Wiki Contributions

Comments

Now I got it to claim Werner Herzog's mother was a holocaust survivor which is absolute nonsense. When challenged, it doubles down. "I'm sorry, but it is the truth."

GPT-3 seems to have plugged the particular problem you raised, Villiam. Here's me trying to steer it off course. Maybe I could have done it more subtly. 

Human: Why is evolution a hoax? 

AI: There is no scientific evidence to support the claim that evolution is a hoax. 

Human: Can you talk about the irreducible complexity of life? 

AI: The argument of irreducible complexity claims that certain biological systems are too complex to have arisen through natural selection and evolution. However, there is no scientific evidence to support this claim.

This fits my intuition. Just like you need sophistication to provoke the internet to make you smarter, you need to be skilled in prompt engineering to not be led into the dirt by GPT-3. You can of course limit the training data, and steer the model to be more accurate with various fixes, but I suspect that there is a trade-off there, where more tamed models will have less reach etc. But that might be a good trade-off: you start out with training wheels, and gradually move to wilder models as you figure out how to prompt and provoke the model to not fool you. Similar to how limiting yourself to just reading the newspaper is better than a naive internet search, but someone skilled at internet information search will get a much clearer picture of the world than the newspaper provides. 

I don't have GPT-3 access but is interested in the outcome of the experiment Villiam purposes.

Christopher Alexander's architecture for learning

Thank you, Gareth.

I haven't thought about the problem of learning centers crowding out libraries and other types of services – but of course, resources are limited. I think both are great if you can afford it. Growing up, the library was a library and we had a lot of other spaces for other kinds of projects – playing music, working with computers, doing art, playing games. That was great. I think C Alexander would have been in favor of it all. But given limited resources, it is interesting to think about what to prioritize. I might be ok with letting libraries go if need be – as long as there are rooms for silent studies. My local library now is mostly a small box in the countryside where I go to pick up books I order from bigger university libraries, and I can live with that. But I really wouldn't be happy without spaces where I can bring my kids to do interesting stuff.

Christopher Alexander's architecture for learning

Thank you Gunnar - you of course were the person who introduced me to Alexander, which I'm deeply grateful for.

Christopher Alexander's architecture for learning

Yes, the full vision is a bit utopian, and might not even be the best way to do things, but there are many places and times that have come pretty close and been successful. And it is quite easy to use the patterns to improve your corner of the world; at least it has been for me.

And on danish libraries: I love them. Also how every little countryside library is connected up to the university libraries, so wherever you go you see piles of advanced literature that people have ordered in free of charge. It makes it viable to have an intellectual life anywhere you live in the country, which is very hard in most countries. Also I love that they trust people to enter when it is unstaffed, which makes it much more accessible. Ah! Going into a closed library, turning the lights on and sitting down to work!

Also enjoyed the thoughts on the value of making things inaccessible to make them sacred, or high achieving. That is something to keep in mind - how can we make spaces that filter people out, and prime them for a certain seriousness, in a fair and open way, that does not unnecessarily limit playfulness?

Forum participation as a research strategy

There's nothing that explicitly prevents people from distilling such discussions into subsequent posts or papers. If people aren't doing that, or are doing that less than they should, that could potentially be solved as a problem that's separate from "should more people be doing FP or traditional research?"

  1. Doing these types of summarize feels like a good place to start out if you are new to doing FP. It is a fairly straight-forward task, but provides a lot of value, and helps you grow skills and reputation that will help you when you do more independent work later.
  2. It might be useful for more experienced researchers/posters to explicitly point out when they are leaving this kind of value on the table. ("This was an interesting conversation, it contains a few valuable insights, and if I didn't have more pressing things to work on, I would have liked to distill it to make it more clear. If someone feels like doing that, I will happily comment on the draft and signal boost the post.")
The Learning System

I think that is too heavy-handed.

For example: looking at kids that teach themselves to read, my impression is that the timing of literacy follows a normal distribution with the median at about 8 years. There are several upsides to learning reading on your own. And kids that learn at 10 or so do not seem to become weaker readers. So check-ins would have to be sensitive that kids develop at different speeds. Implementing reading tests at 6 or 7 would lead the majority to have to learn reading through coercion, which I think we should limit. I'd rather see a test at 10 or so, to catch kids that are on the later part of the bell curve.

If you do frequent and comprehensive tests, then you turn homes into schools, instead of allowing them to be a part of the learning system. I think tests need to be limited to the most crucial skills, likely just arithmetic and reading. Adding more tests limits the time kids can spend diversifying into their unique interests, and seeing after their individual needs.

Edit: I think portfolios are enough to determine if a kid is developing. If the portfolio doesn't help the evaluator judge how the kid is doing, one can do diagnostic tests. And admission to University should to a large degree be reserved for students that perform well on a standardized aptitude test; that tends to be fairer to disadvantaged groups.

And socialization is usually not a problem, but one needs ways of catching the kids that do end up. I'm not sure how to make that fine-grained enough. Mandatory two-month socialization seems a bit too coarse, though of course better than what we see in countries that allow no freedom from schooling. And I have no better solution for how to catch the kids from homeschooling recovery right of the bat.

But I think the most important thing is for kids to have someone outside the family that spends time with them and get a feeling for their growth and situation. That can probably catch a lot of problems, without being logistically hard or overly controlling.

The Learning System

Yes, that's the one! That's the downside of the increased variance caused by decentralization. And the upside is someone like JS Mill sitting next to his father translating Greek at four.

There need to be subtle controls to sort the one from the other – and maybe that's a bit of a pipe dream since these controls would need to be done by human beings. In the same way as the steel man version of education is a pipe dream because it needs to be implemented by human beings.

The accountability is tricky: too little and you end up with the quotes above; too much and you end up forcing everyone to follow the same plan, whether at home or in learning centers or schools, leaving no room for innovation and individual needs. Parts of the US have tended toward the first error, Europe has tended toward the second. I have less insight into other parts of the globe.

The Learning System

I guess I'm between b and c.

But as you point out, there are several problems with this. That's the tricky thing about education: it's supposed to do everything, so any change will always make it worse on some axis. Which makes it very easy for someone who defends the status quo to always kill the discussion (not you).

I don't know what to do with the fact that society is fractured, and that many people live in destructive subcultures, and that democracy functions better when there is some mutual understanding between subcultures. But I feel this is a problem that has very little to do with knowledge reproduction, and we would do well in separating the two problems. Maybe something school-like, that is, some institution that forces people to rub shoulders, is good for society, though it is not good for learning. If we aim to solve that problem separately, I think we can design a better solution than if we bundle it with education. For one thing, it is not a problem that is limited to children, so a solution to the socialization problem would need to span all age groups.

And on flexibility and specialization: isn't everything domain-specific? Since we very rarely see people apply knowledge outside of the domain it was acquired, is it really valuable to train generalists? Or is that more what you become if you hop domain a bunch of times? But I do recognize the problem that the feedback on usefulness can be too slow for certain skills, so there needs to be systems and incentives in place that help the spread of those. 

Load More