vote count without being able to verify it."
"To get the exact verified count, I'd recommend checking official UK electoral records or parliamentary archives."
Yup. An LLM is an LLM: nothing more. It’s a good party trick but it’s not “artificial intelligence” in any meaningful sense.
I would say its more than a party trick, they are lots of genuinely useful things that they aid with e.g. coding, but it is a just another tool. What I wouldn't want to be doing is some sort of job that is text or coding heavily that is overwhelmingly a very repetitive task with little innovation, because tools can certainly be produced that will then only require 1 of you rather than current 5 or 10.
Even for coding - they can help you finish something off (but my word, you better check the output) - but they cannot innovate or problem solve.
Have you tried the full-fat o1 model? If you give it the context it's really quite good on the problem solving. Sadly. I wrote it off originally while giving it similar prompts to gpt-*, but once you realise it needs cajoled in a different way it's really quite good. Again, sadly.
It's worth a quick play with the 'QwQ-32B-Preview' model to see how even an open-source 'free' version is working through a problem using a similar paradigm.
I have been very disappointed by the o1 model. Maybe I don't prompt it correctly, but I find too often it has gone off down the wrong track.
Feels like "palliative care" has just become yet another culture war artifact.
It depressed me that the debate in Parliament, amongst friends and family, and in places like this was respectful: but then the usual toilets took away all the nuance.
I was surprised to learn despite all the hype, OpenAI only has 10 million paying customers despite it currently being very cheap to subscribe. I presume Anthropic will be at an least order of magnitude less than that as nobody outside of particularly geeky people are aware of it as they don't do the PR hype thing.
I feel like "only" is doing some quite hard work there. I suspect they are also playing a line between acquiring paid customers and fulfilling their non-profit "for the benefit of all" claims. At least until the current spat with Musk is played out.
We've also no insight into how much their API via their own service or MS Azure is being used.
Wouldn't surprise me at all if they are burning cash like firelighters. But it's quite opaque as to how their finances are going.
They are losing $5bn a year at the moment.
The "non-profit" angle is basically dead now they have taken such a massive amount of Big Daddy Microsoft money (and between them and Nvidia, they are absolutely beholden to them to keep going) and large numbers of the high ups in AI safety team have left.
I've seen various numbers for them losing/making money. But they are still a private company so we've really no idea.
Wouldn't at all surprise me if they were burning VC/MS/Whoever cash like it was going out of fashion, but that's SV startups for you.
Even if only 1 in a 100 makes it as a going concern in the end it'd probably be a better vc decision than investing in the likes of WeWork.
I have a problem with the framing. They are all Conservative PCCs (Thames Valley, Kent, Surrey, Hampshire), and do not mention the party alignment anywhere as far as I can see in the Youtube descriptions.
It's quite political, and imo that needs to be acknowledged.
Ideally imo it needs to be cross-party if presented as "Police and Crime Commissioners".
I'm not convinced that these four are as knowledgeable as I would like.
I've just read a headline on the Sky News website that refers to a "train station".
I despair.
You catch trains from a train... station.
You catch trains from a railway station. A train station sounds like a section of a gymnasium.
A train is what you catch. You don't catch a railway.
But you don't catch crabs etc at a prophylactic station.
(Sorry. Currently reading Ellis 'The sharp end of war' on the rather miserable and sometimes short life of the Allied squaddie in ww2.)
Was there such a thing as an allied squaddie in the second world war? Surely American and British troops would have been armed and fed quite differently.
Sure, they did usually get different reatment in detail (which the book covers, actually). But I might plead in my defence that the Canadians, for instance, would get similar treatment to the British, and the French with one or the other according to the timing and circumstances (I believe).
Edit: and as for the basic facts of the short lives of infantry and tank crews ...
The Yanks had ice cream ships and air conditioning in their submarines. But almost no working torpedos for about 2 years.
‘Italian’ purees likely to contain Chinese forced-labour tomatoes, BBC finds.
Don't pick 650kg of tomatoes a day? Hung by your arms from your cell ceiling for the night.
Has there been any update on Shein listing on the UK stock market? The US won't have them because of their use of slave labour picked cotton.
Honey is another supply chain that is totally "contaminated" with both Chinese honey and fake honey from China i.e. coloured sugar.
I doubt it's possible to make supermarket £1 pots of honey without contamination. Even the uncontaminated stuff is from bees fed exclusively on sugar syrup. Honey from bees which forage in the wild doesn't come much under £5. Good local stuff is £8:
I have a problem with the framing. They are all Conservative PCCs (Thames Valley, Kent, Surrey, Hampshire), and do not mention the party alignment anywhere as far as I can see in the Youtube descriptions.
It's quite political, and imo that needs to be acknowledged.
Ideally imo it needs to be cross-party if presented as "Police and Crime Commissioners".
I'm not convinced that these four are as knowledgeable as I would like.
Is there anybody who doesn't have a podcast these days?
I have a problem with the framing. They are all Conservative PCCs (Thames Valley, Kent, Surrey, Hampshire), and do not mention the party alignment anywhere as far as I can see in the Youtube descriptions.
It's quite political, and imo that needs to be acknowledged.
Ideally imo it needs to be cross-party if presented as "Police and Crime Commissioners".
I'm not convinced that these four are as knowledgeable as I would like.
Is there anybody who doesn't have a podcast these days?
I have a problem with the framing. They are all Conservative PCCs (Thames Valley, Kent, Surrey, Hampshire), and do not mention the party alignment anywhere as far as I can see in the Youtube descriptions.
It's quite political, and imo that needs to be acknowledged.
