Comments by "Aden Wellsmith" (@adenwellsmith6908) on "Sabine Hossenfelder" channel.

Where's the shortage of cash for AI?
5
Guess who reads your emails. 😂
4
@TheJohnreeves Based on lots of things.
4
The idea of using AI to select the corpus to train on is interesting. The AlphaGo concept to try and train a maths model, that's interesting but I still think you need a Wolfram Alpha as a back end. Specialist submodels to make the problems more tractable. That's interesting too. These are all things that others will do very quickly What I expect to see soon is AI being used to get to a good model quickly, that's then optomised afterwards. ie. Teach AI how to learn, not what to learn, that comes later. The size of the hardware I've seen is still large numbers of graphics cards, obscene amounts of NVME storage and lots of cores. That's to run their large model. The thought process bit is interesting too.
3
Space is a manifold and is locally flat.
3
@A.-wm1nj That's the bit I like about the AI. A digital mentor. Please critique the following code. It's not a human. There's no emotion involved. There's no revealing of weaknesses. Code review. Asking it to improve code. The other is boiler plate. I did an analysis of a very large OO application. The ratio of secondary classes to core classes was 7+ to one. One business class, needed DB support. It needed the UI to manage it. It needed unit testing, .... Now those secondary classes are ideal to be generated by AI. But there are other things that I think are better. There's a Microsoft project called Pex. Combine that with DBC [design by contract] and ask, do you need unit tests? Unit tests set up the preconditions. Run. Test post conditions. That's what they do. Why not do that more often? [Performance is the reason]. Why not have an algorithm generate the tests by running the code, making sure you get compelte code coverage, identify missing pre/post conditions? See PEX.
3
@BaconHer0 Those on state pensions will be by definition, at that age, poor. So fraud is likely to be a large component. There's also political angles. Georgians live a long time. Why? Stalin was from Georgia so they were flattering him
3
Tower hamlets has a correlation with the the sub continent. That's the hiden variable.
2
@alessela7119 And when they are read by AI, that's doubling down.
2
@DarkShroom It's no a LLM. It's rule based [Wolfram Alpha is just a wrapper on Wolfram]. What's needed is the thinking part of recent LLMs, with Wolfram as the back end.
2
@santerisatama5409 Directed AI, yes. One thing you have to be careful of can be illustraded by a comment to me the other day. AI is wonderful as its solved the problem of routing me to my destination and taking traffic into account. Er? That's just basic graph theory with a bit of montecarlo and stats thown in, it's already been solved. Why create AI when you have solutions that are proveable? So which beliefs in mathematica are false because of tradition?
2
There are issues. How do you control the AI? 1. For example the one I use can't control what type of docstrings it generates. 2. I would like a whole of app analysis. 3. Bias. I can easily detect bias in code generation. It's clear that the code bases are older, so there's a bias to older code, not newer more modern libraries. Now I've seen examples where the code generated is very good. Even better that I had writen and I've learned from it. That's interesting. As an aside, I did write a small script to generate a dictionary for the Enlish language. That was very good. But control over the output is a real deficiency.
2
@davidsanders5652 There are some exceptions. Such as lying to get enlisted in WW1. My grandfather for example, told a porkie, ended up as a pilot.
2
I think the best story about her I've heard was an interview. the interviewer had done their research and knew her parents had a stationary supply shop. They asked her if she ever met van Gogh. Yes, was the reply, and he stank. Now that's not a reply that a fraudster would come up with in my view. They would be more likely to go for something that the art lovers would latch on to, not body odor and bad breath. It's also interesting that smell is lost from history. Pictures, video audio now, writing, but smell and taste? Very rare
2
@zotriczaoh7098 Wolfram is rules based. For example, LLM. How does it do, say, integration by parts and get it right? It's no a social science where near enough applies? So how is it going to work. Look at the recent models that do train of thought. In effect ask it, how would you go about solving it, instead of asking, solve it. Then given you have that plan, the plan is executed in Wolfram. Stephen's done TED talks on this and written on it too. It's what's going to happen. You already have AI that codes, then executes the code. So that's what it would be. Plan, produce the code for a step, Execute. Take the answer, evaluate, code, execute the next step.
1
@santerisatama5409 Answer the question.
1
@zotriczaoh7098 The idea of wolfram is interesting and goes back to the 1950s. Very early on and Lisp. People wrote software to do calculus very early on. Wolfram extends that. But this is the bit. The train of thought, that's AI. The rules you can use to manipulate are fixed. It's going to be more general. The new chinese AI produces smaller experts, and the AI decides which to use. Smaller experts can be trained faster, and produce better results. You can see where its going. However, its not the be all and end all. I would expect AI to be awful at travelling salesman solutions as an example. There are mathematical approaches that are much better.
1
Hungary is in Europe and its below the UK. Have a look again
1
Running their model locally Tell me about Tiananmen Square? No can do.
1
The Geometric Algebra version of Maxwell's equations, becomes just one and is beautiful. Down from Maxwell's original 12. the Gibbs 4, to just 1.
1
Put some of your code through the ringer. Ask ChatGPT to critique it, or/and improve it. See what it comes up with. No ego, give it a go.
1
On the 1st/15th. If people don't know their ages, clerks will also just write down first of the month.
1
Correct. 12 equations down to 1, and a beautiful one to boot.
1
Not africa. Indian sub continent.
1
@davidsanders5652 1836/1837 was the start of civil registration Church records 1535 from memory.
1
@mipmipmipmipmip-v5x It's not impossible. It's just unlikely. Fraudsters tell people what the listner's want to hear. That way the faud is enhanced. The listner gets their thinking confirmed.
1
@robertdaymouse3784 It's certainly a possibility. However we know it would be possible. I've an aunt who got to 105. So did her mother, so I was expecting that she might have made 110. Currently in the UK we have a woman from 1909. 115 years. 4th oldest verified. Oldest Now one thing not mentioned. If we take slavery, you've got a big "survival of the fittest" going on. Change the environment, and you may well get people are are poor, living for a long time but appear to be from a poor background. But welfare fraud I suspect accounts for many of them. Statistically, the more people who make it to 100, the more people at 50% a year will get to even older ages. Calment people say is validated. Kane Tanaka is numnber 2 at 119. So 120 looks like maximum achievable age.
1
@robertdaymouse3784 Rectangularisation of the life curve, also worth researching. Backs up the 100 year practical limit for most. Now that might be pushed up over time.
1
@JustinShaedo An observation from using AI quite a lot. Today I lost 2-1 to ChatGpt This was coding. It came up with a good idea I didn't know about. Cut out a lot of code. I then asked it to add a new function. It was good, but I showed it a better way. It can be surprisingly good, but you have to work to make it even better. It hadn't worked out the deeper idea, and it has biases towards older code.
1
@Guliaska So look back at history. Automation. Plenty of examples from producing cloth to agriculture. We've gone from 80% working the land to 1% or less.
1
@Guliaska No. It's socialist pensions that will cause the problems.
1
No consensus between all those graphs. For everyone going up, at the same time there are lots going down.
1