mm321

> I can be confident that a Perl script I write today will run unaltered 10 years from now, modulo external collaborators.

Nothing is eternal. Some dependencies, like CPAN modules, need to build stuff in C (during installation). It is impossible to know if these dependencies will install and build 'unaltered 10 years from now'.

Is it possible that reddit tried to pull of the same monetization trick as twitter, however they forgot to check with the target audience. And that public is really stretched financially, it does not consist of establishment figures who have an extra five dollars to spare for their hobby.

But how is it possible that the reddit bosses don't know who their users are?

it was relatively low, for quite some time.

> Luckily Xiaoping entered the chat and the rest is Chinese history.

I think Deng Xiaoping is a direct result of these developments during the Mao years.

Things would have been very different for China, without the rapprochement with the west. Without friendly terms with the west they would have to maintain an enormously big army.

Think of a China today that is more like North Korea.

the war in Ukraine is not funny at all, it has all the potential to turn into WW3.

Depends on what Xi is planning over Taiwan, that is where things get complicated.

They are all trying to build alliances for that contingency.

both China and the Soviet Union were supporting North Vietnam at the same time - up until 1968 (when they had this war over an island). The Sino Soviet split happened much earlier, in 1961.

https://en.wikipedia.org/wiki/China_in_the_Vietnam_War

The great friendship between China and the USA started after that border conflict in 1968, as China was looking for a counterweight.

https://en.wikipedia.org/wiki/Sino-Soviet_border_conflict

I mean Mao was playing Nixon and Kissinger against Brezhnev.

The funny part is where Kissinger is taking all of the credit for the US China rapprochement.

Looks like some game of chess: China stopped supporting North Vietnam, but that didn't stop North Vietnam from invading South Vietnam, after the pullout. However at that point no one in the US cared about any domino theory, because the US was now friends with China against the Soviet Union.

I wonder what the chess pieces were all thinking...

That is possible, strange things happen at sea.

Now in the case that they crashed due to a bug, then the programming profession is safe from program verification and AI, I like that option!

I agree. I mean lets assume it is true: how could the US government have failed a real alien ship? Did the poor aliens fail at filling out some paperwork? I mean this ship had to spend so much energy just to get to us, what else could have failed them?

that one is 150 million kilometers away, and we have an atmosphere for our protection.

The effect/kinetic energy of the water involved would be comparable to that released by a nuclear explosion.

Total madness.

https://www.theguardian.com/world/2023/jun/06/nova-kakhovka-...

"The reservoir it contains holds an estimated 18 cubic kilometres of water, about the same volume as the Great Salt Lake in Utah. Bursting the dam could send a wall of water flooding settlements below it, including Kherson, which Ukrainian forces recaptured in late 2022"

In 1941 the retreating Red Army blew a different dam, the Dnieper/Dnipro dam, up to a hundred thousand people died on that occasion...

https://en.wikipedia.org/wiki/Dnieper_Hydroelectric_Station#...

Still it sounds funny that they would need to employ the CIA in order to finance the Museum of Modern art in... New York, doesn't it?

The USA had the "Congress for Cultural Freedom", the Soviet Union had it's "World Peace Council", funny kind of influence peddling.

wikipedia has a list of "Communist front organizations" https://en.wikipedia.org/wiki/Category:Communist_front_organ...

and one for "CIA front organizations" https://en.wikipedia.org/wiki/Category:Central_Intelligence_...

And nowadays the spooks are all busy with trolling on twitter (the taxpayers money is doing some real work :-) )

wasn't René Magritte and "Ceci n'est pas une pipe" long before there was a CIA? https://en.wikipedia.org/wiki/The_Treachery_of_Images

Also Picasso was a real commie https://en.wikipedia.org/wiki/Pablo_Picasso#Political_views - also Diego Rivera, and Frida Kahlo.

There could have been more bank failures, given these figures ? I mean that's quite of lot of failing credit...

maybe he was rationalizing his own fate, like Rubashov in Koestlers "Darkness at noon".

A common response by those arrested during Stalin's purges is said to have been "this is a all a mistake, it will be sorted out soon"...

very impressive!

Just one questions: what about older versions of linux that don't have io_uring, does it fall back gracefully to older system calls or are these older versions of linux no longer supported?

guess everything depends on the context, however the context can be misunderstood.

The best way is to take a break, this makes it easier to evaluate the information in it's proper context.

(joking) maybe they fed the LLM some postmodern text, so it got some notion of relativism and post structuralism...

But no, LLM's make things up, and it's a known problem and it is called 'hallucination'. even wikipedia says so: https://en.wikipedia.org/wiki/Hallucination_(artificial_inte...

The machine currently does not have it's own model of reality to check against, it is just a statistical process that is predicting the most likely next word, errors creep in and it goes astray (which happens a lot)

Interesting that researchers are working to correct the problem: see interviews with Yoshua Bengio https://www.youtube.com/watch?v=I5xsDMJMdwo and Yann LeCun https://www.youtube.com/watch?v=mBjPyte2ZZo

Interesting that both scientist are speaking about machine learning based models for this verification process. Now these are also statistical processes, therefore errors may also creep in with this approach...