Ideally imo it needs to be cross-party if presented as "Police and Crime Commissioners".
I'm not convinced that these four are as knowledgeable as I would like.
Is there anybody who doesn't have a podcast these days?
I have a problem with the framing. They are all Conservative PCCs (Thames Valley, Kent, Surrey, Hampshire), and do not mention the party alignment anywhere as far as I can see in the Youtube descriptions.
It's quite political, and imo that needs to be acknowledged.
Ideally imo it needs to be cross-party if presented as "Police and Crime Commissioners".
I'm not convinced that these four are as knowledgeable as I would like.
Is there anybody who doesn't have a podcast these days?
Cladding will be fixed on high-rise buildings in England by 2029, says Angela Rayner
Giles Grover, from the campaign (End Our Cladding Scandal), said of the announcement: “It’s not really going to make much of a difference on the ground, it’s just making an already complicated approach even more complicated. And to be honest, at this stage, it all feels a bit performative, really.
“It doesn’t look like there’s any oversight, you’ve got too many funding schemes rather than a properly joined-up approach. It just looks like they’ve just added further layers of bureaucracy, it’s all still too vague.”
Not gone down well. I remember hearing an senior engineer who works on this talk on the radio saying the complicated regulations around doing any of this work means limited supply of people capable of doing it and so many hurdles to getting it done.
He says it's due to the prosecutions being politically motivated. So now Trump saying his prosecutions were politically motivated now rings true, whether it is or not.
Would have been better to skip the explanation and just pardon him.
He says it's due to the prosecutions being politically motivated. So now Trump saying his prosecutions were politically motivated now rings true, whether it is or not.
Would have been better to skip the explanation and just pardon him.
Right that is what I meant. Did nobody who works for Biden think about this?
Also I think Hunter Biden was supposed to be sentenced in a couple of weeks. For all we know he gets a light sentence anyway. I presume the tax evasion he has already pled guilty to, he gets a big fine and then this gun charge are they really going to lock him up and throw away the key?
He says it's due to the prosecutions being politically motivated. So now Trump saying his prosecutions were politically motivated now rings true, whether it is or not.
Would have been better to skip the explanation and just pardon him.
Right that is what I meant. Did nobody who works for Biden think about this?
Also I think Hunter Biden was supposed to be sentenced in a couple of weeks. For all we know he gets a light sentence anyway.
Perhaps Biden should pardon Trump too before he leaves office.
"Q: How many votes did John Major receive in the Huntingdon constituency at the 1987 general election?
A: In the 1987 general election, John Major, representing the Conservative Party, received 38,904 votes in the Huntingdon constituency."
Wrong, as usual. It's like someone who tries their best to estimate it and hopes no-one notices the mistake because the person they're answering to is too busy doing something else.
Its why Claude is far better...
"While I aim to be precise with numbers, for this specific detail about John Major's exact vote count in the Huntingdon constituency during the 1987 general election, I should be transparent that I could hallucinate specific numbers. While I know Major represented Huntingdon and won the seat comfortably in 1987, I don't feel confident providing the exact vote count without being able to verify it."
"To get the exact verified count, I'd recommend checking official UK electoral records or parliamentary archives."
Yup. An LLM is an LLM: nothing more. It’s a good party trick but it’s not “artificial intelligence” in any meaningful sense.
I would say its more than a party trick, they are lots of genuinely useful things that they aid with e.g. coding, but it is a just another tool. What I wouldn't want to be doing is some sort of job that is text or coding heavily that is overwhelmingly a very repetitive task with little innovation, because tools can certainly be produced that will then only require 1 of you rather than current 5 or 10.
Even for coding - they can help you finish something off (but my word, you better check the output) - but they cannot innovate or problem solve.
Have you tried the full-fat o1 model? If you give it the context it's really quite good on the problem solving. Sadly. I wrote it off originally while giving it similar prompts to gpt-*, but once you realise it needs cajoled in a different way it's really quite good. Again, sadly.
It's worth a quick play with the 'QwQ-32B-Preview' model to see how even an open-source 'free' version is working through a problem using a similar paradigm.
I have been very disappointed by the o1 model. Maybe I don't prompt it correctly, but I find too often it has gone off down the wrong track.
Relatively simple tasks can have high value, though. tooling matters for building quality products.
at @stripe, we have deeply simple way for any employee to file a bug or nit: email a screenshot or video and a LLM makes a ticket and routes to proper team.
"Q: How many votes did John Major receive in the Huntingdon constituency at the 1987 general election?
A: In the 1987 general election, John Major, representing the Conservative Party, received 38,904 votes in the Huntingdon constituency."
Wrong, as usual. It's like someone who tries their best to estimate it and hopes no-one notices the mistake because the person they're answering to is too busy doing something else.
Its why Claude is far better...
"While I aim to be precise with numbers, for this specific detail about John Major's exact vote count in the Huntingdon constituency during the 1987 general election, I should be transparent that I could hallucinate specific numbers. While I know Major represented Huntingdon and won the seat comfortably in 1987, I don't feel confident providing the exact vote count without being able to verify it."
"To get the exact verified count, I'd recommend checking official UK electoral records or parliamentary archives."
Yup. An LLM is an LLM: nothing more. It’s a good party trick but it’s not “artificial intelligence” in any meaningful sense.
I would say its more than a party trick, they are lots of genuinely useful things that they aid with e.g. coding, but it is a just another tool. What I wouldn't want to be doing is some sort of job that is text or coding heavily that is overwhelmingly a very repetitive task with little innovation, because tools can certainly be produced that will then only require 1 of you rather than current 5 or 10.
