Richard G Clegg
richardclegg.bsky.social
Richard G Clegg
@richardclegg.bsky.social
720 followers 160 following 940 posts
Academic studying complex networks at Queen Mary University of London and amateur scuba instructor. I do a lot of research using the Raphtory software for temporal networks, a fast efficient way to analyse network data: https://www.raphtory.com/ He/Him
Posts Media Videos Starter Packs
I suppose at least reform campaigners love what they are saying. Could be worse though could be a tory campaigner.
These things were true of the Internet in 1999.
I read the original series of Open AI tweets. It was clear to me that the person was describing an indexing problem - a very difficult problem in mathematics (spotting when an existing solution matches a problem) - but you needed to read beyond first message.
So wholesome watching sumo do tourism has lightened my week. Using hire bikes was my fave.
Time will tell I guess. I am no expert at all. I can find quite a few "this time we definitely reject the archbishop of Canterbury and we really definitely mean it" moments. The one I pointed out being the most recent before the one this week.
It is a quite bleak scenario that they paint for anyone on the left but it is a lot different to those painted by older organisations.
I looked into it when you posted that poll but did not comment:
en.wikipedia.org/wiki/Opinion...
That reflects what I was seeing: they peg reform and green 2-5 points higher than others and labour 5 or more points lower.
Opinion polling for the next United Kingdom general election - Wikipedia
en.wikipedia.org
I went over Niagara falls in a barrel for NOTHING! Not even a single gold coin.
Yeah. Govt investment typically asks for a much larger pay back ratio for the "worthwhile" category in other areas.
Honestly at the edge of my knowledge there so I would not like to say.
I am blushing at the audacity of their maths if I understand it correctly. The economic cost of the work days lost isn't huge. I *think* their trick is piling up lost work years from deaths "130,000 future work-years due to premature deaths" which is audacious if that is the trick.
High scepticism: "The report, commissioned by Moderna,".
"Whatever you do keep taking heroin" according to dealer. We pile up these "unemptied rubbish bins cost the UK economy 100 billion a year" reports. Seems our economy could be in the quintillions if we sorted them.
But if you *had* lived in the midlands in the early 80s it would be super relevant. (I got one minute into "if you play a 45 at 33" when explaining time dilation and frequency shifting in signal processing).
Absent other concerns (infection of other people etc) if I were offered "90 quid or you definitely get COVID" I would probably take the money and I am not that short of money.
*nod* I am struggling with my own mental arithmetic here. It is mild in me (last time I had it I barely noticed - tired for a day back in gym second day) but infection of others is a concern.
*nod* average worker is taking 4.4 sick days a year in 2024. Average worker produces approx 135 quid productivity a day. Unless there's something I am not seeing in those averages COVID would need to be near 20% of sick days before you get to break even for a 90 quid jab.
I suspect COVID is nowhere near. It is way less common than flu now and the vaccine is crazy expensive. Also a mass vaccination may just be pushing around which varieties spread and how quickly it resists that particular vaccine. (Flu vaccine only targets a few types.)
I can't imagine how difficult that is going to be. Shipping is constant. The currents are fairly ripping and the channel is blessed with bottom features rising several metres (wrecks as we divers call them).
To be clear ChatGPT only used code when I told it it was wrong or when I asked it to check. If I did not do that it did computation "by hand" which looked correct but was not. Real "danger zone" it once got first term only right so it might elude you "hand checking" it.
You should try it on Huffman encoding it is amazing. It chuffs out a few binary digits, realises it got a 2 or a 3 in one of them. It then realises there must be a mistake and goes back again. It can't do it at all but keeps trying. Eventually I just say "give it up mate you can't do this".
Little bit of signal processing humour for a Monday morning.
From the hand worked part:
"Compute real part:

1−2+3+7+1+1+1+1=12"

Oops.
Kind of thing I do all the time. It's so interesting to me that this type of AI failure was predicted five decades ago.
Asked it the same question telling it to check. It produced the first value by hand getting it wrong. It produced all the values with python FFT getting it right. It did not spot the discrepancy in X[0]. When I prompted it noticed and it said it was a manual arithmetic error.
Privacy products are very heavily sold in some parts of the internet. There is money in convincing people of these things. You must have a VPN 100% of the time is a not uncommon belief.