Amusing analogy: the Androids in "Do Androids dream of electric sheep" by Philip K Dick also make things up, just like an LLM. The book calls this "false memories"

If someone doesn't take it personal, isn't that a sign that he/she doesn't care too much?

Speaking for myself: I wouldn't desire to be disinterested in my work.

i found a detailed article on Drakon in Russian [1], they say that development of the Drakon language started in 1986 (that would have been two years before the launch) and that work on the ГРАФИТ-ФЛОКС dev environment was completed in... 1996.

it was used with the Sea Launch program, Fregat - that's a space tug for Soyuz/Zenit, the modernized Proton-M, ДМ-SL-Б - that's a space tug used with the Zenit launcher.

However the article is a bit old, from 2009. It could have been used for later systems.

https://old.computerra.ru/readitorial/418507/

Wikipedia mentions hybrid variants of Drakon - like Drakon-C, here the they would put C code into each of the boxes of the flowchart (each square box would stand for a basic block, and each diamond box a boolean expression?). I think that would make more sense, this way you could actually write some more complex programs and the flowchart would not blow up in complexity.

https://ru.wikipedia.org/wiki/%D0%94%D0%A0%D0%90%D0%9A%D0%9E...

i guess that with Drakon-C a function call would be just a line in the basic block, would make more sense.

and that he knows a thing or two about pointer arithmetic.

still C and C++ have a totally different approach to program design: A C project of non trivial size will have lots of insane tricks with macros and callback functions, whereas C++ has entirely different methods of abstraction.

I think that's a very significant difference.

Also C++ has this std::unique_ptr<> thing that helps a lot with preventing memory leaks. And C++ has stl, so you won't have to write your own list/array/map abstractions.

Can someone please explain the "Scaled dot-product attention" function from the article?

Attention(Q, K, V ) = softmax( (Q * TRANSPOSED(T)) / sqrt(Dk) ) V

That's where i start to shake my head.

Denny Zhou is the founder and lead of the reasoning team @Google Brain. https://twitter.com/denny_zhou

EV battery is degrading over time - a replacement costs ... between 5000-20.000$ [1]. I am not sure that we have objective data on the scale of this issue.

[1] https://www.cars.com/articles/how-much-do-electric-car-batte...

Do you know the current system prompt for bing? I asked that as a direct question, and got the following response: 'I’m sorry but I prefer not to continue this conversation. I’m still learning so I appreciate your understanding and patience'

Data denormalization makes sense if the data is written once and never updated - like with a data warehouse / analytics.

If you need to update the data then denormalization can turn into a big source of trouble. For example you end up with many copies of the same stuff, and you must make an extra effort to update all the duplicates upon update.

Or you end up with multiple entries, where the validity of an entry is determined by some extra 'isValid == true' or 'deleted == false' field. Now all these 'invalid' entries then start to clog up the table/collection, and performance may quitely deteriorate.

I once had to use a denormalized schema for nested data, as lookup through too many reference would have suffered. But that wasn't funny at all.

I don't understand this whole business of page long prompts, the following article says that this kind of short LLM prompts work the best - and they actually measured the performance of the different kinds of prompts! https://arxiv.org/pdf/2212.04037.pdf

"we devise the following straightforward procedure:

1. Obtain a small set of manually created prompts for the task.

2. Expand the set of prompts with automatic paraphrasing using a LM (e.g., GPT3) and backtranslation (see Section 3).

3. Rank the list of prompts by perplexity (aver- aged on a representative sample of task inputs, e.g. 1,000).

4. Choose the k (e.g., 3) lowest perplexity prompts.

Using this algorithm, we show empirically that it is best to prioritize experimenting with the lowest perplexity prompts, as they perform better than manual prompts on average, and are more stable"

How is the automatic paraphrasing of the prompts achieved?

* first they paraphrase the hand written prompts: "We prompt it with a meta-prompt for paraphrasing to generate variations of one of our seed prompts. An example of such a meta-prompt is: Write a paraphrase for the following sentence: <seed prompt> Paraphrase:. The 7 meta-prompts used in this step are listed in Section A in the Appendix. We choose GPT3 as our paraphrasing model because of its well-documented generation abilities. This is also to ensure that there is a separation between the model we use to create the prompts and the models we use to rank them (OPT and Bloom, see Section 4 for details)"

* then they translate the text from the previous step to a different language and back into english.

I guess this process should create a prompt that is close to the relevant text from the training set - and that such a prompt would perform best.

from the article: "we devise the following straightforward procedure:

1. Obtain a small set of manually created prompts for the task.

2. Expand the set of prompts with automatic paraphrasing using a LM (e.g., GPT3) and backtranslation (see Section 3).

3. Rank the list of prompts by perplexity (aver- aged on a representative sample of task inputs, e.g. 1,000).

4. Choose the k (e.g., 3) lowest perplexity prompts.

Using this algorithm, we show empirically that it is best to prioritize experimenting with the lowest perplexity prompts, as they perform better than manual prompts on average, and are more stable"

the Androids in "Do Androids dream of electric sheep" by Philip K Dick also make things up, just like an LLM.

The book calls this "false memories", where we call it "hallucinations"...