Even for coding - they can help you finish something off (but my word, you better check the output) - but they cannot innovate or problem solve.
Have you tried the full-fat o1 model? If you give it the context it's really quite good on the problem solving. Sadly. I wrote it off originally while giving it similar prompts to gpt-*, but once you realise it needs cajoled in a different way it's really quite good. Again, sadly.
It's worth a quick play with the 'QwQ-32B-Preview' model to see how even an open-source 'free' version is working through a problem using a similar paradigm.
I have been very disappointed by the o1 model. Maybe I don't prompt it correctly, but I find too often it has gone off down the wrong track.
Relatively simple tasks can have high value, though. tooling matters for building quality products.
at @stripe, we have deeply simple way for any employee to file a bug or nit: email a screenshot or video and a LLM makes a ticket and routes to proper team.
My issue is specifically with the o1 "thinking model". This approach of the output of a query is then fed back in with some secret additional token in a loop. I have found this worse than the non thinking model, because the o1, at least in my case, goes off on the wrong track. I find it better to be in the loop in which the output first comes to me and I can say no not that before it runs 10s of iterations trying to improve its output.
He says it's due to the prosecutions being politically motivated. So now Trump saying his prosecutions were politically motivated now rings true, whether it is or not.
Would have been better to skip the explanation and just pardon him.
Right that is what I meant. Did nobody who works for Biden think about this?
Also I think Hunter Biden was supposed to be sentenced in a couple of weeks. For all we know he gets a light sentence anyway. I presume the tax evasion he has already pled guilty to, he gets a big fine and then this gun charge are they really going to lock him up and throw away the key?
Poor decision, IMO.
While I agree that the Hunter Biden prosecution had a large element of political motivation, and none of the political allegations which the GOP spent five years pursuing had any substance, Biden had made it a touchstone of his approach that he wouldn't interfere in the process.
He says it's due to the prosecutions being politically motivated. So now Trump saying his prosecutions were politically motivated now rings true, whether it is or not.
Would have been better to skip the explanation and just pardon him.
Right that is what I meant. Did nobody who works for Biden think about this?
Also I think Hunter Biden was supposed to be sentenced in a couple of weeks. For all we know he gets a light sentence anyway. I presume the tax evasion he has already pled guilty to, he gets a big fine and then this gun charge are they really going to lock him up and throw away the key?
Poor decision, IMO.
While I agree that the Hunter Biden prosecution had a large element of political motivation, and none of the political allegations which the GOP spent five years pursuing had any substance, Biden had made it a touchstone of his approach that he wouldn't interfere in the process.
In the other hand, he didn't pardon him and then appoint him ambassador to France...
Its even more stinky than just pardoning on as he sees it a trumped up charge relating to buying a gun.....
The pardon covers any potential federal crimes Hunter Biden may have committed from January 2014 through December 2024. This period covers Hunter Biden's tenure as a board member of Burisama, a Ukrainian company that has been accused of bribery.
‘Italian’ purees likely to contain Chinese forced-labour tomatoes, BBC finds.
Don't pick 650kg of tomatoes a day? Hung by your arms from your cell ceiling for the night.
Good investigation by the BBC.
Sadly, I suspect there are all sorts of things like this in our supply chains that we simply don't know about.
Although this is part corrupting of the supply chain, I am always surprised the wide variance in food regulation. Vanilla Ice Cream, does not need to have any vanilla, cream or milk in it. Natural flavours, that can mean literally anything, and not what we think of as "natural". It seems honey can refer to literally coloured liquid sugar that hasn't been near a bee. Where as other things, are massively protected.
He says it's due to the prosecutions being politically motivated. So now Trump saying his prosecutions were politically motivated now rings true, whether it is or not.
Would have been better to skip the explanation and just pardon him.
Right that is what I meant. Did nobody who works for Biden think about this?
Also I think Hunter Biden was supposed to be sentenced in a couple of weeks. For all we know he gets a light sentence anyway. I presume the tax evasion he has already pled guilty to, he gets a big fine and then this gun charge are they really going to lock him up and throw away the key?
Max 17 years sentence, and you're taking your chances with a Trump appointed judge. US system is so broken.
He says it's due to the prosecutions being politically motivated. So now Trump saying his prosecutions were politically motivated now rings true, whether it is or not.
Would have been better to skip the explanation and just pardon him.
Right that is what I meant. Did nobody who works for Biden think about this?
Also I think Hunter Biden was supposed to be sentenced in a couple of weeks. For all we know he gets a light sentence anyway. I presume the tax evasion he has already pled guilty to, he gets a big fine and then this gun charge are they really going to lock him up and throw away the key?
Max 17 years sentence, and you're taking your chances with a Trump appointed judge. US system is so broken.
If he got some crazy sentence, then Biden Snr could use his get out jail free card as it definitely looks like a political hit job. But in reality, very rich well connected people the US rarely get the full force of the law.
The fact he has basically given him immunity for the past 10 years is the real story now. That goes far beyond getting your son off a trumped up gun charge.
He says it's due to the prosecutions being politically motivated. So now Trump saying his prosecutions were politically motivated now rings true, whether it is or not.
Would have been better to skip the explanation and just pardon him.
Right that is what I meant. Did nobody who works for Biden think about this?
Also I think Hunter Biden was supposed to be sentenced in a couple of weeks. For all we know he gets a light sentence anyway. I presume the tax evasion he has already pled guilty to, he gets a big fine and then this gun charge are they really going to lock him up and throw away the key?
Max 17 years sentence, and you're taking your chances with a Trump appointed judge. US system is so broken.
If he got some crazy sentence, then Biden Snr could use his get out jail free card as it definitely looks like a political hit job. But in reality, very rich well connected people the US rarely get the full force of the law.
The fact he has basically given him immunity for the past 10 years is the real story now. That goes far beyond getting your son off a trumped up gun charge.
Yes not sure why he didn't wait, maybe he thought that would look worse. Maybe he wanted his son not to worry.
I'd hazard a guess that rich well connected Democrats are going to see quite a bit of the force of the law over the next 4 years.
Sounds like there was a lot more going on over that Luton decision than just electric cars.
Nissan are also in big trouble...and Audi...and Jag don't even make cars these days, they are in the presume and fashion business (that is what we have to guess from their rebrand).
‘Italian’ purees likely to contain Chinese forced-labour tomatoes, BBC finds.
Don't pick 650kg of tomatoes a day? Hung by your arms from your cell ceiling for the night.
Good investigation by the BBC.
Sadly, I suspect there are all sorts of things like this in our supply chains that we simply don't know about.
Although this is part corrupting of the supply chain, I am always surprised the wide variance in food regulation. Vanilla Ice Cream, does not need to have any vanilla, cream or milk in it. Natural flavours, that can mean literally anything, and not what we think of as "natural". It seems honey can refer to literally coloured liquid sugar that hasn't been near a bee. Where as other things, are massively protected.
I've bought that squeezy runny honey in a plastic bottle a few times before (can't remember what it's called, not Ronseal but similar) and it totally calcifies after a couple of months.
Sounds like there was a lot more going on over that Luton decision than just electric cars.
No, electric cars - or rather, no electric cars - is their problem.
The failure of western car companies, and governments, to realise the speed and scale of the transition, and the threat of cheap mass market Chinese cars, will likely doom some of them.
Even S Korea, which did realise, but relied on its comparatively unsubsidised industry, is going to have to work very hard to compete.
Sounds like there was a lot more going on over that Luton decision than just electric cars.
No, electric cars - or rather, no electric cars - is their problem.
The failure of western car companies, and governments, to realise the speed and scale of the transition, and the threat of cheap mass market Chinese cars, will likely doom some of them.
Even S Korea, which did realise, but relied on its comparatively unsubsidised industry, is going to have to work very hard to compete.
Given there has always been massive oversupply in the car market it is probably not a bad thing if there is a bit of a shake out of the industry.
Sounds like there was a lot more going on over that Luton decision than just electric cars.
No, electric cars - or rather, no electric cars - is their problem.
The failure of western car companies, and governments, to realise the speed and scale of the transition, and the threat of cheap mass market Chinese cars, will likely doom some of them.
Even S Korea, which did realise, but relied on its comparatively unsubsidised industry, is going to have to work very hard to compete.
Sounds like there was a lot more going on over that Luton decision than just electric cars.
No, electric cars - or rather, no electric cars - is their problem.
The failure of western car companies, and governments, to realise the speed and scale of the transition, and the threat of cheap mass market Chinese cars, will likely doom some of them.
Even S Korea, which did realise, but relied on its comparatively unsubsidised industry, is going to have to work very hard to compete.
People just dont want to buy them
Or rather:
The people who can afford to buy the somewhat limited cars on the market at the moment are a small subset of the car-buying public. Until range improves, and prices come down, the market is hence limited.
Sounds like there was a lot more going on over that Luton decision than just electric cars.
No, electric cars - or rather, no electric cars - is their problem.
The failure of western car companies, and governments, to realise the speed and scale of the transition, and the threat of cheap mass market Chinese cars, will likely doom some of them.
Even S Korea, which did realise, but relied on its comparatively unsubsidised industry, is going to have to work very hard to compete.
People just dont want to buy them
Or rather:
The people who can afford to buy the somewhat limited cars on the market at the moment are a small subset of the car-buying public. Until range improves, and prices come down, the market is hence limited.
Tesla dominating the news also does not help IMV.
Nicely put and I cant see much changing in the short term
99 Kriegsminister Streichholz und Benzinkanister Hielten sich für schlaue Leute Witterten schon fette Beute Riefen: „Krieg!“, und wollten Macht Mann, wer hätte das gedacht? Dass es einmal so weit kommt Wegen 99 Luftballons
99 Kriegsminister Streichholz und Benzinkanister Hielten sich für schlaue Leute Witterten schon fette Beute Riefen: „Krieg!“, und wollten Macht Mann, wer hätte das gedacht? Dass es einmal so weit kommt Wegen 99 Luftballons
Sounds like there was a lot more going on over that Luton decision than just electric cars.
No, electric cars - or rather, no electric cars - is their problem.
The failure of western car companies, and governments, to realise the speed and scale of the transition, and the threat of cheap mass market Chinese cars, will likely doom some of them.
Even S Korea, which did realise, but relied on its comparatively unsubsidised industry, is going to have to work very hard to compete.
People just dont want to buy them
Or rather:
The people who can afford to buy the somewhat limited cars on the market at the moment are a small subset of the car-buying public. Until range improves, and prices come down, the market is hence limited.
Tesla dominating the news also does not help IMV.
Nicely put and I cant see much changing in the short term
It's changing every single day.
Every day, the proportion of EVs on the road increases.
Sounds like there was a lot more going on over that Luton decision than just electric cars.
No, electric cars - or rather, no electric cars - is their problem.
The failure of western car companies, and governments, to realise the speed and scale of the transition, and the threat of cheap mass market Chinese cars, will likely doom some of them.
Even S Korea, which did realise, but relied on its comparatively unsubsidised industry, is going to have to work very hard to compete.
People just dont want to buy them
Or rather:
The people who can afford to buy the somewhat limited cars on the market at the moment are a small subset of the car-buying public. Until range improves, and prices come down, the market is hence limited.
Tesla dominating the news also does not help IMV.
Nicely put and I cant see much changing in the short term
It's changing every single day.
Every day, the proportion of EVs on the road increases.
Unfortunately for that argument the rate of increase is pedestrian.
99 Kriegsminister Streichholz und Benzinkanister Hielten sich für schlaue Leute Witterten schon fette Beute Riefen: „Krieg!“, und wollten Macht Mann, wer hätte das gedacht? Dass es einmal so weit kommt Wegen 99 Luftballons
Biedermann und die Brandstifter?
Genau
"Das Stück Biedermann und die Brandstifter ist eine Mischung aus komischen und makabren Elementen mit düsterem Thema und Ende (eine Burleske). Allerdings ist es kein tragisches Stück, denn der Protagonist Biedermann geht nicht bewusst und zwingend um eines erhabenen Wertes willen in eine Katastrophe, sondern er erleidet aus Feigheit, Dummheit und Verblendung ein vermeidbares „Schicksal“. Es ist die dichterische Gestaltung eines prototypischen Geschehens mit Personen in ihren unverkennbaren, typischen Rollen."
So what happens to the $300k+ in back pay rent that Hunter Biden owes my family from 2019-2020? Is that pardoned now? Thanks Joe (This is a true story)
So what happens to the $300k+ in back pay rent that Hunter Biden owes my family from 2019-2020? Is that pardoned now? Thanks Joe (This is a true story)
‘Italian’ purees likely to contain Chinese forced-labour tomatoes, BBC finds.
Don't pick 650kg of tomatoes a day? Hung by your arms from your cell ceiling for the night.
Good investigation by the BBC.
Sadly, I suspect there are all sorts of things like this in our supply chains that we simply don't know about.
Although this is part corrupting of the supply chain, I am always surprised the wide variance in food regulation. Vanilla Ice Cream, does not need to have any vanilla, cream or milk in it. Natural flavours, that can mean literally anything, and not what we think of as "natural". It seems honey can refer to literally coloured liquid sugar that hasn't been near a bee. Where as other things, are massively protected.
I've bought that squeezy runny honey in a plastic bottle a few times before (can't remember what it's called, not Ronseal but similar) and it totally calcifies after a couple of months.
So what happens to the $300k+ in back pay rent that Hunter Biden owes my family from 2019-2020? Is that pardoned now? Thanks Joe (This is a true story)
So what happens to the $300k+ in back pay rent that Hunter Biden owes my family from 2019-2020? Is that pardoned now? Thanks Joe (This is a true story)
So what happens to the $300k+ in back pay rent that Hunter Biden owes my family from 2019-2020? Is that pardoned now? Thanks Joe (This is a true story)
I don't think not paying rent is a Federal crime..
$300k is quite some rental bill for a year or two! As an aside, Joe Biden could presumably just pay him back if he wanted... he must be fairly wealthy by now.
Sounds like there was a lot more going on over that Luton decision than just electric cars.
No, electric cars - or rather, no electric cars - is their problem.
The failure of western car companies, and governments, to realise the speed and scale of the transition, and the threat of cheap mass market Chinese cars, will likely doom some of them.
Even S Korea, which did realise, but relied on its comparatively unsubsidised industry, is going to have to work very hard to compete.
People just dont want to buy them
Or rather:
The people who can afford to buy the somewhat limited cars on the market at the moment are a small subset of the car-buying public. Until range improves, and prices come down, the market is hence limited.
Tesla dominating the news also does not help IMV.
Nicely put and I cant see much changing in the short term
It's changing every single day.
Every day, the proportion of EVs on the road increases.
Indeed, but AUI (in the UK at least) much of that rise is fleet buyers. And the secondhand market for EVs appears to be pants, with leasing prices apparently increasing spectacularly because of this.
So what happens to the $300k+ in back pay rent that Hunter Biden owes my family from 2019-2020? Is that pardoned now? Thanks Joe (This is a true story)
Sounds like there was a lot more going on over that Luton decision than just electric cars.
No, electric cars - or rather, no electric cars - is their problem.
The failure of western car companies, and governments, to realise the speed and scale of the transition, and the threat of cheap mass market Chinese cars, will likely doom some of them.
Even S Korea, which did realise, but relied on its comparatively unsubsidised industry, is going to have to work very hard to compete.
Sounds like there was a lot more going on over that Luton decision than just electric cars.
No, electric cars - or rather, no electric cars - is their problem.
The failure of western car companies, and governments, to realise the speed and scale of the transition, and the threat of cheap mass market Chinese cars, will likely doom some of them.
Even S Korea, which did realise, but relied on its comparatively unsubsidised industry, is going to have to work very hard to compete.
People just dont want to buy them
Or rather:
The people who can afford to buy the somewhat limited cars on the market at the moment are a small subset of the car-buying public. Until range improves, and prices come down, the market is hence limited.
Tesla dominating the news also does not help IMV.
Nicely put and I cant see much changing in the short term
It's changing every single day.
Every day, the proportion of EVs on the road increases.
Indeed, but AUI (in the UK at least) much of that rise is fleet buyers. And the secondhand market for EVs appears to be pants, with leasing prices apparently increasing spectacularly because of this.
Surely its a good thing if secondhand EVs are cheap to buy?
My 4 year old Kia eniro has proven reliable and the range is as good as new. Lovely to drive too. If you can get one cheap second hand it's a great buy.
Sounds like there was a lot more going on over that Luton decision than just electric cars.
No, electric cars - or rather, no electric cars - is their problem.
The failure of western car companies, and governments, to realise the speed and scale of the transition, and the threat of cheap mass market Chinese cars, will likely doom some of them.
Even S Korea, which did realise, but relied on its comparatively unsubsidised industry, is going to have to work very hard to compete.
People just dont want to buy them
Or rather:
The people who can afford to buy the somewhat limited cars on the market at the moment are a small subset of the car-buying public. Until range improves, and prices come down, the market is hence limited.
Tesla dominating the news also does not help IMV.
Nicely put and I cant see much changing in the short term
It's changing every single day.
Every day, the proportion of EVs on the road increases.
Indeed, but AUI (in the UK at least) much of that rise is fleet buyers. And the secondhand market for EVs appears to be pants, with leasing prices apparently increasing spectacularly because of this.
The problem with EVs is that the cars are stilling going through the iPhone 1 to (roughly) X stage where there are significant improvements with every release.
Add on some stories about lack of support even when under warranty and an petrol car that will last 15 or so years seems a safer bet...
Sounds like there was a lot more going on over that Luton decision than just electric cars.
No, electric cars - or rather, no electric cars - is their problem.
The failure of western car companies, and governments, to realise the speed and scale of the transition, and the threat of cheap mass market Chinese cars, will likely doom some of them.
Even S Korea, which did realise, but relied on its comparatively unsubsidised industry, is going to have to work very hard to compete.
People just dont want to buy them
Don't they ?
From chatting to people, lots of people want to buy EVs. But they just cost too much, and range anxiety is a real thing (as is charging...). Then there are the many, many people for whom buying a new, as opposed to second-hand, car is not an option.
Sounds like there was a lot more going on over that Luton decision than just electric cars.
No, electric cars - or rather, no electric cars - is their problem.
The failure of western car companies, and governments, to realise the speed and scale of the transition, and the threat of cheap mass market Chinese cars, will likely doom some of them.
Even S Korea, which did realise, but relied on its comparatively unsubsidised industry, is going to have to work very hard to compete.
People just dont want to buy them
Or rather:
The people who can afford to buy the somewhat limited cars on the market at the moment are a small subset of the car-buying public. Until range improves, and prices come down, the market is hence limited.
Tesla dominating the news also does not help IMV.
Nicely put and I cant see much changing in the short term
It's changing every single day.
Every day, the proportion of EVs on the road increases.
Indeed, but AUI (in the UK at least) much of that rise is fleet buyers. And the secondhand market for EVs appears to be pants, with leasing prices apparently increasing spectacularly because of this.
Surely its a good thing if secondhand EVs are cheap to buy?
My 4 year old Kia eniro has proven reliable and the range is as good as new. Lovely to drive too. If you can get one cheap second hand it's a great buy.
A very high depreciation is not good for new car sales. And it also leads to questions about *why* the cars have decreased so much in value.
I don't know the answer to that. Is it just the used-battery fears, or something deeper?
‘Italian’ purees likely to contain Chinese forced-labour tomatoes, BBC finds.
Don't pick 650kg of tomatoes a day? Hung by your arms from your cell ceiling for the night.
Good investigation by the BBC.
Sadly, I suspect there are all sorts of things like this in our supply chains that we simply don't know about.
Although this is part corrupting of the supply chain, I am always surprised the wide variance in food regulation. Vanilla Ice Cream, does not need to have any vanilla, cream or milk in it. Natural flavours, that can mean literally anything, and not what we think of as "natural". It seems honey can refer to literally coloured liquid sugar that hasn't been near a bee. Where as other things, are massively protected.
With honey, my understanding is that honey from local bees, that feed off the pollen from the same plants that give you hayfever, can have a beneficial effect in reducing hayfever. Also, unless you're really unlucky and come across a small-scale wrong-'un, local honey is much less likely to be rebadged liquid sugar.
The utter blistering hypocrisy of Biden and the Democrats
They spend two years concocting often dubious law cases against Trump - basically trying anything to get him in jail. But that’s ok because the Democrats aren’t criminals and they don’t break the law like Trump so they are allowed to bend the law a bit ahem
And now Biden flat out pardons his convicted criminal son. There will be an awful lot of Americans feeling very vindicated in their vote for Trump
The utter blistering hypocrisy of Biden and the Democrats
They spend two years concocting often dubious law cases against Trump - basically trying anything to get him in jail. But that’s ok because the Democrats aren’t criminals and they don’t break the law like Trump so they are allowed to bend the law a bit ahem
And now Biden flat out pardons his convicted criminal son. There will be an awful lot of Americans feeling very vindicated in their vote for Trump
And the republicans did all they could to keep Trump out of jail, and will continue to do so now. I don't see any righteous anger from you about that.
The presidential pardon system is a cancerous ulcer in America's body politic. Politically, Biden should not have done this. But from a human point of view, I see why he did.
‘Italian’ purees likely to contain Chinese forced-labour tomatoes, BBC finds.
Don't pick 650kg of tomatoes a day? Hung by your arms from your cell ceiling for the night.
Good investigation by the BBC.
Sadly, I suspect there are all sorts of things like this in our supply chains that we simply don't know about.
Although this is part corrupting of the supply chain, I am always surprised the wide variance in food regulation. Vanilla Ice Cream, does not need to have any vanilla, cream or milk in it. Natural flavours, that can mean literally anything, and not what we think of as "natural". It seems honey can refer to literally coloured liquid sugar that hasn't been near a bee. Where as other things, are massively protected.
I've bought that squeezy runny honey in a plastic bottle a few times before (can't remember what it's called, not Ronseal but similar) and it totally calcifies after a couple of months.
I won't make that mistake again.
Just warm it up and it will go back to liquid.
Yeah but it doesn't.
Everyone says that and once it's stiffened up in the bottle I can't get it to shift again, even after soaking in near boiling water.
Sounds like there was a lot more going on over that Luton decision than just electric cars.
No, electric cars - or rather, no electric cars - is their problem.
The failure of western car companies, and governments, to realise the speed and scale of the transition, and the threat of cheap mass market Chinese cars, will likely doom some of them.
Even S Korea, which did realise, but relied on its comparatively unsubsidised industry, is going to have to work very hard to compete.
People just dont want to buy them
Or rather:
The people who can afford to buy the somewhat limited cars on the market at the moment are a small subset of the car-buying public. Until range improves, and prices come down, the market is hence limited.
Tesla dominating the news also does not help IMV.
Nicely put and I cant see much changing in the short term
It's changing every single day.
Every day, the proportion of EVs on the road increases.
Indeed, but AUI (in the UK at least) much of that rise is fleet buyers. And the secondhand market for EVs appears to be pants, with leasing prices apparently increasing spectacularly because of this.
Surely its a good thing if secondhand EVs are cheap to buy?
My 4 year old Kia eniro has proven reliable and the range is as good as new. Lovely to drive too. If you can get one cheap second hand it's a great buy.
A very high depreciation is not good for new car sales. And it also leads to questions about *why* the cars have decreased so much in value.
I don't know the answer to that. Is it just the used-battery fears, or something deeper?
Used battery fears will be part of it, the other is that the technology is developing so quickly, which will cause faster depreciation in older models.
Comments
It's worth a quick play with the 'QwQ-32B-Preview' model to see how even an open-source 'free' version is working through a problem using a similar paradigm.
https://www.youtube.com/watch?v=Es6yuMlyfPw
"In this profound keynote, Vector co-founder Geoffrey Hinton explores the philosophical implications of artificial intelligence and its potential to surpass human intelligence. Drawing from decades of expertise, Hinton shares his growing concerns about AI's existential risks while examining fundamental questions about consciousness, understanding, and the nature of intelligence itself."
(He makes some quite good (academic) jokes. So it's not all doom'n'gloom)
I've no idea what the state of research on this question is. It feels like something conceptually very different to what LLMs are doing.
The "non-profit" angle is basically dead now they have taken such a massive amount of Big Daddy Microsoft money (and between them and Nvidia, they are absolutely beholden to them to keep going) and large numbers of the high ups in AI safety team have left.
Wouldn't at all surprise me if they were burning VC/MS/Whoever cash like it was going out of fashion, but that's SV startups for you.
One huge big problem they are blasting GPU / data centres so hard, that equipment will need constant replacement. Its a bit like Bitcoin mining in that respect, you haven't bought yourself 100k GPUs and that is cap-ex investment for the next 10 years, they are going through them fast. That is why a lot of work on more efficient inference.
They are best as an efficiency booster and to allow you to more naturally communicate things inside your brain. They will not be replacing jobs.
"A faint worry that the show should still seem relevant is raised in the catalogue to Electric Dreams. Given the dross so far produced by AI, the curators need not fear. The art in this show is innocent, idealistic, absorbing and highly intelligent. It is both charmingly archaic and constantly prophetic."
https://www.theguardian.com/artanddesign/2024/dec/01/electric-dreams-art-and-technology-before-the-internet-tate-modern-london-review-prophetic-and-unexpectedly-poignant
"Electric Dreams: Art and Technology Before the Internet review – charmingly archaic and constantly prophetic"
I'm a bit of a fan of a deck building card game called "Dominion". The game revolves around a market which contains ten piles of ten cards (each pile is usually but not always comprised of identical cards) chosen at random during the game setup from a total pool of about 600 possible piles (across 16 expansion sets).
The gameplay usually revolves around building up a deck where the card characteristics compliment each other - a key card to win in one game may be almost valueless in the next.
The cards have text based instructions to tell you want to do when playing them - the precise phrasing on the text can be critical in edge cases.
Anyway, there are a couple of digital implementations, one of which is a phone app. This app has a neural network based AI run on the phone which is seriously impressive; it beats me in slightly over 50% of games and I generally hovered on the edge of being in the top 1k of players when I played competitively. Given how hard a problem this is to solve by conventional programming (there are a couple of hundred quadrillion different possible sets of cards to go in the market if you have all expansions available, and because the cards are pretty interdependant you can't pre-assign weights to them very effectively), it's an impressive "real world" use for an AI, at which it is probably better than ChatGPT is at regurgitating Wikipedia.
* when I say near, I mean a lot nearer than humans have got.
‘Italian’ purees likely to contain Chinese forced-labour tomatoes, BBC finds.
Don't pick 650kg of tomatoes a day? Hung by your arms from your cell ceiling for the night.
Honey is another supply chain that is totally "contaminated" with both Chinese honey and fake honey from China i.e. coloured sugar.
https://www.youtube.com/watch?v=rcmwS3ltVBc
I'd be interested in thoughts on the content.
I have a problem with the framing. They are all Conservative PCCs (Thames Valley, Kent, Surrey, Hampshire), and do not mention the party alignment anywhere as far as I can see in the Youtube descriptions.
It's quite political, and imo that needs to be acknowledged.
Ideally imo it needs to be cross-party if presented as "Police and Crime Commissioners".
I'm not convinced that these four are as knowledgeable as I would like.
Squaddies - ask JackW, who was probably there.
https://farmhousedeli.co.uk/products/cotswold-honey-set-or-runny
https://www.youtube.com/@justgetatesla
https://x.com/disclosetv/status/1863381567919292875
Giles Grover, from the campaign (End Our Cladding Scandal), said of the announcement: “It’s not really going to make much of a difference on the ground, it’s just making an already complicated approach even more complicated. And to be honest, at this stage, it all feels a bit performative, really.
“It doesn’t look like there’s any oversight, you’ve got too many funding schemes rather than a properly joined-up approach. It just looks like they’ve just added further layers of bureaucracy, it’s all still too vague.”
https://www.theguardian.com/society/2024/dec/02/cladding-will-be-fixed-on-high-rise-buildings-in-england-by-2029-says-angela-rayner
Not gone down well. I remember hearing an senior engineer who works on this talk on the radio saying the complicated regulations around doing any of this work means limited supply of people capable of doing it and so many hurdles to getting it done.
Would have been better to skip the explanation and just pardon him.
Also I think Hunter Biden was supposed to be sentenced in a couple of weeks. For all we know he gets a light sentence anyway. I presume the tax evasion he has already pled guilty to, he gets a big fine and then this gun charge are they really going to lock him up and throw away the key?
tooling matters for building quality products.
at @stripe, we have deeply simple way for any employee to file a bug or nit: email a screenshot or video and a LLM makes a ticket and routes to proper team.
result: we 6x'd # of bugs filed, leading to 1000s of fixes just this year.
https://x.com/jeff_weinstein/status/1857506026729398307
While I agree that the Hunter Biden prosecution had a large element of political motivation, and none of the political allegations which the GOP spent five years pursuing had any substance, Biden had made it a touchstone of his approach that he wouldn't interfere in the process.
The pardon covers any potential federal crimes Hunter Biden may have committed from January 2014 through December 2024. This period covers Hunter Biden's tenure as a board member of Burisama, a Ukrainian company that has been accused of bribery.
https://www.bbc.co.uk/news/live/c4ngnw2qr01t
Sadly, I suspect there are all sorts of things like this in our supply chains that we simply don't know about.
US system is so broken.
The fact he has basically given him immunity for the past 10 years is the real story now. That goes far beyond getting your son off a trumped up gun charge.
I'd hazard a guess that rich well connected Democrats are going to see quite a bit of the force of the law over the next 4 years.
Boss of car making giant Stellantis abruptly quits
https://www.bbc.com/news/articles/cvgx5pe4r0xo
Sounds like there was a lot more going on over that Luton decision than just electric cars.
F1: utter fluke with the bet, but given my atrocious fortune earlier in the season I'll take it.
I won't make that mistake again.
The failure of western car companies, and governments, to realise the speed and scale of the transition, and the threat of cheap mass market Chinese cars, will likely doom some of them.
Even S Korea, which did realise, but relied on its comparatively unsubsidised industry, is going to have to work very hard to compete.
https://youtu.be/NXcbyxFNCSc?si=TRzhMKd5npvaNyFb
The people who can afford to buy the somewhat limited cars on the market at the moment are a small subset of the car-buying public. Until range improves, and prices come down, the market is hence limited.
Tesla dominating the news also does not help IMV.
Streichholz und Benzinkanister
Hielten sich für schlaue Leute
Witterten schon fette Beute
Riefen: „Krieg!“, und wollten Macht
Mann, wer hätte das gedacht?
Dass es einmal so weit kommt
Wegen 99 Luftballons
Every day, the proportion of EVs on the road increases.
"Das Stück Biedermann und die Brandstifter ist eine Mischung aus komischen und makabren Elementen mit düsterem Thema und Ende (eine Burleske). Allerdings ist es kein tragisches Stück, denn der Protagonist Biedermann geht nicht bewusst und zwingend um eines erhabenen Wertes willen in eine Katastrophe, sondern er erleidet aus Feigheit, Dummheit und Verblendung ein vermeidbares „Schicksal“. Es ist die dichterische Gestaltung eines prototypischen Geschehens mit Personen in ihren unverkennbaren, typischen Rollen."
🤣🤣🤣
https://x.com/shaunmmaguire/status/1863442352456491202
Yup. True story. Hunter was our tenant in Venice, CA. Didn't pay rent for over a year. Tried to pay w/ art made from his own feces.
https://x.com/shaunmmaguire/status/1863442726588498117
As an aside, Joe Biden could presumably just pay him back if he wanted... he must be fairly wealthy by now.
My 4 year old Kia eniro has proven reliable and the range is as good as new. Lovely to drive too. If you can get one cheap second hand it's a great buy.
Add on some stories about lack of support even when under warranty and an petrol car that will last 15 or so years seems a safer bet...
German Chancellor Olaf Scholz arrives in Kyiv for the first time in two and a half years, DPA reports. He is to meet with President Zelenskyy.
https://x.com/EuromaidanPress/status/1863475571558137954
I don't know the answer to that. Is it just the used-battery fears, or something deeper?
NEW THREAD
They spend two years concocting often dubious law cases against Trump - basically trying anything to get him in jail. But that’s ok because the Democrats aren’t criminals and they don’t break the law like Trump so they are allowed to bend the law a bit ahem
And now Biden flat out pardons his convicted criminal son. There will be an awful lot of Americans feeling very vindicated in their vote for Trump
The presidential pardon system is a cancerous ulcer in America's body politic. Politically, Biden should not have done this. But from a human point of view, I see why he did.
Everyone says that and once it's stiffened up in the bottle I can't get it to shift again, even after soaking in near boiling water.